A dispatching method for the lots of different priorities in 450-mm semiconductor manufacturing

Abstract

Semiconductor industry grows rapidly in recent years. It has an impact on the economies all over the world. For the mass production requirement, the wafer size starts to be increased from 300 to 450 mm. The technician cannot handle a large number of wafers since the size and weight of the wafers are increased. The automated material handling system has become the basic requirements of the wafer factory, and it is important to provide on-time delivery service and shorten cycle time to meet customer demand for different priority products. Hence, this study proposed an effective conveyor dispatching rule, which can reduce the handling delay of hot lots and detect the dispatching of lots with different priorities. The simulation experiments were conducted by building a 450-mm manufacturing environment and compared the nearest job first with conveyor dispatching rule. The results are sound. The conveyor dispatching rule can reduce 97.57% of the total average delivery variable time of hot lots and 0.28% of normal lots. This study proves that conveyor dispatching rule can efficiently improve productivity and reduce the cycle time of hot lots in every load configuration.

Keywords

Automated material handling system 450-mm wafer dispatching priority hot lots

Introduction

The semiconductor industry is becoming one of the most important industries in the world. By the concept of Moore’s Law,¹ semiconductor manufacturers will continue improving productivity, while reducing cycle times and costs. The cost of wafer manufacturing will reduce by increasing the wafer size. The International Semiconductor Technology Roadmap (ITRS)² predicts that the 450-mm wafers are key requirement for semiconductor manufacturers in the next generation of plants (NGF), and the ITRS expects a large number of 450-mm wafer fabrication tools to be available in 2016 and beyond (see Figure 1).

Figure 1.

ITRS roadmap of the 450-mm wafer generation.

Usually, the heavier 450-mm wafer fabs are shipped by hand, so the automated material handling system (AMHS) is more important in shipping, including automatic guided vehicles (AGVs), rail-guided vehicles (RGVs), overhead shuttles (OHSs), overhead lifts (OHTs), and conveyors in the fabs. For traditional working environment, the AGV is applied.³ In the 300-mm wafer fabs, the AMHS usually uses the OHT technology dispatching policies to transport the lots. However, instead of the existing vehicle-based AMHS, a conveyor-based AMHS is emerging as an effective alternative to provide high-speed, high-throughput deliveries in next-generation factories.^4,5

There are two main components of AMHS: one is the interbay loop transfer between production centers and the other is intrabay loop transfer within a production center. The interbay material handling system is centrally located and connects all the compartments. Figure 2 shows an intrabay loop, including sensors for detecting the finished production, conveyor track, and the load port of the device. The batches are moved along the conveyor track and continuously move in the same direction and then the batch arrives at the device and the device automatically captures the batch.

Figure 2.

Interbay and intrabay loops.

However, in the 450-mm semiconductor fab, heavier batch and process time increased, and some scholars will transport conveyor as the main transport tool. Nazzal et al.⁶ delivered a higher transmission policy, with shorter transit times (TTs), and lower cost than other traditional AMHS methods. As the space, equipment and carrier transmission is not blocked, the conveyor will be the next generation of the new concept of the plant AMHS.

Wang et al. proposed a Rotacaster heuristic preemptive dispatching method (R-HPD) to operate the AMHS for 450-mm wafer fabrication. The R-HPD method included some effective dispatching rules to control and dispatch wafer lots in the AMHS. The simulation results showed that the R-HPD outperformed the differentiated preventive dispatching policy in terms of average lot delivery time (ALDT). The improvements of 49.8% and 61.9% of ALDTs for hot lots and normal lots were observed, respectively.⁷ A heuristic preemptive dispatching method (HPDB) was proposed for controlling the movements of wafer lots in a convey-based AMHS which was restructured based on activated roller belt (ARB) and to be used for 450-mm wafer fabrication. The experimental results show that HPDB could better solve the traffic-jam problem and reduce transportation time.⁸ A dispatching rule called “heuristic preemptive dispatching rule using stocker (HPDS)” is proposed to control wafer movements in a conveyor-based AMHS for 450-mm wafer fabrication. Simulation has been conducted and the results showed that the HPDS performed well in terms of average variable time that includes block and waiting times.⁹ Wang et al. developed an effective dispatching method named the “conveyor preemptive dispatching” (CPD) method for conveyor-based AMHS to minimize the average variable time of hot lots with the minimum impact to the transport of normal lots. Experimental results revealed that the CPD demonstrates better performance than the existing method. The CPD method reduced the average variable time by 59% (from 14.572 to 5.904 s) for hot lots and 31% (from 3962.394 to 2741.889 s) for normal lots. It expedites the movement of lots and significantly reduces lot average variable time and will effectively enhance the throughput of lots transportation.¹⁰

