Sage Journals: Discover world-class research

Abstract

Researchers are working hard to handle the intricate interplay and interconnected relationships between fiber properties, process parameters, and consumer requirements in yarn manufacturing. These relationships make a complex network and researchers tried to develop prediction models through artificial neural network (ANN). This paper provides a systematic and comprehensive up to date (2024) overview on applications of ANN in the textile yarn manufacturing (fiber to yarn) sector. In addition, diverse methodologies, approaches, and specific applications of ANN in fiber to yarn are criticized in this study. The limitations, challenges, and future scopes of this subject have been explored by synthesizing the selected relevant publications in the field of textile yarn manufacture and ANNs. The study contributes to the theoretical and practical implications through this review and concludes with a valuable future opportunity. Notwithstanding the fact that the research shows a clear improvement in the implementation of ANN in the yarn manufacturing sector, substantial work remains to be done in these areas.

Keywords

Artificial Neural Network back propagation yarn manufacturing predictive model fiber properties yarn quality

Introduction

Nowadays, people typically satisfy their second fundamental need—that of textiles and clothing—without considering the consequences. Globalization has facilitated the manufacturing of textile and clothing products in poor nations by taking advantage of low labor costs and relaxed environmental and working standards regulations.¹ Textile manufacturing is dynamic, evolving in conjunction with the state of the art in science and engineering.² Yarn production is an essential initial phase in the worldwide textile industry when the yarn is made from fibers and/or filaments. The characteristics of yarn are a crucial factor in creating appealing and essential clothing.³ Ring spinning is widely regarded as the most prominent spinning process when compared to other spinning systems including rotor, vortex, and friction spinning. This is because it produces superior-quality yarn and utilizes highly adaptable technology.⁴ The yarns undergo additional preparation stages such as winding, doubling, twisting, and waxing, depending on the desired clothing type.⁵

The structure and physical characteristics of yarns are mainly determined by the characteristics of the fibers used, the spinning technique employed (such as ring, rotor, and vortex spinning), and other process variables including draft, twist-insertion rate, rotor speed, nozzle pressure, etc.⁶ The intricate interplay between fiber properties, process parameters, yarn structures that are raised to a complex network of interconnected relationships, necessitating the development of prediction models.⁷

The characteristics of fiber, sliver, and yarn can be forecasted using the statistical and experimental framework of artificial neural networks (ANNs). While ANN is more robust and capable of self-validation and instruction in complicated, multiparameter, and nonlinear scenarios those do not have easily identifiable or direct analytical solutions.^8–12

Pareira et al.¹³ explored the application of Fuzzy logic, ANN, and expert systems as tools of artificial intelligence (AI) in textile industries. Chattopadhyay¹⁴ explained the ANN model in detail and predicted the yarn characteristics from fiber properties as input materials. A number of reviews by scholars such as Sikka et al.,¹⁵ He et al.,¹⁶ Mukhopadhyay and Suddiquee,¹⁷ Chattopadhyay and Guha¹⁸ highlighted the applications of AI tools specially ANN in various operations from fiber to clothing in textile industries.

In the textile and clothing sectors, there is a complex relationship of numerous elements. Due to significant fluctuations in raw materials characteristics, complex processing stages, and limited control on process parameters, the correlation between these variables and yarn properties depends on human expertise.¹⁹ However, it is impractical for individuals to retain all the specific details of process-related data over time.²⁰ The majority of yarn manufacturing sections are transitioning from traditional to automated and sophisticated systems that utilize cutting-edge technology.²¹ The testing results on various parameters of fibers and yarns takes long time which awaits the yarn manufacturers for long time and disrupts the continuity of production flow. The yarn manufacturers are seeking faster evaluation of fibers, intermediate materials, and yarns.²² With the significant enhancement of computational capacity over the preceding decade, ANN can effectively learn from datasets to uncover the unknown relationships among the various variables.²⁰

Over the last decade, substantial advances have been made in the use of ANN in textile processing, which contributes to sustainability in the clothing industry. Nevertheless, research has been somewhat divided, with ANN concentrating exclusively on material advancements while overlooking the process prediction. For the convenient of academicians and professionals of yarn manufacturing, the study aims to make a comprehensive and systematic literature review on the scope of ANN in yarn manufacturing. This review investigates the scholars’ findings and application of ANN on fibers, slivers, roving, and yarn that are manufactured through the utilization of contemporary spinning techniques like ring, rotor, and air-jet. This paper also provides a critical analysis of the scopes, limitations and challenges, along with the promise of employing intelligent techniques on the yarn manufacturing process, to establish a solid foundation for future scholarly investigations.

The present study contributes to the current body of literature by investigating the utilization of ANN within the scope yarn manufacturing industry.

This research investigates the existing ANN application on attributes of fibers, slivers, roving, and yarn that are manufactured through the utilization of ring, rotor, and air-jet spinning techniques. Additionally, identify future research directions for forecasting the characteristics of materials in process at each stage of the short-staple yarn manufacturing industry.

For the first time, the current study presents a novel approach to conducting a systematic and thorough assessment of the utilization of ANN in the yarn production industry, focusing specifically on scholarly publications sourced from Scopus.

Paper selection and review process

This evaluation examines scholarly articles that are included in the Scopus database. The Scopus search engine was employed to search for the precise keyword ‘Artificial Neural Network and fiber and yarn’ with restrictions to the English language on January 8, 2024. The search time for the articles’ spans from 1996 to 2024. A total of 150 documents were identified, with 79.3% being journal articles and 13.3% being conference papers. The categorization of the documents is depicted in Figure 1. The materials are additionally classified according to their subject area and shown in Figure 2. The documents consist of 21.20% papers related to engineering and 30.06% about material sciences. Following a series of consecutive sorts, a total of 67 articles were selected for review. Only the field study about the process of converting fiber into yarn has been taken into account while excluding any reviews or studies that are not relevant to the yarn manufacturing technology. Figure 3 illustrates the process of selecting articles.

Figure 1.

Documents by type.

Figure 2.

Documents by subject area.

Figure 3.

Paper selection outline for the review.

An overview of Artificial Neural Networks (ANNs)

The ANNs are statistical learning models used in cognitive science and machine learning. These models mimic biological neural networks (BNN). Its algorithms are designed to achieve a certain purpose, and it learns and stores knowledge using synaptic weights like the human brain.¹³ Synaptic weights, numerical values provided to neuron connections, help neural networks learn and adapt to new inputs. The network’s experience can change these weights. ANN cannot solve the problem alone; it must be integrated into a system engineering technique.²³ Neural networks are powerful because of their parallel distribution structure and capacity to learn and generalize. Generalizations allow neural networks to operate well with untrained inputs. ANNs have a unique neuron connection ‘architecture’. Connection weights are calculated using an activation function and ‘training’, ‘learning’, or ‘algorithm’.^24,25 However, it has multiple drawbacks, such as direct convergence to local minima, slow convergence, or lack of convergence, overfitting, and an ambiguous optimal number of hidden network layers.^26,27

Network selection

A variety of function approximation problems have been solved using feed-forward neural networks. By adding units to the input, hidden, and output layers, the neural network can be constructed as Figures 4 and 5. The network must be trained using back-propagation on the specified data sets. This requires adjusting network parameters like the number of hidden layers, the number of units inside each layer, the activation function, the learning rate, and the training cycles.¹⁴

Figure 4.

Artificial neural network without hidden layer.¹⁴

Figure 5.

Artificial neural network with hidden layer.¹⁴

Optimizing network parameters

Hidden layers exponentially increase processing time. Thus, single-hidden-layer neural networks that mimic yarn properties might be better. The number of hidden units greatly affects model success.^28,29 An excessive number of units might prolong training and cause overfitting. Exercises with each dataset will determine the hidden layer’s unit count. Training of a network with a learning rate of 0.1 is simple and straightforward process, and network error drops rapidly and eventually minimize the errors.¹⁴

Data and performance estimation

Artificial Neural Networks (ANN) utilize the backpropagation technique, which involves a variable learning rate and numerous linear regressions. These methods have demonstrated efficacy in the field of fiber grading, sliver roving, and yarn quality prediction in the spinning industry.¹⁵ The evaluation of models’ performance is commonly measured using many metrics, such as the correlation coefficient (R), R-square (R²), Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Error (ME), and Relative Error (RE).^16,30 These are measured with following equations (1) to (7). A model can be evaluated with a high degree of accuracy when the RMSE is at the minimum and R² is high (i.e. ⩾0.8)³¹

R (M, P) = \frac{\sum_{i = 1}^{n} (M_{i} - \bar{M}) (P_{i} - \bar{P})}{\sqrt{\sum_{i = 1}^{n} {(e_{i} - \bar{e})}^{2} . \sum_{i = 1}^{n} {(p_{i} - \bar{p})}^{2}}}

(1)

R^{2} = {(\frac{\sum_{i = 1}^{n} (M_{i} - \bar{M}) (P_{i} - \bar{P})}{\sqrt{\sum_{i = 1}^{n} {(M_{i} - \bar{M})}^{2} . \sum_{i = 1}^{n} {(P_{i} - \bar{P})}^{2}}})}^{2}

(2)

M A E = \frac{1}{n} \sum_{i = 1}^{n} | M_{i} - P_{i} |

(3)

M S E = \frac{1}{n} \sum_{i = 1}^{n} {(M_{i} - P_{i})}^{2}

(4)

R M S E = \sqrt{\frac{1}{n}} \sum_{i = 1}^{n} {(M_{i} - P_{i})}^{2}

(5)

M E = \frac{1}{n} \sum_{i = 1}^{n} \begin{matrix} (M_{i} - P_{i}) X 100 % \end{matrix}

(6)

R E = \frac{1}{n} \sum_{i = 1}^{n} \frac{M_{i} - P_{i}}{M_{i}} X 100 %

(7)

