The dynamic transfer batch-size decision for thin film transistor–liquid crystal display array manufacturing by artificial neural-network

https://doi.org/10.1016/j.cie.2011.01.012Get rights and content

Abstract

In the thin film transistor–liquid crystal display (TFT–LCD) manufacturing process, array manufacturing is an important process. Transporting activities in array manufacturing are an important factor because of the frequent tasks. The transporting activity in array manufacturing is performed by an automated material handling system (AMHS). Automated guided vehicle (AGV) is the transporter used to carry glass substrates that are stored in a cassette. The capacity of a cassette is known as the transfer batch-size. Prior research of decisions in transfer batch-size, has addressed an optimal methodology, where one optimal transfer batch-size is assumed to have known conditions. However, in the volatile production environment, there may be multiple kinds of transfer batch-sizes. Therefore, we present an application of using a dynamic transfer batch-size strategy within a volatile production environment. In order to obtain the appropriate transfer batch-size for the current production environment, we adopt a neural-network based methodology as the core of the decision-making mechanism. This mechanism has the capability to identify the suitable transfer batch-size to allow an effective and efficient transportation under numerous conditions within the current production environment. This methodology is compared with the fixed transfer batch-size strategy in a real practical case. The results show that the dynamic transfer batch-size is superior to the fixed batch-size transportation.

Introduction

The liquid crystal display (LCD) industry adopts highly advanced manufacturing techniques and, in terms of wealth creation, is one of the important industries over recent years. In thin film transistor–liquid crystal display (TFT–LCD) manufacturing, there are three principal processes, specifically the array, Cell, and Module. The array manufacturing process, with the longest processing time and having the most production steps during manufacturing, is recognized as predominant in targeting efficiency and productivity gains.

The manufacturing process time consists of processing, waiting, and transporting. The waiting time between transferring batches is non-value-adding and one kind of waste needing reduction (Johnson, 2003). There are over 20 production stages connected by an automated material handling system (AMHS) in array manufacturing; of which the transporting time is a relatively high percentage of the total cycle time. Consequently, it is important to reduce materials handling activities.

AMHS consist of numerous components, including the batch container. Few researchers have investigated the effect of reducing transfer batch-sizes (Langevin et al., 1999, Liu, 2003, Veral, 1995); the tendency has been to calculate the optimal initial transfer batch-size. However, there is not an optimal transfer batch-size in the stochastic and dynamic environment. The transfer batch-size decision will be affected by the number of product types and also the utilization of workstations. Thus, the transfer batch-size is a decision factor affected by numerous variables. Chiu, Chang, and Lee (2004) considered that overall system efficiency would be better in variant transfer batch-sizes than for fixed transfer batch-sizes. Consequently, our research constructs a decision-making mechanism of dynamic transfer batch-sizes to accommodate the volatile production environment.

Due to unpredictable and complex interactions in the manufacturing environment, applying deterministic methods, such as mathematical programming, lacks empirical credibility. Therefore, we propose a neural based decision mechanism to solve the issue of dynamic transfer batch-size. To develop this model, production parameters were identified from a company-case. Adjustment to the level of environmental variables was made to reflect transfer batch-sizes, the volume and variance of WIP, throughput, utilization, the volume of input, and number of product types. Finally, we simulated these scenarios. By altering quantities in each scenario, a transfer batch-size giving the best performance for one scenario was identified. Iterations were conducted and a model was developed for the decision-making mechanism of dynamic transfer batch-size. The artificial neural-network (ANN) was trained using samples, and this formed a decision-making mechanism for identifying the suitable transfer batch-size. The neural based decision-making mechanism will suggest the appropriate transfer batch-size for current conditions based upon the previous production state (based on the variables described above).

Our research adopts an empirical case to evaluate the methodology; which draws comparison with a fixed transfer batch-size strategy to demonstrate improved effectiveness.

The remainder of this paper is organized as follows: Section 2 reviews the pertinent literature. Section 3 provides details of the proposed methodology. The background information for the case study is discussed in Section 4, and experimental analysis is presented in Section 5. Conclusions and future research opportunities are addressed in the final section.

Section snippets

Literature review

The definition of a transfer batch-size is the number of parts that are accumulated before being transferred to the next workstation (Hoop & Spearman, 2008). Therefore, the smaller the batch-size transferred, the lesser the time waiting for accumulation; and hence a reduced cycle time. However, it also has drawbacks, such as more transportation activities.