While the conveyor-based AMHS systems maintain the advantages of existing transport systems, new systems can be improved with reference to studies of existing systems. Kong developed a two-step simulation for 300-mm factory. The capability of AMHS can be predicted in advance.¹¹ Kim and Park¹² proposed a policy of idle vehicle flow that affects AMHS productivity. A vehicle should be dispatched to somewhere if necessary even without a new assignment to prevent blocking of other vehicles. Wu et al.¹³ proposed an adaptive multi-parameter based (AMP) policy to get better cycle time, throughput, maturity satisfaction, and vehicle utilization when compared with traditional single-attribute and multi-attribute scheduling methods.

Wafer fabrication requires short cycle times and on-time delivery of production services to meet customer requirements. High-priority products are typically approved as heat lots in semiconductor fabs. The operation of the hot lot may be a capacity reserved for no waiting service or a greater profit. Some scholars pay attention to priority to deal with batch processing problems. Liao and Wang¹⁴ used neural network combined with heuristic method to deal with priority batch and delivery time prediction within 300 mm. In addition, an effective OHT scheduling rule has been developed; a differential priority scheduling (DPD) strategy has been developed to reduce the potential blocking effects during transport of heat lots in 300 isolation systems.¹⁵ Liang and Wang proposed an analytic methodology to estimate the loop-to-loop delivery time for differentiated lots in a 300-mm AMHS environment. Combining simulation and statistics techniques, the study developed a modularized simulation method (MSM) for delivery time forecast of priority lots. The time differences between MSM and simulation for both priority lots and regular lots were 0.2 and 0.1 s, respectively. Using the MSM method to forecast AMHS delivery time was a great contribution for streamlining shop floor operations in the 300-mm automatic environment.¹⁶ The simulation technique is used to verify the application of an interbay and several intrabays of the environment. Another efficient scheduling rule, named Heuristic Preemptive Scheduling (HPD), was developed to address this traffic jam in the AMHS environment. The HPD used a simulation method for a 300-mm wafer fabrication plant to analyze the problem and verify the results.¹⁷

In semiconductor industry, there are three major global originations in the world. They are ITRS,¹⁸ SEMATECH,¹⁹ and SEMI.²⁰ However, the latest roadmap of ITRS 2013 summary report¹ showed that the 450-mm factories will start from 2016. ITRS provides bi-year report from the whole group of global semiconductor major companies. Next, ITRS summary report of 2015 will be issued in 2016. The major semiconductor companies and originations all over the world are driving the technology by following this ITRS roadmap. The OHT is the major AMHS tool of 300-mm factories. For 450-mm AMHS, the Intel’s scientists,^4,5 several scholars,^6,7 and SEMATECH¹⁹ all suggested the conveyors could be a better choice for 450-mm AMHS. That’s why this research chooses conveyors to be our study for 450-mm AMHS.

The purpose of this study is to simplify the principle of the system. This study uses simulation software to effectively discuss the conveyor dispatching rules (CDRs), providing a good transport service for batches under a 450-mm semiconductor fab. The objective of this article is to provide a conveyor conveying system to minimize the transport delay of a 450-mm hot batch in AMHS.

The remainder of this article is organized as follows. Section “Research design and dispatch M” proposes the approach of this study. Experimental designs and result analysis based on realistic data are presented in section “Empirical design and results.” Finally, conclusions are given in section “Conclusion” with future research directions.