Here M_i, P_i, and $e_{i}$ are the measured, predicted, and error output of the model respectively.^16,32

ANN in the yarn manufacturing industry

Desired quality of products in the yarn manufacturing is assured through the evaluation of various physical and mechanical properties of components such as fibers, slivers, rovings, and yarns. Generally, this procedure is complex, time consuming, and necessitates the involvement of proficient operatives and the application of expensive machinery.^20,33 ANN research aims to reveal this complex functional relationship between fiber and yarn structure. Identifying yarn traits allows for the optimization of qualities for different usage and lowering costs as well.³⁴ In the mid-1990s, researchers used ANN, a computer program that learns mathematical correlations between yarn and fiber properties, to predict yarn qualities from fiber parameters.³⁵ Management may perceive a neural network trained in this manner as a valuable instrument for decision-making.¹⁸ The degree of uniformity in the yarn significantly impacts the quality of the textile, as well as its effect on subsequent processes and the appearance of the fabric.¹⁹ Forecasting yarn quality measurements is consistently a prominent area of study for textile experts.³⁶ Yarn property modeling is a fascinating area of research in the realm of textile engineering.³⁷

Prediction in fiber properties

In the textile and clothing industries, raw material qualities determine product quality and market value.^38–41 Every spinner encounter difficulty in manufacturing high-quality yarn while minimizing expenses. The machine-learning technology can predict yarn quality in the cotton spinning trade, replacing human skill.⁴² An overview of the application of ANN in the prediction of correlation between fiber and yarn characteristics has been shown in Table 1.

Table 1.

ANN in the prediction on fiber properties and correlation between fiber and yarn.

Sl.	Referrences	Year	Material used	Technique	Aims
1	Das and Chakraborty⁴³	2021	Cotton fiber properties.	Adaptive Neuro-Fuzzy Inference System (ANFIS) along with Regression, Fuzzy logic, ANN, and Support Vector Machine (SVM).	To investigate the relationship between cotton fiber properties and yarn characteristics.
2	Farooq et al.⁴⁴	2018	Cotton fiber length, uniformity index, strength, and linear density.	Back propagation ANN	To predict the maturity of cotton fiber based on other fiber properties.
3	Majumdar⁴⁵	2006	Yarn evenness, hairiness, tenacity, and elongation.	Back propagation ANN	To choose cotton fiber by creating a reverse model that compares the attributes of yarn and fiber.
4	Jayadeva et al.⁴⁶	2003	Yam properties, namely, Count Strength Product (CSP), unevenness, and imperfections.	Feed forward ANN	To assess the relative significance of fiber qualities.

Das and Chakraborty⁴³ developed an ANFIS to make a relationship between fiber and yarn characteristics. This system relies on six input variables of cotton fibers, including strength, fineness, elongation, length, uniformity, and short fiber content. Their prediction ability was thoroughly validated based on the estimated values of R, Mean Absolute Percentage Error (MAPE), RMSE, ME, and Variance Performance Index (VPI). The investigation asserted that the ANFIS model incorporated the beneficial characteristics of both the fuzzy control system and the neural network. Farooq et al.⁴⁴ developed a predictive algorithm to estimate cotton fiber maturity. ANN inputs were fiber length, strength, uniformity index, and linear density, and output was fiber maturity. The network’s projected values were compared to the tested values to get the MAE. Training networks were validated using 10% cross-validation. The MAE of 0.0491 between real and projected values indicated high neural network modeling accuracy. Majumdar et al.⁴⁵ demonstrated the feasibility of yarn engineering by creating a model that can convert yarn into fibers using ANN. The yarn’s tenacity, breaking elongation, unevenness, and hairiness index served as the input parameters in the model. The observed MAE of prediction for Spinning Consistency Index (SCI) and micronaire were 4.38% and 1.92%, respectively. Jayadeva et al.⁴⁶ investigated the potential application of neural networks in assessing the significance of fiber qualities concerning yarn properties, specifically lea strength in Count Strength Product (CSP), unevenness, and defects. The relative significance of input parameters was assessed using two distinct methods: ‘skeletonization’ for trimming neural networks and a newly constructed first-order sensitivity analysis. The sensitivity analysis was determined to be highly effective.

The physical characteristics of fiber, including maturity, SCI, and linear density, have been estimated from other metrics obtained through High Volume Instrument (HVI). The essential characteristics of strength, trash content, and color grade remain unexamined. The existing approach concentrates exclusively on a single aspect of raw materials in the yarn manufacturing sector: the challenges associated with fiber production, comprising both natural and synthetic fibers, including ginning process and fiber baling till retention. Majumdar⁴⁵ presented a model to predict the yarn properties from fiber properties and also showed a reverse model from desired yarn properties to required fiber properties. This study presented the scope of ANN in prediction of output characteristics based on input and vice versa.

Prediction in blending, mixing, and sliver properties

The uniformity of yarn has a growing importance in the textile sector, with sliver evenness being a crucial aspect in ensuring the production of high-quality yarn.⁴⁷ To widen the product mix and minimize the product costs while ensuring desired quality, manufacturers are practicing blending and mixing of various fibers. An overview of the application of ANN in the prediction of correlation between fibers, intermediates, and yarn characteristics has been shown in Table 2.

Table 2.

Application of ANN in Prediction on blending, mixing, and sliver properties.

Sl.	Authors	Year	Material used	Technique	Aims
1	Yang et al.⁴⁸	2017	Spectral reflectance of colored fibers.	Modular Artificial Neural Network (MANN)	To develop a technique for forecasting the recipe of mélange yarn.
2	Farooq⁴⁹	2014	Sliver doubling, Labelling Action Point (LAP), Labelling Intensity (LI) of draw frame.	Supervised multi-layer feed-forward ANN	To investigate the influence the draw frame variables on the fiber cohesion in the drafted slivers.
3	Abd-Ellatif⁵⁰	2013	Fiber characteristics, sliver fineness, drafting roller gauge, and CV of sliver.	Back propagation ANN and Multiple Linear Regression	To predict the LAP in finisher, draw frame.
4	Farooq and Cherif⁵¹	2012	Draft, doubling, and delivery speed of draw frame.	Feed forward ANN	To determine the relationships between draw frame, sliver, and yarn quality.

Yang et al.⁴⁸ produced mélange yarn from blended pre-dyed fiber and used ANN and Modular ANN (MANN) to forecast recipes better. The study used 500 training and 100 blind test samples of 19.4 tex ring spun dyed yarns for 100% cotton knitted apparel. The 60 dye formulations produced fibers with 1.85–2.32 dtex fineness. The 500-tex of rovings were drafted 30 times in the ring frame during yarn manufacture. Correlation Coefficient (CC), RMSE, training execution time (T), and Color Measurement Committee color difference assessed trained networks. MANN beats ANN in correlation, error, and training time. For 100 testing samples, the MANN model’s projected spectrum’s average CMC was 1.26, significantly higher than in practice (0.60), yet it still works well for recognizing pre-dyed fibers’ basic structures. Farooq⁴⁹ used draw frame features to train an ANN model to predict sliver cohesion. Cotton, polyester, and a cotton-polyester blend were studied. The materials were processed with draw frame variables. The study showed that ANN can accurately estimate cohesive force in drafted slivers. Abd-Ellatif⁵⁰ predicted the appropriate LAP for production and material conditions using ANN. In the blow room, extra-long, long, and medium-long staple Egyptian cotton was collected. The most critical elements were considered when creating ANNs. The inputs were fiber length, homogeneity, fineness, sliver liner density, evenness, and main draft gauge. They developed five statistical models to determine the optimal LAP value using the most critical fiber and process characteristics. ANN model performance was compared to statistics models. Farooq and Cherif⁵¹ determined the draw frame variables’ effects on sliver and yarn quality. Training neural networks with the Levenberg-Marquardt method and Bayesian regularization was hybrid. This method improved network generalization. Performance was assessed by cross-validating each training network. Finally, they developed a sliver and yarn quality predicting system.

The notion of auto-leveling in the finisher draw frame and the adjustment of LAP and LI are critical factors that influence the properties of the drawn sliver and the produced yarn. Nonetheless, the draft concept and the frequency of doubling in draw frames and comber machines, along with technical parameters such as feed per nip, nips per minute, top comb penetration, and cylinder comb settings, have been constrained research in predicting the characteristics of materials processed in the short staple fiber spinning industry. The research on ANN application for in-process-materials along with various technical parameters are still demanding.

Prediction in ring-spun yarn characteristics

The future applications of yarns are significantly impacted by the quality of their manufacturing process.⁵² Precise yarn quality prediction is essential in spinning mills to maintain consistent production, ensure superior product quality, and efficiently control costs.⁵³ The ring-spinning technology employs a variety of yarn types, including core-spun compact, solo, and siro, mélange, and slub, each of which is regulated by a unique set of parameters.⁵⁴ Roving properties, process parameters of ring frame such draft, twist, twist multipliers, spacer and size, traveler weight, and spindle speeds are the main factors that determine ring spun yarn quality.^55,56 An overview of the Application of ANN in the prediction of ring-spun yarn characteristics has been shown in Table 3.

Table 3.

Application of ANN in prediction of ring-spun yarn characteristics.

