Rong-Kwei Li

Process capability analysis for an entire product

Process capability indices (PCIs) are powerful means of studying the process ability for manufacturing a product that meets specifications. Several capability indices including C p , C pu , C pl and C pk have been widely used in manufacturing industry to provide common quantitative measures on process potential and performance. The formulas for these indices are easily understood and can be straightforwardly applied. However, those process capability indices are inappropriate for asymmetric tolerances and could not be applied to evaluate multiprocess products. Based on C p , C pu , C pl and C pk , this research aims to develop one process capability analysis chart (PCAC) for precisely measuring an entire product composed of symmetric tolerances, asymmetric tolerances, larger-the-better and smaller-the-better characteristics. The process capability analysis chart evaluates the capabilities of multi-process products and provides chances for continuous improvement on the manufacturing process.

A heuristic rescheduling algorithm for computer-based production scheduling systems

Generation of new production schedules in response to changes that take place on a factory shop-floor is a common requirement in todays complex and flexible manufacturing environment. The current approaches use either a regeneration rescheduling method, which results in unsatisfactory response times or methods that need too much manual intervention for identifying the operations to be changed. We propose a new rescheduling algorithm to address these problems. Our algorithm is based on the construction of a scheduling binary tree and a net change concept adopted from MRP system. It revises schedules by rescheduling only those operations that need to be revised. Our algorithm can be embedded in existing simulation-based scheduling systems or electronic Gantt chart approaches to improve their effectiveness.

A novel means of generating assembly sequences using the connector concept

Assembly planning attempts to arrange product assembly sequences on the basis of the input of product description, similar assembly plans and assembly constraints. Assembly planning is frequently viewed as a creative thinking process, which requires extensive assembly planning experience. Assembly product modeling with mechanical connection functions has increasingly focused on assembly sequence planning. When considering the connection functions, the relationships between assembled components can act as a foundation for assembly clustering. Despite the merits of assembly sequence planning in lieu of considering connection functions, previous investigations have focused primarily on product representation with the connector concept. However, to our knowledge, generation of the assembly sequence on the basis of the connector concept has not been previously discussed. In this study, we present a novel means of generating assembly sequences on the basis of the connector concept. The proposed method initially defines different types of connectors and representation schemes. Through the definition and representation scheme, an assembly product can be decomposed into a set of connector-based assembly elements. The connector-based assembly elements then serve as input for an assembly sequence generation algorithm, which is connector-based, to generate an assembly sequence. The final output is a connector-based assembly sequence graph, resembling an assembly precedence diagram. In addition, this graph can act as the input for any line balance method when designing work elements for each assembly workstation.

Heuristic PAC model for hybrid MTO and MTS production environment

Abstract Two distinct types of semiconductor plants in Taiwan are integrated device manufacturing (IDM) plants and foundry plants. Most IDM plants are make-to-stock (MTS) operations, focusing on throughput and machine utilization. However, foundry plants are make-to-order (MTO) operations, focusing on due date and cycle time. Besides the challenge of different process technology, the mode of hybrid operation (a combination of MTO and MTS operations) is also a formidable task for these plants. This study develops a heuristic production activity control model to achieve the two different criteria in a hybrid wafer production environment.

Enhancement of Theory of Constraints replenishment using a novel generic buffer management procedure

Managing a distribution system requires the right inventory in the right place at the right time. A Theory of Constraints replenishment solution is presented to aggregate inventory buffers at the central warehouse in plant and change the mode of operation from push to pull. The solution is powerful, but the optimal amount of buffer remains undetermined. In Theory of Constraints, choosing a specific buffer size is not crucial if the buffer is accurately monitored in a timely manner. Accordingly, Theory of Constraints is offered a buffer management approach for monitoring the buffer. Such buffer management is feasible and effective but is insufficiently rigorous. This paper elucidates a generic buffer management procedure, based on the concept of Theory of Constraints buffer management that rigorously defines a method of monitoring to size and adjust the buffer. An example demonstrates the feasibility of the proposed generic buffer management procedure.