Research design and dispatch M

The transfer job is defined as a macro that transfers the command in three steps: first, when the process machine completes the process, the machine transfers the unprocessed batch signal to the material control system (MCS) and sends the batch to the conveyor. Next, the MCS finds the empty destination of the next destination for the batch and moves the batch to the destination. Third, if no empty destination machine is available, move the batch to the internal bay of the destination machine. Wait in the internal bay until the empty target machine is available and moved to this available machine. The purpose of the conveyor schedule is to minimize the lead time for the hot batch while minimizing the impact on the normal batch.

Analyzing conveyor operating and theoretical travel times, the study found that batch wait time is non-profit. High-priority products should have a lower priority of priority transport. In order to express our idea, this study proposes a heuristic conveyor scheduling method to speed up the movement of thermal ore to minimize the delivery time of the conveyor. The conveyor will continue to move in the same direction. Considering a set of transfer jobs transmitted by the conveyor, the available machines are assigned to the highest priority job. That is, once the highest priority job must retain the available machines to service the job, any other ongoing transfers on the conveyor before the highest priority job is completed. If such a highest priority job does not exist, all jobs are first satisfied, first for the normal transfer job.

The heuristic algorithm is designed and introduced in detail as below and demonstrated in Figure 3.

Figure 3.

Conveyor transport policies.

Overall policies

Step 1. On first come first served basis, the conveyor delivers the lot from interbay into intrabay and moves the lot to the nearest available processing machine.

Step 2. After a hot lot job is issued, this lot will follow the hot lot rule.

Step 3. Unprocessed lots are loaded into the machine by the intrabay conveyor.

Step 4. Processed lots are sent by the machines to unload into the intrabay conveyor for delivery.

Step 5. The intrabay sensor checks and determines whether the lot has been processed or not. When it is processed, this lot will be released to the interbay; otherwise, it will remain re-entrant in the intrabay to wait for processing.

Step 6. The interbay sensor checks and determines whether the lot has been completely processed or not. When this lot finishes the whole process, it will be moved to the warehouse; otherwise, it will return to Step 1 of the overall rule.

Hot lot policies

Step 1. The AMHS controller checks the location of all available machines and reserves the nearest available machine to process that unprocessed hot lot.

Step 2. If more than one unprocessed hot lot jobs are issued, first come first served is adopted.

Step 3. If there are not any available machines, the unprocessed hot lots continue moving in this intrabay.

Step 4. Unprocessed hot lots are loaded into the machine by the intrabay conveyor.

Step 5. Processed hot lots are sent by the machines to unload into the intrabay conveyor for delivery.

Step 6. The intrabay sensor checks and determines whether the hot lot has been processed or not. When it is processed, the hot lot will be released to the interbay; otherwise, it will remain re-entrant in the intrabay to wait for processing.

Step 7. The interbay sensor checks and determines whether the lot has been completely processed or not. When this hot lot finishes the whole process, it will be moved to the warehouse; otherwise, it will return to Step 1 of the hot lot rule.

Intrabay conveyors load untreated heat batches into the available machines. Then, go back to Step 6 of the overall strategy.

Empirical design and results

The main purpose of experiments is to compare the performances of different dispatching rules. Different rules need to compare in the same environment. However, according to the ITRS roadmap,² the 450-mm fabs could be available from 2016. To convey our idea of a 450-mm fab thermal-automated material handling service, we conducted a simulation based on actual data from a local 300-mm fab.^21,22 According to the international SEMATECH 450-mm guide, the dimensions of the machine are similar between 450 and 300 mm.²³ Therefore, this study assumes that the 450-mm fab process equipment specifications and 300-mm fab similar.

The simulation model in this study was implemented using the Flexsim simulation software,²⁴ a discrete event simulation package from the Canyon Park Technology Center (Orem, UT, USA). In semiconductor manufacturing, automated material transport systems are fairly complex systems. It is too complex to allow realistic simulations; therefore, this study attempts to simplify the SEMATECH ISMI 450-mm rule-based system.²³Figure 4 illustrates several objects defined in the simulation model.