Sl.	Authors	Year	Material used	Technique	Aims
1	Ghanmi et al.⁵⁷	2023	Cotton fiber properties, yarn count, and twist per inch.	Back-propagation ANN and Fuzzy expert system	To utilize hybrid models to accurately and precisely predict the global quality index of yarn.
2	Abd-Elhamied et al.⁵⁸	2022	Fiber type, yarn count, and twist factor.	Image processing with Back-propagation ANN	To estimate the tenacity, elongation, Coefficient of mass variation% (CVm%), imperfections of ring-spun and compact cotton yarns.
3	El-Geiheini et al.⁵⁹	2020	Hundred percent cotton yarn and polyester/cotton (50/50) blended yarn.	Image processing with Back-propagation ANN	To utilize machine vision and learning approaches to accurately forecast the elongation and tenacity of different yarn kinds.
4	Doran and Sahin⁶⁰	2019	Fiber properties, process parameters like yarn fineness, the delivery speed of ring frame in m/min etc.	Back-propagation ANN and SVM	To construct predictive models for the intrinsic characteristics of cotton/elastane core yarn.
5	Kavitha et al.⁶¹	2019	Spindle speed, top roller pressure, traveler mass, temperature, current and voltage.	Feed-forward ANN and Principal Component Analysis (PCA)	To diagnosis the ring-spun yarn faults.
6	Ahmad¹²	2016	Fiber properties, yarn count and twist.	Feed-forward ANN and Graphical User Interface (GUI)	To predict the yarn strength.
7	Ghanmi⁶²	2015	Fiber properties, yarn count, and yarn twist.	Feed-forward ANN and Fuzzy expert system	Ring-spun yarn worldwide quality index prediction.
8	Unal and Ozdil⁶³	2015	Fineness, yarn fineness and twist.	Multilayer feed-forward ANN and Regression	To analyze yarn diameter of cotton ring-spun yarn.
9	Almetwally et al.⁶⁴	2014	Linear density of cotton fiber and spandex filament, drawing ratio, and twist multiplier.	Feed-forward ANN and Regression	To forecast the tensile characteristics of cotton/spandex core-spun yarns.
10	Mozafary and Payvandy⁶⁵	2014	Wool and polyester fiber properties and process parameters such as blend ratio, doubling, oil, drawing, RH, and temperature.	ANN and data-mining methods containing clustering	To predict the quality of worsted spun yarn.
11	Das et al.⁶⁶	2013	Cotton fiber properties.	Hybrid model of Back propagation ANN and Genetic Algorithm	To develop a equation for prediction of cotton yarn unevenness.
12	Ghosh et al.⁶⁷	2013	Fiber properties and yarn strength.	Back propagation ANN, Generic Algorithm and Non-dominated Sorting Genetic Algorithm II (NSGA-II)	To maximize of cotton yarn strength and minimize of raw material quality.
13	Bo⁶⁸	2012	Front roller speed, spindle speed, nip gauge, break draft, and yarn twist.	Back propagation ANN and Multiple Regression	To the predict the end breakage rates of cotton ring spinning yarn.
14	Soltani et al.⁶⁹	2012	Twist multiplier, hairiness, unevenness, and yarn tenacity.	Back propagation ANN	To create ANN models for forecasting the migratory patterns of fibers in cotton yarn.
15	Bo⁷⁰	2011	Front roller speed, spindle speed, nip gauge, break draft, and roving twist.	Back propagation ANN	To predict the end breakage rate of ring spun polyester/cotton yarn.
16	Admuthe and Apte⁷¹	2010	Cotton fiber properties.	Back propagation ANN and ANFIS	To predict the yarn properties.
17	Bo⁷²	2010	Roving twist, spindle speed, nip gauge, front roller speed.	Back propagation ANN and Regression	To predict the hairiness of polyester/viscose/cotton yarn.
18	Furferi and Gelli⁷³	2010	Fiber properties, percentage of fiber in blended roving, and yarn count.	Back propagation ANN	To predict yarn strength.
19	Huijun and Xinhou⁷⁴	2011	Rotor diameter, rotor speed, navel type, opener roller speed.	Back propagation ANN	To forecast the tensile strength and variation in mass of cotton yarns spun using rotor technology.
20	Majumdar⁷⁵	2010	Rotor diameter, navel type, rotor speed, opener roller speed and yarn count.	Back propagation ANN and Neuro Fuzzy System	To predict the cotton rotor-spun yarn hairiness.
21	Admuthe and Apte⁷⁶	2009	Yarn properties by customer demand.	Hybrid ANN and Genetic Algorithm,	To simulate and predict the fiber-to-yarn spinning process and cost optimization.
22	Khan et al.⁷⁷	2009	Fiber properties, yarn twist and yarn fineness, process parameters.	Multilayer Perceptron (MLP), Multivariate Linear Regression (MLR) and ANN	To prediction of hairiness of worsted-spun wool yarn.
23	Majumdar et al.⁵³	2008	Cotton fiber properties.	Back propagation ANN and Fuzzy logic	To predict the unevenness of cotton carded ring-spun yarn.
24	Mwasiagi et al.⁷⁸	2008	Cotton fiber properties, yarn count and twist, and four spinning process parameters.	Back propagation ANN	To evaluate ANN prediction models and recommend the best one for yarn breaking elongation prediction.
25	Üreyen and Gürkan⁷⁹	2008	Yarn count, twist, and roving properties.	ANN and Linear Regression	To predict the ring cotton yarn properties.
26	Mwasiagi et al.⁸⁰	2007	HVI cotton fiber properties, processing parameters, and yarn quality.	Multilayer feed forward perceptron ANN	To forecast the strength, elongation, and uniformity of cotton ring-spun yarn.
27	Chen et al.⁸¹	2007	Fiber properties and yarn count, and twist.	Back propagation ANN	To predict the yarn tenacity.
28	Beltran et al.⁸²	2006	number of ﬁbers in cross-section and spinning ends-down.	Multilayer perception ANN	To predict worsted yarn performance.
29	Babay et al.⁸³	2004	Fiber characteristics, and yarn twist and fineness.	Back propagation ANN	To estimate the hairiness of yarn based on fiber and yarn characteristics.
30	Beltran et al.⁸⁴	2004	Fiber characteristics, yarn twist and fineness, and process parameters.	Back propagation ANN	To predict the worsted spinning performance.
31	Majumdar and Majumdar⁸⁵	2004	Cotton fiber properties, yarn count and twist.	Back propagation ANN	To forecast ring-spun cotton yarn breaking elongation.
32	Shanmugam et al.⁸⁶	2001	Fiber characteristics.	Back propagation ANN	To predict the micro-pun yarn lea CSP.

Ghanmi et al.⁵⁷ validated the use of Fuzzy expert systems with back-propagation ANN for global yarn quality prediction. Several variables include yarn fineness and twist per meter, in addition to fiber length, strength, fineness, and elongation. The data showed that the hybrid models could use the input factors that were chosen to predict the new yarn quality index very well. Abd-Elhamied et al.⁵⁸ assessed ring-spun and compact yarn strength, elongation percentage, coefficient of variation in mass, and defects using image processing and ANN. Two cotton mills, ring spinning and compact spinning provided yarn samples. Back-propagation neural networks were fed feature vectors. Each of two systems has three modules to evaluate yarn properties. Multilayer network topologies improve network performance and parameter modeling. A cost-effective method assessed yarn quality for several varieties.

El-Geiheini et al.⁵⁹ conducted a study to construct a cost-effective image processing-ANN system to model yarn tenacity and elongation. Using an appearance tester, the lab wound yarn samples around the blackboard. The specimens were digitally photographed. Two ANNs were trained via Levenberg-Marquardt backpropagation after filtering and scaling. Network training used 80% of the data, validation and testing 10%. Picture enhancement and multilayer neural networks enhanced yarn quality estimate. Doran and Sahin⁶⁰ suggested utilizing ANN and SVM to predict cotton elastane core yarn quality. The best models predicted most yarn quality metrics. Optimized models in MAPE and R exceeded 90% with the required model. Predictions of cotton elastane core yarn irregularity, yarn hairiness, and Resskilometer quality were 91%, 93%, and 95% accurate. They state additional experiments improve ANN and SVM. Experiments and larger training sizes can cause the model to store more data, increasing its inaccuracy. Ahmad¹² designed a simplified graphical user interface for ANN, which was verified on a novel dataset. Input factors were yarn count, twist, and fiber properties. The study produced 32 cotton ring-spun yarns using 12 cotton blends. Yarn counts ranged from 6.96 to 24.06 Ne, and twists from 338 to 921 turns per meter. After yarn strength testing, ANN analyses were done. Comparisons were made using real data and the Soloviev relationship used to predict cotton yarn tensile strength. ANN results were better than Soloviev’s. Kavitha et al.⁶¹ used ANN with and without PCA in ring spinning. The collection includes 1000 samples of conventional and typical ring spinning conditions. Spindle speed, top roller pressure, traveler mass, temperature, current, and voltage are used to train ANN and PCA models. The rates of end breakage, yarn irregularity, and yarn imperfection were projected. PCA reduced input data dimensionality. PCA-ANN model performance and classification accuracy have improved. Ghanmi et al.⁶² created a hybrid model by combining two soft computing methodologies, namely ANN and fuzzy expert system. This model was used to analyze the yarn qualities and assign an index value between 0 and 1 to each combination of these properties. The hybrid model was trained using fiber characteristics, yarn count, and twist level as inputs. The output of the model is a quality index that encompasses the key physical attributes of ring-spun yarn. Unal and Ozdil⁶³ analyzed the diameter of cotton ring spun yarns by employing regression and ANN models, with a focus on the qualities of the cotton fibers. The derived equations and networks were also compared with the yarn diameter of cotton ring-spun yarns, as established by Peirce.⁸⁷ The findings demonstrated that the generated regression equation provided a reliable calculation of yarn diameter in comparison to Pierce’s equations. Almetwally et al.⁶⁴ studied cotton/spandex core-spun yarn breaking strength, elongation, and work of rupture using ANN and multiple regression. The company made 40 30/1 Ne cotton/spandex core-span yarns. The yarn had a spandex filament with 22, 33, 44, and 78 dtex liner densities. All-cotton yarn sheath. 2.4, 2.8, 3.3, 4, and 4.4 drawing ratios were used for each filament. Individual yarn samples were made using twist multipliers 4.0 and 4.2. Using MSE, Mean Bias Error (MBE), and R², the two models were tested. The study found that ANN predicts better than multiple linear regression. Mozafary and Payvandy⁶⁵ used clustering and ANN to predict yarn quality. In one year, the quality control laboratory collected almost 150,000 data points. The data included fiber characteristics, manufacturing process parameters, and yarn quality measures. Clustering and ANN were used to predict yarn quality in unevenness, nep, thinness, and thickness. The data-mining method outperformed the ANN in accuracy. Das et al.⁶⁶ developed mathematical formulas to predict cotton yarn strength and irregularity based on fiber properties. A genetic algorithm (GA) uses ANN equations to find the best yarn-spinning fibers based on natural selection and survival of the fittest. Tenacity and unevenness were desired for a spinning mechanism. Ghosh et al.⁶⁷ improved yarn strength and raw cotton quality with ANN and regression models. Elitist multi-objective evolutionary algorithms using Non-dominated Sorting Genetic Algorithm II solved problems most efficiently. Multi-objective optimization maximized cotton yarn strength and minimized SCI. The evolutionary algorithm is a unique multi-objective optimization method that finds multiple optimal solutions in one simulation run.