The key benefit of splitting to transfer batch is the reduction in the manufacturing cycle time for a job (Ramasesh, Fu, Fong, & Hayya, 2000

Methodology

This section provides a detailed description of the decision-making mechanism for deciding on a transfer batch-size using an ANN.

Empirical illustration

In order to demonstrate the superiority of applying the decision-making mechanism of a dynamic transfer batch-size in the TFT–LCD array manufacturing within an AMHS environment, the methodology is explained via a practical company-case. First, company-case is described.

Comparisons and sensitivity analysis

In this section, comparison is made with applying the dynamic transfer batch-size strategy to the performance of applying a fixed transfer batch-size. Next, sensitivity analyses are performed with respect to input-volume, number of product types, and load/unload time to illustrate the performance of the dynamic transfer batch-size strategy under difference production circumstances.

Conclusions

The ubiquitous application of TFT–LCD has driven the development and importance of its manufacturing in recent years. The design of the AMHS and cassette transportation operation makes management in TFT–LCD manufacturing operation more complex. Among the design of AMHS, the cassette, which is designed to prevent the glass substrates from spoiling due to pollutants, has a capacity limitation of 20 pieces. This upper-limit capacity restriction is a major constraint and greatly hinders any

Acknowledgments

The authors thank the anonymous company for providing the case study. This work was supported, in part, by the National Science Council of Taiwan, Republic of China, under Grant NSC-98-2221-E006-100-MY3.

References (29)

  • E. Veral

    Using production and transfer batches in flowshops under MRP control

    Journal of Operations Management

    (1995)
  • T. Yang et al.

    A neural network-based prediction model for fine pitch stencil printing quality in surface mount assembly

    Engineering Application of Artificial Intelligence

    (2005)
  • X.-H. Yu et al.

    Efficient backpropagation learning using optimal learning rate and momentum

    Neural Networks

    (1997)
  • G. Zhang et al.

    Forecasting with artificial neural networks: The state of the art

    International Journal of Forecasting

    (1998)
  • Cited by (6)

    • The multi-stage multi-product batch-sizing problem in the steel industry

      2020, Applied Mathematics and Computation
      Citation Excerpt :

      Wang and Guignard [17] proposed a new hybrid technique for a variant of the problem considered by Sarker et al. [16]. Yang and Shen [18] adopted a neural-network based methodology as the core of the decision-making mechanism to obtain the appropriate transfer batch-size for the current production environment. Shim et al. [19] studied a capacitated problem to determine the lot sizes and the sequence of lots on a single machine.

    • Median based conversion of SGPA into percentage by cognitive methods

      2015, Applied Mathematics and Computation
      Citation Excerpt :

      Some of the major applications of ANN in different fields of science and technology are as follows: The problem of wireless sensors network localization [12], Optimization of a RFID-based positioning system [19], Prediction of spark ignition engine performance for an alternative fuel [1], Predicting the mechanical properties of glass fiber reinforced polymers [8], Optimization of laser welding process parameters for super austenitic stainless steel [32], Forecasting of integrating multi-stage supply-chains [5], Prediction of surface roughness in hard turning process [27], Process synthesis of biodiesel production plant [6], Prediction of TiO2 nanotube morphology [39], Simulation of rate dependent behavior of adhesive materials [23], Detection of cyclic behavior in auto correlated process [28], and also [2,22,38] shows some other application of ANN in prediction and optimization problems. Genetic algorithms are one of the best ways to solve a problem for which little is known.

    • Preliminary cost estimates for thin-film transistor liquid-crystal display inspection and repair equipment: A hybrid hierarchical approach

      2012, Computers and Industrial Engineering
      Citation Excerpt :

      The TFT-LCD industry is second only to the semiconductor industry as a driver of economic growth in Taiwan. Because LCD production involves hundreds of complex processes, different panel manufacturers have different patented processes and production lines requiring highly customized manufacturing equipment, which is expected to increase complexity when producing varying sizes of next-generation panels (Yang & Shen, 2011). Despite significant advances in project management tools and techniques (PMI, 2004), project managers (PMs) in the engineer-to-order manufacturing industry still encounter problems associated with guesswork and subjective judgment that often result in inaccurate estimates.

    • Transfer batch size impact on a job shop environment performance

      2017, International Journal of Manufacturing Research
    View full text