Figure 4.

Objects defined in the simulation models.

There is an interbay loop and five intrabay loops. The steering sensor is set at the intersection to determine if the product is complete. If the product is complete, it is released to the next process. Otherwise, it returns to the original process.

This model is depicted in Figure 5.

Figure 5.

Simulation model.

Based on the information presented above, the simulation environment in this study is described as follows:

Machine processing time of 24 hours a day, 7 days a week, a total of 2 weeks, a total of 14 days. The first week is the warm-up time for this simulation.

In this simulation model, there are 69 sets of processors. They are assigned to five intrabays.

The factory has a 145-foot-long central channel throughout the simulated plant model. There are five intrabays, each walkway is 100-feet long. Each intrabay or interbay does not have any inventory.

The average conveyor speed is 1 ft/s. Load time = offload time = 5 s.

During the simulation, the conveyor and the equipment are free of faults and maintenance activities.

All conveyors move in the same direction. Therefore, the moving direction corresponds to the conveying direction.

Each loading or unloading machine can only carry out batch loading and unloading operations. If there are many previous load or unload operations being performed, the back of the batch must be stopped until these operations are complete.

Since the focus of this study is on the AMHS performance of conveyors, the relationship between the two processing units is used, rather than the overall process flow of the wafer product.

The arrival time of the transmission request is probabilistic and is assumed to be an exponential distribution.

This study analyzed two products: the first product is the normal batch and the second is the hot batch.

In order to achieve better performance, the factory always tries to reduce the volume lead time (LDT). This study defines the formula for calculating LDT, including TT, loading and unloading time (L&ULT), waiting time (WT), and blocking time (BT). The TT is the time for one lot to keep delivering non-stop since the start of the interbay until going out of this interbay. It does not include a batch to re-enter an intrabay time. The L&ULT is the time to load unprocessed batches into the empty machine plus the time to unload the processed batch from the machine and move it to the conveyor. WT is the time for a batch to re-enter an intrabay because it has not completed processing in this intrabay. BT is the dead time for a batch due to loading or unloading of other batches. In theory, the TT and L&ULT can be regarded as a fixed time while the WT and BT are variable times and they can be reduced by a good dispatching method. The main purpose of experiments is to compare the performances of different dispatching rules. A good rule can reduce the WT and BT. Therefore, a good rule can shrink the lot delivery time. So, this research takes lot variable time (LVT) as the main index of performance to demonstrate the efficient result. Equation (1) defines the formula for calculating the LVT, which includes WT and BT

LVT = WT + BT

(1)

If there is one batch, the average batch variable time (ALVT) can be determined by equation (2), where i represents the ith batch

ALVT = \frac{{\sum_{i = 1}^{l} LVT}_{i}}{l}

(2)

This study observed the dynamics of the conveyor system. A heavy-duty system is used to highlight the impact of resource contention and the increase in the hot batch ratio will add significant latency to the normal batch. Due to the dynamics of the material handling system based on the conveyor, for each intrabay j this study uses two dominant control variables, the bottleneck module loading ratio $(R_{BL} (j))$ and the thermal batch ratio $(R_{h} (j))$ to simulate the intrabay of the current situation.

The higher $R_{BL} (j)$ and $R_{h} (j)$ are the causes of traffic jams in intrabay j. A heavily loaded system is used to highlight the impact of heuristic rules in a finite resource environment.

Equation (3) defines the formula for $R_{BL} (j)$ used to calculate the internal bottleneck j, which is the average of the hourly input batches divided by the maximum number of output batches per hour

R_{BL} (j) = \sum_{i = 1}^{m} \frac{3600 s}{S T_{i} + P T_{i} + L & UL T_{i}} machine i (i = 1, \dots, m) \in Intrabay j

(3)

where j is the bay j, m is the total number of machines in intrabay j, $S T_{i}$ is the setup time of machine i, $P T_{i}$ is the processing time of machine i, and $L & UL T_{i}$ is the loading and unloading time of machine i.