Bo⁶⁸ predicted cotton ring-spun yarn end breakage using ANN and MLR. The ANN model inputs were front roller speed, spindle speed, nip gauge, rear draft zone time, and ring yarn twist, while the output was end breakage rate. Validating the MSE and R² for test data prediction measured model performance. Results showed that ANN outperformed various linear models. Soltani et al.⁶⁹ suggested a new way to predict the migratory patterns of siro, solo, compact, and conventional ring-spun solar yarns. Spun yarns’ physical and mechanical properties and current requirements were investigated. The improved ANN method predicted spun yarn migration. The results showed that ANN models can reliably predict yarn migratory qualities based on physical and mechanical factors. Bo⁷⁰ modeled ring-spun polyester/cotton yarn end breaking with a three-layer BP neural network. An FA506 ring frame machine spun 50% cotton and 50% polyester into 18.2 tex yarn. Ring-spun yarn end breakage was compared to front roller delivery, spindle speed, nip gauge, break draft, and roving twist using the backpropagation neural network model. The simulation accuracies were good, and extending the experimental database for network training may improve the BP neural network approach. Admuthe and Apte⁷¹ used ANFIS and subtractive clustering to predict yarn properties. The trial included 30 cotton varieties. Fiber length, strength, micronaire, maturity, short fiber content, and elongation were inputs. After training both systems, 100 samples were tested to predict yarn properties including unevenness ratio (UR), thin place, thick place, elongation, and count. Overall, the ANFIS approach was tested on numerous datasets and compared to the ANN model. Bo⁷² predicted polyester/viscose/cotton ring-spun yarn hairiness using multiple linear regression and ANN algorithms with processing parameters. The yarn was 40% polyester, 40% viscose, and 20% cotton. At 14.6 tex, yarn was fine. A mixed-effects regression model with five process parameters predicted yarn hairiness. Regression model predictions were less reliable than ANN model predictions. Furferi and Gelli⁷³ created a feed-forward backpropagation ANN model to help technicians predict yarn strength. Instead of spinning yarn, this method uses roving properties to create accurate forecasts. Blending fibers from a spinning mill in Prato, Italy produced six roving kinds. Each was measured for fiber strength, length, yarn twist, and count. Huijun and Xinhou⁷⁴ predicted 50 Ne combed yarn unevenness by using BP ANN. The micronaire, length, uniformity, strength, whiteness, yellowness, and neps were input parameters, as were yarn unevenness, knots (+50%), and neps (+140%). Every projected coefficient above 0.93763, and yarn evenness and knots (+50%) predicted correlation coefficients exceeded 0.94247. Majumdar⁷⁵ improved cotton yarn hairiness prediction with ANN and adaptable neuro-fuzzy inference models. Cotton fiber mean length, short fiber content, maturity, and yarn linear density were model inputs. Even in unseen test samples, the method estimates cotton yarn hairiness 2% inaccurately. Admuthe and Apte⁷⁶ employed an ANN and a Genetic Algorithm technique to predict the fiber characteristics necessary for achieving the target yarn quality and to optimize their cost. The hybrid approach technique is proposed as a powerful model that has the potential to greatly enhance the predictability and profitability of the yarn production business. Khan et al.⁷⁷ tested MLP and MLR ANN models to predict worsted-spun wool yarn hairiness. Many toppings, yarn, and process characteristics affected projections. Different top specs and processing parameters were tested in 75 yarn sets. MLP model showed yarn hairiness-input parameter correlation nonlinearity, performing marginally better. The MLP model’s sensitivity analysis showed that yarn twist, ring size, and average fiber length most affected yarn hairiness. Majumdar et al.⁵³ used ANN and neural-fuzzy systems to predict ring-spun yarn irregularity. ANN, neural fuzzy, and linear regression models were compared for prediction accuracy. All three models predicted well, although the ANN and neural-fuzzy models outperformed the linear regression model. The neural-fuzzy system’s language rules show how input factors affect yarn consistency. Mwasiagi et al.⁷⁸ changed design variables to test neural networks’ ability to forecast cotton ring-spun yarn breaking elongation. Two transfer functions differed significantly from training functions. Levenberg-Marquardt outperformed the other five training functions. Input component analysis ranked yarn twist, yarn count, fiber elongation, length, length consistency, and spindle speed highest. Üreyen and Gürkan⁷⁹ developed an ANN model to predict ring cotton yarn qualities using High Volume Instrument (HVI) fiber property measurements. The ANN model was compared to a statistical regression model for effectiveness. A total of 180 cotton ring-spun yarns were made from 15 blends. Yarn counts and twist multipliers were limited to 20–35 Ne and 3.8–4.6 TM. Tenacity and breaking elongation of the yarn were measured and analyzed using an ANN. Finally, the R² value of ANN and regression models were assessed. Mwasiagi et al.⁸⁰ examined how ANN algorithms predicted cotton yarn strength, elongation, and evenness. Kenyan spinning mills provided cotton lint and carded ring-spun yarn samples. To reduce interjection differences, machinery technology, work culture, quality, and maintenance rules should be as same as possible. The main inputs were fiber qualities, processing parameters, and yarn quality. Chen et al.⁸¹ created a soft computing model to predict yarn strength using fiber and yarn properties. To discover the most important fiber quality and yarn properties for the small-scale ANN model, they were ranked. Also, an ANN model was created to analyze the relationship between fiber quality, yarn properties, and tenacity. Results show ANN model makes accurate predictions. Beltran et al.⁸² developed an MLP-based mill-worsted spinning performance prediction tool. Sixteen variables predicted spinning efficiency and yarn quality. Cross-section fiber count, Unevenness (U%), thin areas, neps, yarn toughness, break elongation, thick areas, and spinning ends-down were inputs. Using mill-specific commercial data, the model matched expected values. MLP performed consistently across randomly selected independent test data. Based on Sirolan Yarnspec’s projected results, the ANN technique improves mill-specific forecast accuracy. Babay et al.⁸³ modeled ring yarn hairiness with a backpropagation neural network and linear regression. With the HVI, fiber properties were measured. Data and process characteristics like twist and yarn count predicted hairiness. A leverage-based local overfitting control model selection method selected the right architecture. Less training and generalization mistakes made the neural network model better. These errors were less than the hairiness standard deviation in 5% Uster data. The linear model returned 0.8893, whereas R² between observed and projected hairiness was 0.9353. Neural models held more or equal confidence than linear forecasts. Beltran et al.⁸⁴ examined tested ANNs’ ability to predict worst-case spinning performance. Sirolan Yarnspec randomly created 250 training data sets. Each dataset contained produced inputs and Yarnspec predicted outputs. They recommended improving yarn prediction for unevenness, cross-sectional diameter, thin areas, thick areas, and tenacity. Majumdar and Majumdar⁸⁵ evaluated the anticipated effectiveness of mathematical, statistical, and ANN models. A prediction was made regarding the breaking elongation of ring-spun cotton yarns. Cotton fibers and yarn count were inputs to the models. The HVI instrument measured cotton fiber properties’ importance, which the ANN model assessed. Shanmugam et al.⁸⁶ Predicted micro-spun yarn CSP by backpropagation ANN model effectively. The study examined fiber span length, bundle strength, fineness, breaking elongation, uniformity ratio, and matured fiber %. The neural network with five hidden neurons in a single layer and a 12-epoch made better predictions. To predict Count Strength Product (CSP), the ANN and regression models were compared. The neural network model had 60% lower MSE than the regression model.

The majority of applications of ANN in yarn manufacturing have been thoroughly addressed and effectively presented in literature concerning the technical parameters of ring frames and process variables to forecast yarn characteristics such as irregularity, thick places, thin places, neps, imperfections, hairiness, strength, and elongation. Compact spinning is a pivotal aspect of the contemporary advancement of ring spun yarns, enhancing process efficiency and the quality of final garments through reduced hairiness and increased strength. There is still scope in application of ANN in prediction the characteristics of compact, core-spun, slub, mélange, and other fancy yarns.

Prediction in rotor- and vortex-spun yarn properties

The selection and design of machine components have a considerable impact on the quality and performance of rotor-spun yarn.⁸⁸ Vortex-spun yarn has core and wrapping. Since fiber separation happens everywhere on the fiber bundle’s outer edge, multiple wrapper fibers create a ring spun look with improved durability.⁸⁹ An overview of the application of ANN in the prediction on rotor- and vortex-spun yarn characteristics has been shown in Table 4.

Table 4.

Application of ANN in prediction on rotor- and vortex-spun yarn characteristics.

