Intelligent monitoring,diagnosis and control in mechanical engineering

Health monitoring, diagnostic and automatic control systems are commonly found in modern machinery and mechanical systems. Intelligent systems have a capability to acquire and apply knowledge in an intelligent manner and have the capabilities of perception, reasoning, learning and making decisions from incomplete information. Therefore, intelligent system approaches for monitoring, diagnosis and control can pave a practical way for a variety of mechanical engineering applications in the absence of human interaction. The objective of this Special Collection is to summarize the recent emerging development and ideas in intelligent monitoring, diagnostic and control systems and their applications for mechanical engineering.

After a rigorous review process, 18 out of a total of 52 submitted papers have been selected in this Special Collection, which fall into two main areas including (1) Intelligent monitoring and diagnosis and (2) Intelligent algorithms and control. A brief summary of the accepted papers in each area is also given below.

Intelligent monitoring and diagnosis

A fault diagnosis approach for the pitch system of wind turbines is developed by Wu and Liu. The proposed approach combines the interval prediction algorithm with a recursive subspace identification algorithm based on variable forgetting factor. This is a new idea. An improved method for detecting weak abrupt information based on permutation entropy is proposed by Shen et al. In many mechanical systems, the reason for signal abruption is possibly due to impact shock, speed fluctuation, structure deformation or fracture, which will lead to mechanical failure. An adaptive denoising fault feature extraction method based on ensemble empirical mode decomposition and correlation coefficient is proposed by Yang et al. In this study, various feature extraction techniques are applied to a neural-network-based fault classifier for comparison. Yao et al. present a new intelligent approach for diagnosis of railway rolling bearings using multi-scale intrinsic mode function permutation entropy and extreme learning machine classifier. It is well known that the working environment of the railway roller bearing is very harsh, so the bearings fail easily. It is believed that the proposed fault diagnostic system can enhance railway safety. An intelligent fault isolation method with expert guidance learning based on a regularization framework is proposed by Fan et al. In this article, the proposed method is verified by a robotic-arm-based spray marking process. Wu et al. propose an adaptive threshold algorithm for fault detection based on grey theory. In this method, the threshold can be dynamically adjusted according to the real-time mean and variance of the residual. The proposed method is validated through the fault detection of robot sensors. Xie and Chen study the fault pattern recognition of a gas blower based on a discrete Fourier transform interpolation algorithm. In noisy conditions, the proposed algorithm could achieve high precision, strong compatibility and robustness. Duan et al. propose an integrative intrinsic time-scale decomposition and hierarchical temporal memory approach for gearbox fault diagnosis under variable operating conditions. The effectiveness of the proposed approach is demonstrated by comparing neural-network-based approaches and fuzzy c-means clustering method. Jeong et al. design a simple monitoring system for welding spatter using a mobile phone, which is very useful to the welding industry.

Apart from the novel intelligent monitoring and diagnostic techniques, Pérez-Ruiz et al. evaluate existing gas turbine diagnostic techniques under variable fault conditions. In the study, multi-layer perceptron, radial basis network, probability neural network and support vector machines are compared intensively. Their conclusion shows that all the methods have their own pros and cons. In addition, Cao and Li conduct a survey on ambient energy sources and harvesting methods for structural health monitoring in which 97 recent articles are reviewed and discussed.

Intelligent algorithms and control

Jin et al. apply intelligent algorithms to improve the efficiency of the refilling process in a fast medicine dispensing system. Moreover, a comprehensive comparison of various intelligent algorithms is conducted and discussed. Yang et al. develop a random neural Q-learning algorithm for the obstacle avoidance of a mobile robot in unknown environments. The study shows that the proposed random neural Q-learning algorithm is superior to the Q-learning algorithm based on the conventional back-propagation method. A direct power control method for the three-phase pulse width modulation rectifier based on virtual flux orientation and a model reference adaptive vector controller for the induction motor without speed sensor are proposed by the research team of Fan et al. Both experimental results show that the proposed controllers can achieve good performances. Yang et al. develop a robust kernel-based model reference adaptive controller for unstable aircrafts. The results show that the proposed controller can achieve better identification and tracking performance. Han et al. design an output feedback controller for polynomial linear parameter-varying systems via parameter-dependent Lyapunov functions. By introducing the Lyapunov functions into the control framework, the stability of the control system can be guaranteed. Finally, Wei et al. investigate the dynamic properties and vibration control ability of a sandwich beam with magnetorheological fluid core. The results indicate that the active control ability of magnetic field is influenced by the axial force, moving speed and skin–core thickness ratio.

We hope that the readers will find the Special Collection interesting and stimulating, and expect that the involved papers can contribute to the further advance in the domain of intelligent monitoring, diagnosis and control in mechanical engineering.