This study calculates the loading of each intrabay’s tank to find the intrabay with the smallest cabin load ratio; it will be the bottleneck of all five intrabays. Each calculation will follow this minimum load ratio. The bay-loading time of an intrabay is calculated as 1 h in seconds.^21,22

In this research, the simulation uses three bottleneck loading ratios or 80%, 90%, and 100% for the design specification. Equation (4) defines the formula used to calculate $R_{h} (j)$ , which is the average number of hot lots (ANHL) divided by the average number of lots (ANL) in platform j. The test in this study used three hot batch ratios of 4%, 8%, and 12% because the increased thermal batch ratio imposes a severe time delay on the normal batch

R_{h} (j) = \frac{ANHL}{ANL}

(4)

Each experiment was performed three times. The total number of simulations carried out was 3 (hot batch ratio) × 3 (tank loading ratio) × 3 (copy) = 27. The simulation range was set to 14 days long and 7 days preheat for each experiment. Plant simulations usually require a warm-up time because your simulation may begin to empty (no product on any machine), but the actual plant does not start to work on a weekly basis. Therefore, the results of this study do not include warm-up time. The results were ALVT, with time in seconds, excluding warm-up time.

In this study, the CDR method was discussed and compared with the most recent work (nearest job first (NJF)), developed by Liao and Fu²⁵ from two control variables. Results from different cabin loads were discussed and analyzed. The experimental results are demonstrated in Table 1 and Figure 6. Because the NJF policy does not have the ability to distinguish the movement of batches with different priorities, the results of the hot batch and the normal batch are the same in different configurations.

Table 1.

Experiment results in variable time of the lot delivery time (in seconds).

System configuration		ALVT (in seconds)
Bay loading	Hot ratio	CDR		NJF
Bay loading	Hot ratio	Hot lots	Normal lots	Normal lots
80%	4%	18.52	234.88	256.09
	8%	23.48	252.14	260.41
	12%	28.17	278.36	252.77
Average of 80%		23.39	255.12	256.42
90%	4%	22.86	452.55	509.61
	8%	27.26	505.87	511.44
	12%	34.30	556.22	503.27
Average of 90%		28.14	504.88	508.10
100%	4%	36.84	3028.20	3201.94
	8%	43.26	3273.42	3241.88
	12%	56.80	3403.29	3281.73
Average of 100%		45.63	3234.97	3241.85
Average for all		32.39	1331.66	1335.46

ALVT: average batch variable time; CDR: conveyor dispatching rule; NJF: nearest job first.

Figure 6.

Simulation average results in different configurations.

This study favors the CDR approach to perform better in reducing the average variable delivery time of the heat batch. For 80%, 90%, and 100%, the CDR method reduced the hot batch ALVT by 90.88% (from 256.42 to 23.39 s), 94.46% (from 508.10 to 28.14 s), and 98.59% (from 3241.85 down to 45.63 s) for 80%, 90%, and 100% bay loading, respectively. Compared to the NJF strategy, the CDRs reduced the hot lot ALVT by 97.57% (for all system configurations) and reduced the ALVT of normal lots by 0.51% (from 256.42 to 225.12 s), 0.64% (from 508.10 to 504.88 s), and 0.21% (from 3241.85 to 3234.97 s), for 80%, 90%, and 100% bay loading, respectively. The results of high bay loading (100%) are convergent. This means the system cannot afford the 100% high bay loading. The CDR reduces the ALVT of normal lots by 0.28% for all system configurations. Obviously, the results can indicate that the ALVT of hot lots is reduced significantly. The results are in line with our expectations. The simulation experiments show that in the 450-mm wafer foundry environment, CDR is a good way to reduce the delivery time of priority products.