Sl.	Authors	Year	Material used	Technique	Aims
1	Ghanmi et al.⁹⁰	2019	Fiber properties in blends.	Back propagation ANN and Fuzzy expert system	To create rotor-spun yarn quality prediction method.
2	Erbil et al.⁹¹	2017	Fiber blend ratio, mixing method, number of passages of draw frame, and fineness of yarn	Back propagation ANN and Multiple Linear Regression	To forecast OE rotor spun yarn breaking strength and elongation% from ternary fiber mixtures.
3	Ghorbani et al.⁹²	2016	Machine parameters of rotor spinning frame.	Back propagation ANN and Multiple Linear Regression	To predict the rotor-spun yarns hairiness.
4	Moghassem et al.⁹³	2012	Production speed, break draft and drafting zone setting.	Back propagation ANN and Gene Expression Programming (GEP)	Rotor-spun yarn breaking strength prediction.
5	Zhao³⁰	2012	Opening roller speed and diameter, rotor speed and diameter	Back propagation ANN	To anticipate cotton/polyester rotor-spun yarn hairiness.
6	Pei and Yu⁹⁴	2011	Jet orifice diameter and yarn delivery speed.	MLR neural network	To forecast vortex yarn tenacity-machine parameter correlations.
7	Shanbeh et al.⁹⁵	2011	Rotor diameter, speed, navel type, opener roller speed, ginning waste proportion, yarn linear density	Feed forward ANN	To forecast strength and mass irregularity of rotor-spun yarns made from ginning waste.
8	Bo⁹⁶	2010	Opening roller diameter and rpm, rotor diameter and rpm.	Back propagation ANN	To predict the cotton rotor-spun yarn hairiness.
9	Ghosh and Chatterjee⁹⁷	2010	Cotton fiber characteristics.	SVM and Back propagation ANN	To predict the cotton yarn properties and analysis of prediction capability of ANN and SVM.
10	Majumdar⁹⁸	2010	Cotton fiber properties and yarn count.	Back propagation ANN and Regression	Using cotton fiber qualities to predict ring and rotor yarn hairiness.
11	Nurwaha and Wang⁹⁹	2010	Cotton fiber properties.	Back propagation ANN and Hybrid Fuzzy System	To predict rotor-spun yarn strength.
12	Demiryürek and Koç¹⁰⁰	2009	Blend ratio, yarn count, rotor speed.	Multi-layer perceptron ANN	Predicting polyester/viscose blended open-end rotor-spun yarn unevenness.
13	Demiryürek and Koç¹⁰¹	2009	Polyester viscose blend ratio, fiber properties, and yarn fineness.	Multi-layer perceptron ANN and Mixture Process Crossed Regression Model	To anticipate polyester/viscose blended open-end rotor spun yarn breaking elongation.
14	Majumdar et al.¹⁰²	2005	Cotton fiber properties, ring-rotor yarn fineness, twist multipliers of yarn.	Back propagation ANN, ANFIS and Fuzzy logic	To predict ring-rotor yarn strength using cotton fiber characteristics.
15	Majumdar et al.¹⁰³	2005	Cotton fiber properties and yarn count.	Back propagation ANN and Neuro-fuzzy Model,	Rotor-spun yarn breaking elongation prediction.
16	Jackowska-Strumillo¹⁰⁴	2004	Content of flax in sliver, yarn fineness.	Back propagation ANN	To determine the influence the percentage of flax content on feed sliver and linear density in cotton/flax blended yarn.
17	Majumdar¹⁰⁵	2004	Fiber characteristics and yarn fineness	Back propagation ANN	Predicting ring and rotor spun single yarn tenacity.
18	Jackowski et al.¹⁰⁶	2002	Linear densities of slivers and yarns.	Multilayer perceptron ANN	To predict the coefficient of mass variation of yarn.
19	Basu et al.¹⁰⁷	2002	Delivery speed, nozzle pressure, first nozzle to front roll distance, draft ratio, yarn fineness.	Back propagation ANN	Prediction of air-jet-spun yarn characteristics.

Ghanmi et al.⁹⁰ developed a hybrid model using ANN and fuzzy expert systems. ANN predicted yarn toughness, breaking elongation, and irregularity. These three outputs predicted the innovative quality index using fuzzy expert system. R, RMSE, MAE, and MRPE assessed the prediction model’s accuracy. Analysts found the hybrid model predicted rotor-spun yarn quality accurately. Erbil et al.⁹¹ conducted a stepwise MLR investigation produced a Levenberg-Marquardt-trained ANN model with backpropagation. A comparison of both models’ prediction performance followed. Statistical and ANN models predicted the tensile properties of ternary mix open-end yarns made from polyester (PES), viscose (CV), and polyacrylonitrile (PAN) fibers with six blend ratios and three linear dens Next, prediction model statistical performance criteria were evaluated. For breaking strength and elongation, ANN and MLR models were built. MLR research shows yarn count and polyester ratio affect breaking strength statistically. This significance has 1% confidence and R² = 0.938. These data suggest the blend’s PES ratio influences yarn strength. The statistical study found that PAN content, yarn count, mixing method, and polyester ratio effect breaking elongation. R² = 0.589 and 1% confidence support this result. However, ANN predicted breaking strength better. Ghorbani et al.⁹² studied how rotor type, diameter, doffing-tube nozzle, and torque-stop affect polyester/cotton rotor-spun yarn hairiness. The study characterized yarn hairiness using ANN and regression models. Hairiness was best predicted by two-hidden-layer networks. The study used optimal network-based partial derivatives to evaluate input variable relevance. The rotor type and diameter affected blended yarn hairiness most and least, respectively. Moghassem et al.⁹³ used three draw-frame machine parameters to forecast cotton rotor-spun yarn breaking strength. Comparing GEP and ANN models’ performance. The two proposed models’ MSE and R² values differed by 33.01% and 4.96% in forecasting test data. Zhao³⁰ constructed an ANN model to forecast the level of hairiness in cotton/polyester blended yarn manufactured by rotor spinning. Their outcome indicators, including MSE, MAE, and MAPE, exhibited a significant level of concurrence between the anticipated and experimental results. This implies that the neural network is an exceedingly efficient instrument for forecasting results. Pei and Yu⁹⁴ used ANN and numerical simulation predicted vortex yarn tenacity depending on process and nozzle parameters. The vortex-spinning nozzle’s fiber-airflow interaction and fiber movement were simulated numerically. The jet orifice diameter and yarn delivery speed predicted yarn strength. MLP neural networks predicted modal fiber vortex yarn strength. A good correlation coefficient and little relative errors exist between predicted and measured tenacities. This suggests ANN can predict accurately. Shanbeh et al.⁹⁵ applied an ANN-based predictive model of breaking strength and unevenness of cotton waste rotor-spun yarn comprising ginning waste. Rotor parameters, ginning waste proportion, and yarn linear density were inputs to the ANN models. To produce the best prediction models, ANN model parameters including learning, momentum rate, hidden layers, and hidden processing elements were optimized. The maximum error in forecasting testing data breaking strength and mass variation was 8.34% and 6.65%. Bo⁹⁶ used the ANN model to predict cotton yarn hairiness during rotor spinning. The opening roller rpm, diameter, rotor rpm, and rotor diameter were process factors of the rotor spinning machine. The ANN model with the back-propagation method can predict cotton rotor-spun yarn hairiness based on processing factors. Ghosh and Chatterjee⁹⁷ predicted the properties of cotton yarn, such as tenacity, breaking elongation, evenness, and hairiness using SVM regression approach. The qualities of the fibers and ring and rotor spinning were used to design this yarn. The expected accuracy of SVM and ANN models was compared. The expected simplification accuracy of both models was assessed using k-fold cross-validation. SVM models outperformed ANN models in predicting yarn characteristics. Majumdar⁹⁸ investigated two methods for predicting ring and rotor yarn hairiness. The ANN and linear regression models used cotton fiber properties and yarn count. A correlation coefficient of 0.92 and an MAE < 4% indicated great prediction accuracy for the models. The ANN model demonstrated superior performance compared to regression in the prediction of ring yarn hairiness. Nurwaha and Wang⁹⁹ utilized a hybrid neuro-fuzzy system to predict cotton rotor spun yarn strength based on fiber quality. Their approach combines ANN with fuzzy logic benefits. They also evaluated how different fiber quality affected rotor-spun yarn strength. Fiber strength, length, homogeneity, and thread count improve rotor-spun yarn strength. Conversely, micronaire, yellowness, and short fiber content weaken yarn. Demiryürek and Koç¹⁰⁰ developed an ANN and statistical model to forecast polyester/viscose blended open-end rotor spun yarn irregularity. A rotor spinning machine handled seven polyester/viscose sliver blend ratios with four rotor speeds and four yarn counts. Backpropagation MLP neural networks and mixer process crossing regression models (simplex lattice design) predicted yarn unevenness. The ANN and simplex lattice designs predict well and are more reliable than the ANN alone. Demiryürek and Koç¹⁰¹ developed two models to predict the breaking elongation (%) of polyester/viscose mixed rotor-spun yarns: an artificial neural network (ANN) model and a statistical simplex mixing process crossing regression model. The input parameters utilized in the back propagation feed forward artificial neural network (ANN) model consisted of the blended ratio, yarn count, and rotor speed. The output parameters were identified as the breaking elongation of the yarns. The ANN model exhibited more reliability in comparison to the statistical model. Majumdar et al.¹⁰² used a hybrid neuro-fuzzy system to predict cotton yarn tenacity using HVI fiber metrics like strength, elongation, length, uniformity index, micronaire, reflectance, and yellowness. 4.1 and 4.8 twist multipliers were used to make ring and rotor yarns. Train the neuro-fuzzy inference system with HVI fiber test results and compare its predictive performance to ANN and regression models. Majumdar et al.¹⁰³ used linear regression, ANN, and neuro-fuzzy models to predict rotor yarn elongation. Cotton fiber characteristics and yarn fineness are inputs for the models. The study examined how yarn count and cotton fiber characteristics affect rotor elongation. Model predictions were compared for accuracy. The prioritization of findings was achieved by data analysis and the use of ANN model. Jackowska-Strumillo¹⁰⁴ presented ANN-based yarn spinning model architecture. To mimic the spinning process and analyze yarn quality parameters like tenacity, hairiness, and flaws, partial models were developed. Feed-forward neural networks were employed for modeling purposes. Yarns produced in rotor frame by blending of flax and cotton slivers and 100% cotton slivers. Majumdar¹⁰⁵ developed ANN model to predict the tensile strength of individual yarns created by ring and rotor spinning techniques, using fiber characteristics as input. The input parameters comprised seven fiber attributes and yarn count. The yarn tenacity exhibited an MSE of less than 5% for ring-spun yarns and less than 2% for rotor-spun yarns. The study also evaluated the relative importance of cotton fiber properties and thread count on the strength of individual yarns. Jackowski et al.¹⁰⁶ did an experimental work on the BD 200S rotor spinning machine used rotor rotational speed, opening roller speed, and yarn linear densities. An additional draw frame sliver fed the spinning machine. The partial hybrid yarn mass variation coefficient model considers yarn and sliver linear densities. The ANN estimated yarn mass variation coefficient from silver linear density and tenacity variation. A modest one-hidden-layer multilayer perceptron network was employed for this assignment. Basu et al.¹⁰⁷ developed an ANN model to predict the properties of air-jet yarn by assessing process data. An attempt was made to utilize a reverse model to predict process variables in air-jet spinning that would lead to specified yarn characteristics. A backpropagation algorithm was implemented using a neural network that consisted of only one hidden layer. The inverse model is expected to play a vital role in the control, modeling, and prediction of the spinning process.