The construction of production performance prediction system for semiconductor manufacturing with artificial neural networks

The major performance measurements for wafer fabrication system comprise WIP level, throughput and cycle time. These measurements are influenced by various factors, including machine breakdown, operator absence, poor dispatching rules, emergency order and material shortage. Generally, production managers use the WIP level profile of each stage to identify an abnormal situation, and then make corrective actions. However, such a measurement is reactive, not proactive. Proactive actions must effectively predict the future performance, analyze the abnormal situation, and then generate corrective actions to prevent performance from degrading. This work systematically constructs artificial neural network models to predict production performances for a semiconductor manufacturing factory. An application for a local DRAM wafer fabrication has demonstrated the accuracy of neural network models in predicting production performances.

Integrated scheduling of production and delivery with time windows

This paper deals with an integrated scheduling problem in which orders have been processed by a distribution centre and then delivered to retailers within time windows. We propose a nonlinear mathematical model to minimise the time required to complete producing the product, delivering it to retailers and returning to the distribution centre. The optimal schedule and vehicle routes can be determined simultaneously in the model. In addition, two kinds of genetic-algorithm-based heuristics are designed to solve the large-scale problems. The conventional genetic algorithm provides the search with a high transition probability in the beginning of the search and with a low probability toward the end of the search. The adaptive genetic algorithm provides an adaptive operation rate control scheme that changes rate based on the fitness of the parents. The experimental results have shown that the solution quality of these two algorithms is not significant but that the adaptive genetic algorithm can save more time in finding the best parameter values of the genetic algorithm.

A new ART-counterpropagation neural network for solving a forecasting problem

This study presents a novel Adaptive resonance theory-Counterpropagation neural network (ART-CPN) for solving forecasting problems. The network is based on the ART concept and the CPN learning algorithm for constructing the neural network. The vigilance parameter is used to automatically generate the nodes of the cluster layer for the CPN learning process. This process improves the initial weight problem and the adaptive nodes of the cluster layer (Kohonen layer). ART-CPN involves real-time learning and is capable of developing a more stable and plastic prediction model of input patterns by self-organization. The advantages of ART-CPN include the ability to cluster, learn and construct the network model for forecasting problems. The network was applied to solve the real forecasting problems. The learning algorithm revealed better learning efficiency and good prediction performance. q 2004 Elsevier Ltd. All rights reserved.

A new rescheduling method for computer based scheduling systems

Providing a powerful interactive tool for the scheduler to quickly and easily react to the inevitable rescheduling changes is mandatory in todays complex and flexible manufacturing environment. The conventional approaches either employ the regeneration method, which results in unsatisfactory response times, or use. methods that require too much manual intervention for editing operations to be changed. A new rescheduling method is proposed in this paper which addresses these problems. The fundamental scheme of this rescheduling method is based on a scheduling graph as well as concepts of time effect and relationship effect. The scheduling graph is an alternative representation structure of a schedule Gantt chart. The time effect and relationship effect provide functions of (1) identifying those operations that require revision, (2) revising those identified affected operations (via a partial change of the scheduling graph structure) and (3) updating starting and ending times of those revised operations. T...

Exact solution of inventory replenishment policy for a linear trend in demand - two-equation model

Abstract This study presents a model for obtaining the exact optimal solution of inventory replenishment policy problem. The proposed model, called the “two-equation model”, includes two governing equations and a time-frequency algorithm: the first equation determines the optimal replenishment times for a specified number of replenishments; the second equation determines the optimal number of replenishments. The time-frequency algorithm includes two main procedures to solve the first and second equations. In contrast to many approximation approaches, this model exactly solves the first equation, a simultaneous non-linear equations system using a generalized matrix-based solver. This study also examines the classical no-shortage inventory replenishment policy for linear increasing and decreasing trend demand. According to these results, the solution given by the two-equation model is an exact solution.