Conclusion

It is clear that the size of the wafers increases with the number of batches and the number and complexity of process steps for 450-mm wafer fabrication. Manual delivery of this chip is too difficult. As a result, effective AMHS becomes more important. The conveyor system will be one of the best options for the next generation of 450-mm manufacturing, thanks to the higher transport capacity and the outstanding advantages of the local buffer for the machine. However, it is the same as other systems; conveyor-based AMHS also faces problems with traffic congestion when there are so many loops in the loading or unloading process. In addition, a higher priority segment should have greater traffic privileges than a lower-priority lot. The key issue is how to deliver almost no-wait shipments for hot batches in an automated material handling environment. Therefore, a good scheduling rule for the AMHS system is very helpful.

Based on empirical data, and taking into account the effects and limitations of shipping when applied in 450-mm wafers in 300-mm wafer fabrication, this study proposes a CDR method to minimize the TT for hot batches with minimal impact on normal batch delivery. Compared to NJF policy in 450-mm wafer fabrication, the results demonstrate that CDR policy reduced the average variable time of hot lots by 97.57% (from 1335.46 to 32.39 s) and reduced the average variable delivery time of normal lots by 0.28% (from 1335.46 to 1331.66 s). The results are sound for reducing the average variable time of hot lots with less impact for normal lots. This method will reduce the cycle time and increase the throughput of shop flow control in factory.

The wafer size transition from 300 to 450 mm is a major turning point and the development trends of the semiconductor industry. The replacement of hardware devices faces many difficulties due to cost and huge production scale, so the development of new dispatching methods and algorithms is more focused with final purposes to reduce average delivering time of lot and minimize bottleneck problem. Today, conveyor needs to fulfill the demanding of 450-mm wafer handling.

Because the hot lots have high priority than normal lots, the unprocessed hot lot should be handled with high priority than unprocessed normal lot; previous researchers have handled this problem quite well by applying dynamics algorithms and rules. In addition to inheriting the previous achievements, this study proposes new principles with the aim of reducing the processing time and transport time of lots, especially for hot lots. The new rule is based on the difference of setup time and processing time of machine in intrabay. The machines have short setup time and processing time have high priority to handle hot lot. This research uses new algorithms and rules to help AMHS system more dynamics, flexible and adaptable. The results of this method not only reduce lot delivery time but also deal with the bottleneck very well.

This study proposes the CDR algorithm for solving some material handling problems. As a matter of fact in processing, CDR already dealing with these problems reduces the delivery time of all products, guarantees the delivery time of high-priority products, gets higher performance of systems, and makes system more flexible for transport or processing. CDR completely provides the optimization method and higher performance for production process solution. Moreover, CDR is absolutely compatible for manufacturer, where the requirement quantity of products continuously rise day by day.

Future studies will consider higher product mixes leading to more frequent process changes, or strategies for scheduling multi-level transmissions, automated scheduling strategies for segmented two-track bidirectional loops (SDBLs), or extended simulation models to full-size AMHS applications. In addition, the direction may also consider CDR methods with different speeds in 450-mm wafers or modify the CDRs by combining them with a reservoir. In the future, it can meet the needs of transmission of more different types of products and priorities. In addition, future research will also consider expanding the simulation model in this research to full-scale AMHS applications. Moreover, the calculation formulas will be further studied for the AMHS of 450-mm environment. More mathematical models will be developed for 450-mm material handling policy.

Footnotes

Academic Editor: Kuei Hu Chang

Author Note

Other corresponding authors for this paper are: Min-Tsong Chou, National Kaohsiung University of Applied Sciences, 80778 Kaohsiung, Taiwan. Email: iahucio@yahoo.com.tw

Ruei-Yuan Liao, National Sun Yat-sen University, 80424 Kaohsiung, Taiwan. Email: n954050002@student.nsysu.edu.tw

Chung-Jen Huang, National Kaohsiung University of Applied Sciences, 80778 Kaohsiung, Taiwan. Email: cjhuang@gm.kuas.edu.tw

Declaration of conflicting interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported, in part, by MOST 105-2221-E-151-039 and MOST 105-2622-E-151-006-CC3 from the Ministry of Sciences and Technology. The authors also appreciate the support from National Kaohsiung University of Applied Sciences in Taiwan.