A substantial amount of research is necessary to forecast the vortex yarn and yarn produced from other modern spinning techniques. The domain of innovative technology continues to allow for the prediction of spun yarn produced from friction, wrap, and electrojet spinning methods.

Prediction in spliced yarn properties

The electronic clearer in the winding machine eliminates yarn faults of significant scale that have been stored in the memory.⁵ This technique has also been utilized for the purpose of connecting yarn ends, commonly referred to as splicing of broken ends.¹⁰⁸ The tension, cradle pressure, winding speed, and traverse ratio are the most important technological factors for a winding machine.¹⁰⁹ The great performance of pneumatic splicing is essential for uniform yarn that involves turbulent airflow over yarn ends.^110,111

Ünal¹¹² examined how splicing, fiber, and yarn characteristics affect strength and elongation. Eight cotton types were utilized to make yarns with three fineness and three twist coefficients. Spliced yarn tenacity and elongation were analyzed using ANN and response surface models. Yarn twist and count were chosen based on fiber quality and machine factors like opening air, splicing air, and splicing time. Air pressure during splicing impacts yarn strength and stretching, according to ANN models. In prediction, ANN models outperformed response surface models. This is because ANN models explain non-linear connections better. Unal¹¹³ used ANN and response surface models to predict cotton ring spun yarn’s preserved spliced diameter. Eight cotton kinds were used to make yarns with three fineness and three twisting level. Fiber properties were measured using an AFIS testing machine. The ANN model predicted spliced diameter better than response surface models for ring-spun cotton yarn (Table 5).

Table 5.

Application of ANN in prediction in spliced yarn properties.

Sl.	Authors	Year	Material used	Technique	Aims
1	Ünal et al.¹¹²	2010	Fiber properties, fineness of yarn, and Twist multiplier	Multilayer feed-forward ANN and Response Surface Method (RSM)	To assess the tensile strength and elongation of spliced yarn.
2	Ünal et al.¹¹³	2010	Fiber properties, fineness of yarn, and Twist multiplier	Multilayer feed-forward ANN and RSM	To determine how fiber, yarn, and splicing conditions affect retained spliced diameter.

A least number of studies was identified on the prediction of spliced and conditioned yarn from the technical parameters of winding and yarn conditioning operations. The winding speed, winding tension, winding angle, splicing pressure, splicing duration, splicing length, and the settings of electronic yarn clearers are the key factors that should be prioritized in predicting the characteristics of spliced yarn. These significant factors may indicate the potential for optimal productivity and quality of spun yarns, as well as the finished textile product.

Discussion and future perspectives

Nowadays, the management exercises control over the manufacturing process and product quality through the utilization of technical personnel and skilled operators. The utilization of the ANN model as an analytical instrument can aid in the selection of material specifications and enhance processing parameters based on the anticipated results of the models.⁶³ Feedforward and backpropagation techniques are two of the most common applications of ANN in the textile and apparel sectors. Because of the intricate nonlinear relationship among the raw material parameters, spinning method, and yarn properties, it is challenging to develop a prediction model based on product requirements.^23,114 The examination primarily emphasized in forecasting of ring-spun yarn properties. Research studies were scarce concerning the prediction on vortex-spun yarn quality and process variables, whereas a moderate number of studies concentrated on the quality of rotor-spun yarn. The ANN exhibited better results than any other models except the SVM unveiled better prediction performance than ANN.⁹⁷

Among the selected articles, none was on blow room settings, comber noil adjustment, and yarn conditioning. Further research is necessary to precisely determine the frequency of spinning ends down and neps utilizing the ANN. Further mill-specific data and breakthroughs in the ANN simulations are necessary to enhance the forecasts of quality parameters.²⁰ Hence, there is a substantial possibility to do research with ANN. In addition, robust learning methods are required to enhance the ANN models to be capable to handle from fiber to yarn. Predicting yarn characteristics can reduce both the time and money required for testing. No study has been discovered examining the correlation between manufacturing cost, time, and price of spun yarns. Therefore, there is ample opportunity for research in this area.

The wide range of processing procedures and equipment is a challenge in evolving a universal experimental model that can precisely envisage yarn characteristics across different mills. The Uster Statistics are universally acknowledged as the exclusive globally-accepted quality benchmarks for fiber, sliver, roving, and yarn in the yarn manufacturing sector.¹¹⁵ It is necessary to have a unified worldwide platform of yarn quality that could be also accomplished by the implementation of ANN.

The waste was also reduced due to the optimal exploitation of fiber during twisting. The reduction of pneumafil waste, end breakage rate, and the enhancement of productivity are critical area that are unveiled in prediction of ring spun yarn characteristics. The economic viability is also crucial. The minimization of ring frame waste lessens its direct influence on manufacturing expenses, hence impacting economic sustainability and the working conditions of personnel in the industry. The present research trend is creating eco-friendly and cost-effective clothes from yarn manufacturing waste. Nevertheless, the waste are being utilized through dependence on the process. A significant amount necessitates further repurposing. By merging diverse electro-mechanical processes, they can be utilized for sustainable energy sources. The application of ANN can also be employed in this field to forecast the characteristics of spun yarn. Moreover, this preference depends mostly on selection of cost of raw materials and process control in spinning industries.

Conclusion

In current study, a thorough analysis of the literature that focused on the use of ANN on yarn quality prediction was undertaken. The review examined the attributes of raw materials in the spinning industry, the processing of materials through various phases, and the final output of the initial phase of backward linking in the textile and garment sectors. This assessment demonstrated that ANN is capable to predict the characteristics of in-process materials and final product of spinning industry. The challenges that are associated with textile spinning will be the subject of additional research to facilitate the replacement of human judgment with intelligent systems. Before the data is put into an ANN, training, such as using process-captured online data, is necessary. Various hybrid models integrating with ANN like Genetic Algorithms, Decision Trees, and SVM can optimize the parameters of spun yarns and enhance the architecture of the ANN, thereby strengthening the model as well as improving forecasting accuracy. This study has explored the theoretical implications through a systematic review and comprehensive analysis of scholarly papers. It has practical implications for industrial specialists in predicting yarn qualities without testing, which solves real-life problems in the spinning industry. Finally, the study ended by providing future directions that can serve as a valuable guide for researchers in their subsequent research endeavors.

Footnotes

ORCID iDs

Jamal Hossen

Md. Reajul Islam

Subrata Kumar Saha

Shahriar Raian

Amit Chakrabortty

Emdadul Hoque

Chowdhury Jony Moin

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Ahmad

Hossen

Ali

SM.

Improvement of overall equipment efficiency of ring frame through total productive maintenance: a textile case. Int J Adv Manuf Technol 2018; 94(1–4): 239–256.

Qiu

Zhou

Zhang

Evolving regularised random vector functional link by seagull optimisation algorithm for yarn-dyed fabric colour difference classification. Coloration Technol 2024; 140: 467–482.

Hossen

Saha

SK.

Influence of blending method and blending ratio on ring-spun yarn quality – a MANOVA approach. Tekstilec 2023; 66(3): 227–239.

Abdkader

Hossain

A critical review on recent developments and solutions in the high-speed ring spinning process. Text Res J 2023; 93(22–23): 5485–5504.

Rutkowski

Analysis of the strength parameters of worsted and component spun yarns after the rewinding process. Fibers Textiles Eastern Eur 2018; 26(2): 32–37.

Ramesh

Rajamanickam

Jayaraman

The prediction of yarn tensile properties by using artificial neural networks. J Text Inst 1995; 86(3): 459–469.

Cong

Gao

Evaluation of knitted suit fabric style based on fuzzy neural network. J Eng Fiber Fabr 2020; 15: 1–11.

Haykin

Neural networks: a comprehensive foundation. 2nd ed. Prentice Hall, 1998.

Vassiliadis

Rangoussi

Cay

, et al. Artificial neural networks and their applications in the engineering of fabrics. In: Dubrovski

(ed.) Woven fabric engineering. Intech, 2010, pp.111–134.

10.

Zhao

Wang

. The prediction model of cotton yarn quality based on artificial recurrent neural network. In: Abawajy

Choo

Islam

, et al. (eds) Advances in intelligent systems and computing. Springer Verlag, 2020, pp.857–866.

11.

Ghosh

Ishtiaque

Rengasamy

RS.

Analysis of spun yarn failure. Part II: the translation of strength from fiber bundle to different spun yarns. Text Res J 2005; 75(10): 741–744.

12.

Ahmad

GG.

Using artificial neural networks with graphical user interface to predict the strength of carded cotton yarns. J Text Inst 2016; 107(3): 386–394.

13.

Pareira

Carvalho

Vasconselos

, et al. A review in the use of artificial intelligence in textile industry. In: Machado

Soares

Trojanowska

, et al. (eds) Innovations in mechatronics engineering, Conference paper, 2021, pp.377–392.

14.

Chattopadhyay

Artificial neural networks in yarn property modeling. In: Majumdar

(ed.) Soft computing in textile engineering. Woodhead Publishing Ltd, 2011, pp.105–125.

15.

Sikka

Sarkar

Garg

Artificial intelligence (AI) in textile industry operational modernization. Res J Text Apparel 2024; 28(1): 67–83.

16.

Tran

, et al. Modeling of textile manufacturing processes using intelligent techniques: a review. Int J Adv Manuf Technol 2021; 116: 39–67.

17.

Mukhopadhyay

Siddiquee

Artificial neural networks and their use in textile technology. Asian Text J 2003; 12(3): 72–77.

18.

Chattopadhyay

Guha

Artificial neural networks: applications to textiles. Textile Progress 2009; 35(1): 1–46.

19.

Jiang

Song

Zhang

, et al. Prediction of yarn unevenness based on BMNN. J Eng Fiber Fabr 2021; 16: 1–15.

20.

Suzuki

Artificial neural networks: industrial and control engineering applications. Intech, 2011.

21.

Ramzan

Qureshi

Habib

, et al. Evaluation of human factors on visual inspection skills in textiles and clothing: a statistical approach. J Eng Fiber Fabr 2022; 17: 1–12.

22.

Dayik

Prediction of yarn properties using evaluation programing. Text Res J 2009; 79(11): 963–972.

23.

Shaikh

Agrawal

SA.

Engineering cotton yarns with artificial neural networking (ANN). 1st ed. Woodhead Publishing India Ltd, 2017.

24.

Fausett

LV.

Fundamentals of neural networks: architectures, algorithms and applications. Pearson Education India, 2006.

25.

Haykin

Neural networks: a comprehensive foundation. 2nd ed. Pearson Education, 1999.

26.

Zhang

Zhou

Classifying colour differences in dyed fabrics using an improved hunger games search optimised random vector functional link. J Eng Fiber Fabr 2022; 17: 1–11.

27.

Shi

Yang

Color difference classification of dyed fabrics via a kernel extreme learning machine based on an improved grasshopper optimization algorithm. Color Res Appl 2021; 46(2): 388–401.

28.

Pynckels

Kiekens

Sette

, et al. The use of neural nets to simulate the spinning process. J Text Inst 1997; 88(4): 440–448.

29.

Pynckels

Kickens

Sette

, et al. Use of neural nets for determining the spinnability of fibers. J Text Inst 1995; 86(3): 425–437.

30.

Zhao

. Study on hairiness of polyester/cotton blended yarn in rotor spinning via artificial neural network theory. In: Proceedings of the international conference on information engineering and applications (IEA) (ed Zhong

), 2012, pp. 487–494. Springer.

31.

Kasabov

NK.

Foundations of neural networks, fuzzy systems, and knowledge engineering. 2nd ed. The MIT Press, 1998.

32.

Lawal

Idris

MA.

An artificial neural network-based mathematical model for the prediction of blast-induced ground vibrations. Int J Environ Stud 2020; 77(2): 318–334.

33.

Lien

Lee

A method of feature selection for textile yarn grading using the effective distance between clusters. Text Res J 2002; 72(10): 870–878.

34.

Amani Tehran

Maleki

. Artificial neural network prosperities in textile applications. In: Suzuki

(ed.) Artificial neural networks: industrial and control engineering applications. InTech, 2011, pp.35–64.

35.

Chen

(ed.). Modelling and predicting textile behaviour. 1st ed. The Textile Institute, 2010.

36.

Jiang

Song

Zhang

, et al. Yarn unevenness prediction using generalized regression neural network under various optimization algorithms. J Eng Fiber Fabr 2022; 17: 1–10.

37.

Ghanmi

Ghith

Benameur

Application of artificial neural network in textile: prediction of ring spun yarn properties. IOSR J Eng 2014; 4: 28–34.

38.

Yong

Jian

Xian

, et al. Test and analysis of the porosity of cotton fiber assembly. J Eng Fiber Fabr 2021; 16: 1–7.

39.

Cui

Xue

Liu

Manufacturing a ring spun slub yarn using multi-channel drafting technique. J Eng Fiber Fabr 2020; 15: 1–13.

40.

Kang

Kim

SC.

Objective evaluation of the trash and color of raw cotton by image processing and neural network. Text Res J 2002; 72(9): 776–782.

41.

Fang

Watson

MD.

Clustering analysis for cotton trash classification. Text Res J 1999; 69(9): 656–662.

42.

Wang

Gao

, et al. One-dimensional convolutional neural network with data characterization measurement for cotton yarn quality prediction. Cellulose 2023; 30(6): 4025–4039.

43.

Das

Chakraborty

Adaptive neuro-fuzzy inference system-based modelling of cotton yarn properties. J Inst Eng (India) E 2021; 102(2): 257–272.

44.

Farooq

Sarwar

Ashraf

, et al. Predicting cotton fiber maturity by using artificial neural network. Autex Res J 2018; 18(4): 429–433.

45.

Majumdar

Sarkar

An investigation on yarn engineering using artificial neural networks. J Text Inst 2006; 97(5): 429–434.

46.

Jayadeva Guha

Chattopadhyay

A study on the capability of a neural network ranking fiber parameters having an influence on yarn properties. J Text Inst 2003; 94(3–4): 186–193.

47.

Farooq

Cherif

Use of artificial neural networks for determining the leveling action point at the auto-leveling draw frame. Text Res J 2008; 78(6): 502–509.

48.

Yang

, et al. Recipe prediction of mélange yarn using modular artificial neural network. J Text Inst 2017; 109(5): 629–635.

49.

Farooq

Predicting the dynamic cohesion in drafted slivers at draw frame using artificial neural networks. Text Apparel 2014; 24(3): 286–290.

50.

Abd-Ellatif

SAM

. Optimizing sliver quality using artificial neural networks in ring spinning. Alexandria Eng J 2013; 52(4): 637–642.

51.

Farooq

Cherif

Development of prediction system using artificial neural networks for the optimization of spinning process. Fibers Polym 2012; 13(2): 253–257.

52.

Subramanian

Venkatachalam

Subramanian

Prediction and optimization of yarn properties using genetic algorithm/artificial neural network. Indian J Fiber Text Res 2007; 32(4): 409–413.

53.

Majumdar

Ciocoiu

Blaga

Modelling of ring yarn unevenness by soft computing approach. Fibers Polym 2008; 9(2): 210–216.

54.

Barzoki

Vadood

Johari

MS.

Modeling the properties of core-compact spun yarn using artificial neural network. J Text Polym 2016; 4(2): 101–105.

55.

Hossen

Ahmad

Ali

SM.

An application of Pareto analysis and cause-and effect diagram (CED) to examine stoppage losses: a textile case from Bangladesh. J Text Inst 2017; 108(11): 2013–2020.

56.

Farooq

Khan

Irshad

, et al. Predictive modeling of yarn quality at ring spinning machine using resilient back propagation neural networks. Text Apparel 2023; 33(1): 9–14.

57.

Ghanmi

Ghith

Benameur

Ring spun yarn quality prediction using hybrid neural networks. J Text Inst 2023; 114(1): 66–74.

58.

Abd-Elhamied

Hashima

ElKateb

, et al. Prediction of cotton yarn’s characteristics by image processing and ANN. Alexandria Eng J 2022; 61(4): 3335–3340.

59.

El-Geiheini

ElKateb

Abd-Elhamied

MR.

Yarn tensile properties modeling using artificial intelligence. Alexandria Eng J 2020; 59(6): 4435–4440.

60.

Doran

Sahin

The prediction of quality characteristics of cotton/elastane core yarn using artificial neural networks and support vector machines. Text Res J 2019; 90(13–14): 1558–1580.

61.

Kavitha

Ramya

Kannapiran

. ANN based model with PCA for fault diagnosis in ring spinning process. In: International conference on intelligent techniques in control, optimization and signal processing (INCOS), Tamilnadu, India, 11–13 April 2019. IEEE.

62.

Ghanmi

Ghith

Benameur

Ring yarn quality prediction using hybrid artificial neural network: Fuzzy expert system model. Int J Clothing Sci Technol 2015; 27(6): 940–956.

63.

Unal

Ozdil

Analysis of cotton ring spun yarn diameter using regression and artificial neural network. Industria Textila 2015; 66(6): 316–321.

64.

Almetwally

Idrees

HMF

Hebeish

AA.

Predicting the tensile properties of cotton/spandex core-spun yarns using artificial neural network and linear regression models. J Text Inst 2014; 105(11): 1221–1229.

65.

Mozafary

Payvandy

Application of data mining technique in predicting worsted spun yarn quality. J Text Inst 2014; 105(1): 100–108.

66.

Das

Ghosh

Majumdar

, et al. Yarn engineering using hybrid artificial neural network-genetic algorithm model. Fibers Polym 2013; 14(7): 1220–1226.

67.

Ghosh

Das

Banerjee

Multi objective optimization of yarn quality and fiber quality using evolutionary algorithm. J Inst Eng (India) E 2013; 94(1): 15–21.