References

Schaller

RR.

Moore’s law: past, present and future. IEEE Spectrum 1997; 34: 52–59.

International Technology Roadmap for Semiconductors. Executive summary, http://www.itrs2.net/itrs-reports.html (2014, accessed 12 September 2014).

Qudeiri

Umer

Khadra

. Layout design optimization of dynamic environment flexible manufacturing systems. Adv Mech Eng 2015; 7: 1–9.

Pettinato

Pillai

Technology decisions to minimize 450-mm wafer size transition risk. IEEE T Semiconduct M 2005; 18: 501–509.

Pillai

The future of semiconductor manufacturing. IEEE Robot Autom Mag 2006; 13: 16–24.

Nazzal

Jimenez

Carlo

. An analytical model for conveyor-based material handling system with crossovers in semiconductor wafer fabs. IEEE T Semiconduct M 2010; 23: 468–476.

Wang

Hsu

. Using Rotacaster in the heuristic preemptive dispatching method for conveyor-based material handling of 450 mm wafer fabrication. IEEE T Semiconduct M 2016; 29: 230–238.

Wang

Lee

Hsu

. The heuristic preemptive dispatching method for convey-based automated material handling system of 450 mm wafer fabrication. Comput Ind Eng 2016; 96: 52–60.

Wang

Lee

Hsu

. The heuristic preemptive dispatching rule using stocker for conveyor-based material handling system of 450-mm wafer fabrication. Int J Adv Manuf Tech 2017; 88: 2179–2189.

10.

Wang

Chung

Wang

. The material dispatching method for conveyor system in 450 mm wafer fabrication. J Test Eval 2017; 43: 1–13.

11.

Kong

SH.

Two-step simulation method for automatic material handling system of semiconductor fab. Robot Cim: Int Manuf 2007; 23: 409–420.

12.

Kim

Park

Idle vehicle circulation policies in a semiconductor FAB. J Intell Manuf 2009; 20: 709–717.

13.

Mok

Zhang

An adaptive multi-parameter based dispatching strategy for single-loop interbay material handling systems. Comput Ind 2011; 62: 175–186.

14.

Liao

Wang

CN.

Neural-network-based delivery time estimates for prioritized 300 mm automatic material handling operations. IEEE T Semiconduct M 2004; 17: 324–332.

15.

Liao

Wang

CN.

Differentiated preemptive dispatching for automatic materials handling services in 300 mm semiconductor foundry. Int J Adv Manuf Tech 2006; 29: 890–896.

16.

Liang

Wang

CN.

Modularized simulation for lot delivery time forecast in automatic material handling systems of 300 mm semiconductor manufacturing. Int J Adv Manuf Tech 2005; 26: 645–652.

17.

Wang

Chen

LC.

The heuristic preemptive dispatching method of material transportation system in 300 mm semiconductor fabrication. J Intell Manuf 2012; 23: 2047–2056.

18.

International Technology Roadmap on Semiconductor (ITRS), http://www.itrs2.net/ (2015, accessed 1 December 2015).

19.

SEMATECH, http://public.sematech.org (2015, accessed 1 December 2015).

20.

SEMI, http://www.semi.org (2015, accessed 1 December 2015).

21.

Quinn

Bass

300 mm factory layout and material handling modeling: phase I report. Report, International SEMATECH, Austin, TX, 31 March 1999.

22.

Campbell

Ammenheuser

300 mm factory layout and material handling modeling: phase 2 report. Report, International SEMATECH, Austin, TX, 30 June 2000.

23.

International SEMATECH Manufacturing Initiative. ISMI 450 mm guidelines, http://www.ismi.sematech.org/research/450/docs/ISMI450mmGuidelines.pdf (2009, accessed 12 September 2014).

24.

Flexsim Software, http://www.flexsim.com (2014, accessed 8 September 2014).

25.

Liao

HS.

Speedy delivery—dynamic OHT allocation and dispatching in large-scale, 300-mm AMHS management. IEEE Robot Autom Mag 2004; 11: 22–32.