68.

. Prediction of end breakage rates of cotton yarn in ring spinning processing by applying neural network approach and regression analysis. In: International conference of system science and engineering (IEEE), Dalian, China, 30 June–2 July 2012, pp. 555–558. IEEE.

69.

Soltani

Vadood

Johari

MS.

Modeling spun yarns migratory properties using artificial neural network. Fibers Polym 2012; 13(9): 1190–1195.

70.

. Predicting the end breakage rates of ring spinning polyester cotton yarn. In: Proceedings of the Fourth International Workshop on Chaos-Fractals Theories and Applications (IEEE), Hangzhou, China, 19–22 October 2011, pp.291–294. IEEE.

71.

Admuthe

Apte

Adaptive neuro-fuzzy inference system with subtractive clustering: a model to predict fiber and yarn relationship. Text Res J 2010; 80(9): 841–846.

72.

. Prediction of hairiness of polyester/viscose/cotton blended ring spinning yarn. In: Proceedings - 2010 international conference on optoelectronics and image processing, ICOIP 2010, Haikou, China, 11–12 November 2010, pp.131–134. IEEE.

73.

Furferi

Gelli

Yarn strength prediction: a practical model based on artificial neural networks. Adv Mech Eng 2010; 2: 640103.

74.

Wang

. Predicting yarn unevenness using improved BP Neural Network. Advanced Materials Research. 2011; 331: 219–222.

75.

Majumdar

Modeling of cotton yarn hairiness using adaptive neuro-fuzzy inference system. Indian J Fiber Text Res 2010; 35: 121–127.

76.

Admuthe

Apte

SD.

Neuro-genetic cost optimization Model: application of textile spinning process. Int J Comput Theory Eng 2009; 1(4): 1793–8201.

77.

Khan

Lim

AEK

Wang

, et al. An artificial neural network-based hairiness prediction model for worsted wool yarns. Text Res J 2009; 79(8): 714–720.

78.

Mwasiagi

Huang

Wang

Performance of neural network algorithms during the prediction of yarn breaking elongation. Fibers Polym 2008; 9(1): 80–86.

79.

Üreyen

Gürkan

. Comparison of artificial neural network and linear regression models for prediction of ring spun yarn properties. I. Prediction of yarn tensile properties. Fibers Polym 2008; 9(1): 87–91.

80.

Mwasiagi

Wang

Huang

XB.

Use of input selection techniques to improve the performance of an artificial neural network during the prediction of yarn quality properties. J Appl Polym Sci 2007; 108(1): 320–327.

81.

Chen

Zhang

Chen

, et al. A soft computing model for predicting yarn tenacity. Res J Text Apparel 2007; 11(4): 80–86.

82.

Beltran

Wang

Mill specific prediction of worsted yarn performance. J Text Inst 2006; 97(1): 11–16.

83.

Babay

Cheikhrouhou

Vermeulen

, et al. Selecting the optimal neural network architecture for predicting cotton yarn hairiness. J Text Inst 2004; 96(3): 185–192.

84.

Beltran

Wang

Predicting worsted spinning performance with an artificial neural network model. Text Res J 2004; 74(9): 757–763.

85.

Majumdar

Predicting the breaking elongation of ring spun cotton yarns using mathematical, statistical, and artificial neural network models. Text Res J 2004; 74(7): 652–655.

86.

Shanmugam

Chattopadhyay

Vivekanandan

MV.

Prediction of micro-spun lea CSP using artificial neural networks. Indian J Fiber Text Res 2001; 26: 372–377.

87.

Peirce

FT.

The geometry of cloth structure. J Text Inst Trans 1937; 28(3): T45–T96.

88.

Kumer

RS.

Process management in spinning. 1st ed. CRC Press, 2015.

89.

Karakan Günaydın

Çeven

. Investigation of comfort properties of knitted fabrics made of core-spun vortex yarns. J Eng Fiber Fabr 2023; 18: 1–22.

90.

Ghanmi

Ghith

Benameur

Prediction of rotor-spun yarn quality using hybrid artificial neural network-fuzzy expert system model. Indian J Fiber Text Res 2019; 44: 31–38.

91.

Erbil

Babaarslan

Ilhan

İ.

A comparative prediction for tensile properties of ternary blended open-end rotor yarns using regression and neural network models. J Text Inst 2017; 109(4): 560–568.

92.

Ghorbani

Vadood

Johari

MS.

Prediction of polyester/cotton blended rotor-spun yarns hairiness based on the machine parameters. Indian J Fiber Text Res 2016; 41: 19–25.

93.

Moghassem

Fallahpour

Shanbeh

An intelligent model to predict breaking strength of rotor spun yarns using gene expression programming. J Eng Fiber Fabr 2012; 7(2): 1–10.

94.

Pei

Prediction of the vortex yarn tenacity from some process and nozzle parameters based on numerical simulation and artificial neural network. Text Res J 2011; 81(17): 1796–1807.

95.

Shanbeh

Hasani

Tabatabaei

SA.

Modelling and predicting the breaking strength and mass irregularity of cotton rotor-spun yarns containing cotton fiber recovered from ginning process by using artificial neural network algorithm. Model Simul Eng 2011; 2011(1): 1–8.

96.

Predicting cotton yarn hairiness in rotor spinning. In: Proceedings - 2010 international conference on optoelectronics and image processing, ICOIP 2010, Haikou, China, 11–12 November 2010, pp.127–130.

97.

Ghosh

Chatterjee

Prediction of cotton yarn properties using support vector machine. Fibers Polym 2010; 11(1): 84–88.

98.

Majumdar

Comparative analysis of regression and artificial neural network models for the prediction of yarn hairiness. Res J Text Apparel 2010; 14(3): 85–93.

99.

Nurwaha

Wang

XH.

Prediction of rotor spun yarn strength from cotton fiber properties using adaptive neuro-fuzzy inference system method. Fibers Polym 2010; 11(1): 97–100.

100.

Demiryürek

Koç

Predicting the unevenness of polyester/viscose blended open-end rotor spun yarns using artificial neural network and statistical models. Fibers Polym 2009; 10(2): 237–245.

101.

Demiryürek

Koç

The mechanism and/or prediction of the breaking elongation of polyester/viscose blended open-end rotor spun yarns. Fibers Polym 2009; 10(5): 694–702.

102.

Majumdar

Sarkar

Application of an adaptive neuro-fuzzy system for the prediction of cotton yarn strength from HVI fiber properties. J Text Inst 2005; 96(1): 55–60.

103.

Majumdar

Sarkar

Application of linear regression, artificial neural network and neuro-fuzzy algorithms to predict the breaking elongation of rotor-spun yams. Indian J Fiber Text Res 2005; 30: 19–25.

104.

Jackowska-Strumillo

Jackowski

Cyniak

, et al. Neural model of the spinning process for predicting selected properties of flax/cotton yarn blends. Fibers Text East Eur 2004; 12(4): 16–21.

105.

Majumdar

Sarkar

Prediction of single yarn tenacity of ring-and rotor-spun yarns from HVI results using artificial neural networks. Indian J Fiber Text Res 2004; 29: 157–162.

106.

Jackowski

Cyniak

Chylewska

, et al. Investigating and modelling yarn parameters for different feeding slivers. Fibers Text East Europe 2002; 1(36): 49–51.

107.

Basu

Chellamani

Kumar

PR.

Application of neural network to predict the properties of air-jet spun yarns. J Inst Eng (India) 2002; 83: 1–4.

108.

Cheng

KPS

Lam

HLI

Kowloon

, et al. Evaluating and comparing the physical properties of spliced yarns by regression and neural network techniques. Text Res J 2003; 73(2): 161–164.

109.

Gramsch

Bell

Moghiseh

, et al. Analysis of the package diameter in winding processes by image analysis and a linear regression model. J Eng Fiber Fabr 2022; 17: 1–11.

110.

Das

Ishtiaque

Parida

JR.

Effect of fiber friction, yarn twist, and splicing air pressure on yarn splicing performance. Fibers Polym 2005; 6(1): 72–78.

111.

Moqeet

Jabbar

Hussain

, et al. Influence of splicing parameters on retained splice strength, elongation and appearance of spliced cotton/flax blended yarn. Indian J Fiber Text Res 2013; 38(1): 74–80.

112.

Ünal

Özdil

Taşkın

The effect of fiber properties on the characteristics of spliced yarns part I: prediction of spliced yarns tensile properties. Text Res J 2010; 80(5): 429–438.

113.

Ünal

Arikan

Özdil

, et al. The effect of fiber properties on the characteristics of spliced yarns: part II: prediction of retained spliced diameter. Text Res J 2010; 80(17): 1751–1758.

114.

Zhang

Zhao

, et al. A method for yarn quality fluctuation prediction based on multi-correlation parameter feature subspace mechanism in spinning process. J Eng Fiber Fabr 2023; 18: 1–12.

115.

ITMA Press Conference Presentation on Uster Statistics 2023. The Uster Technologies AG [internet], https://www.uster.com/fileadmin/user_upload/Landing_pages/Media_Kit_ITMA_2023/ITMA_Press_conference_presentation_Uster_Statistics_2023_final.pdf (2024, accessed 4 June 2024).

Application of Artificial Neural Networks in yarn manufacturing industry: A critical review

Abstract

Keywords

Introduction

Paper selection and review process

An overview of Artificial Neural Networks (ANNs)

Network selection

Optimizing network parameters

Data and performance estimation

ANN in the yarn manufacturing industry

Prediction in fiber properties

Prediction in blending, mixing, and sliver properties

Prediction in ring-spun yarn characteristics

Prediction in rotor- and vortex-spun yarn properties

Prediction in spliced yarn properties

Discussion and future perspectives

Conclusion

Footnotes

ORCID iDs

Funding

Conflicting interests

References