A knowledge-based multi-agent evolutionary algorithm for semiconductor final testing scheduling problem

doi:10.1016/j.knosys.2015.03.024

Knowledge-Based Systems

Volume 84, August 2015, Pages 1-9

https://doi.org/10.1016/j.knosys.2015.03.024 Get rights and content

Abstract

The final testing process ensures the quality of the products in the semiconductor manufacturing factory. Scheduling for the final testing process is crucial to the economic efficiency of production. In this paper, an effective knowledge-based multi-agent evolutionary algorithm (KMEA) is proposed for solving the semiconductor final testing scheduling problem (SFTSP). In the KMEA, each agent is represented by a solution, which is a combination of the operation sequence vector and the machine assignment vector. A hybrid initialization mechanism is proposed to balance the diversity and the quality of the initial agents. In each iteration of evolution, the agents evolve by mutual-learning and competition based on the model of agent lattice. Moreover, a knowledge base is employed to store the useful information during the search process. The knowledge base is used to generate new agents in the competition phase. The computational complexity of the KMEA is analyzed, and the influence of parameter setting is also investigated. Finally, numerical simulation and comparisons are provided to demonstrate the effectiveness and efficiency of the KMEA in solving the test instances.

Introduction

To ensure the quality of integrated circuit (IC) products, the products are tested in the semiconductor final testing process after the completion of wafer fabrication. Usually, the testing machines are not many in a semiconductor testing factory because of their high prices. As an essential function in production management, an effective scheduling is important to improve the utilization of resources [1], [2]. Thus, the semiconductor final testing scheduling problem (SFTSP) becomes an important issue to the semiconductor manufacturing companies. In addition, the SFTSP is a specific type of simultaneous multiple resources scheduling problem, which has been proved to be NP-hard [3]. Therefore, it is of significant importance to study the SFTSP in both academic field and engineering field, especially to develop effective and efficient solution algorithms.

As an effective technology of distributed artificial intelligence, the agent-based optimization has gained wide applications in the field of evolutionary computation during recent years. Although the evolutionary algorithms (EAs) have strength on complex optimization [4], [5], the EAs based on multi-agent system have the ability of performing better than the classical EAs [6]. Zhong et al. [7] proposed a multi-agent genetic algorithm (MAGA) by integrating GA and multi-agent system for solving the global numerical optimization problem. In their MAGA, each agent represented a candidate solution of the problem to be solved, and all the agents were fixed on the points of a lattice. In each generation, the agents competed or cooperated with their neighbors to obtain new better solutions. Based on the same model of agent lattice [7], Liu et al. proposed a multi-agent evolutionary algorithm for the constraint satisfaction problems [8] and the combinatorial optimization problems [9], respectively. By combining agents and quantum-bit, Tao et al. [10] proposed a quantum multi-agent evolutionary algorithm for partner selection problems in a virtual enterprise. Zeng et al. [11] presented a multi-agent evolutionary algorithm for assembly sequence planning based on the geometry and assembly process information of product.

Inspired by the above successful applications of the multi-agent EAs, this paper aims to propose a knowledge-based multi-agent evolutionary algorithm (KMEA) for solving the SFTSP. The innovation in designing the algorithm can be summarized as follows: (1) A novel initialization mechanism is proposed to ensure the diversity and quality of the initial solutions; (2) Some effective search operators are designed to explore better solutions in the mutual-learning phase and competition phase; (3) A knowledge base is designed to store the useful information during the search process and to generate promising solutions in the competition phase. In addition, we analyze the computational complexity of the KMEA and investigate the influence of parameter setting. Numerical simulation and comparisons show that the KMEA is more effective and efficient than the state-of-the-art in solving the test instances.

The remainder of the paper is organized as follows: Section 2 presents the literature review about SFTSP and Section 3 describes the SFTSP briefly. Then, the KMEA for solving the SFTSP is introduced in Section 4. The influence of parameter setting is investigated in Section 5, and the computational results and comparisons are provided as well. Finally the paper ends with some conclusions in Section 6.

Section snippets

Literature review

Scheduling for the semiconductor final testing process is one of the most complex and common scheduling problems [12]. To develop a suitable scheduling system, Uzsoy et al. [13] divided the facility or job shop into a number of work-centers and then sequenced one work-center at a time. Besides, the authors used the disjunctive graph representation of the entire facility to capture interactions between work-centers. For the single-machine scheduling problems with sequence-dependent setup times

Problem description

The SFTSP can be described briefly as follows. There are n jobs (IC products) J = {J₁, J₂, … , J_n} to be tested on m machines M = {M₁, M₂, … , M_m}. The testing process of job J_i is to perform a sequence of n_i operations ${O_{i, 1}, O_{i, 2}, \dots, O_{i, n_{i}}}$ according to a given sequence. The operations include functional test, burn-in, scan, bake, tape and reel, package and load. The processing time of O_i_,_j on machine M_k is t_ijk. Due to different product specifications, different jobs may require different steps of

Knowledge-based multi-agent evolutionary algorithm

In this section, the KMEA for solving the SFTSP is introduced in details. First, the flowchart of the KMEA is presented. Then, the solution representation, initialization mechanism, mutual-learning phase, and competition phase are introduced. Besides, the computational complexity of the KMEA is analyzed.

Numerical results and comparisons

In the literature [3], [12], [20], [21], [22], a set of instances (LS1–LS5 and WR1–WR5) were used to test and compare the performances of the algorithms for solving the SFTSP (available at website http://dalab.ie.nthu.edu.tw/newsen_content.php?id=0). In this paper, these instances are also used to carry out numerical test. The parameters of the instances are listed in Table 3.

Conclusions

In this paper, a knowledge-based multi-agent evolutionary algorithm named KMEA is proposed to solve the semiconductor final testing scheduling problem. Considering the characteristics of the problem, the solution as an agent is represented by an operation sequence vector and a machine assignment vector. With the lattice model, agents learn and compete with each other using problem-specific search operators in the mutual-learning phase and the competition phase. In addition, a hybrid

Acknowledgments

The authors would like to thank the Editor-in-Chief and the anonymous reviewers for their constructive comments to improve the paper. This research is partially supported by National Key Basic Research and Development Program of China (Grant No. 2013CB329503), National Science Foundation of China (Grant No. 61174189), and Doctoral Program Foundation of Institutions of Higher Education of China (Grant No. 20130002110057).

References (27)

R. Zhang et al.
A two-stage hybrid particle swarm optimization algorithm for the stochastic job shop scheduling problem
Knowl.-Based Syst.
(2012)
H. Karimi et al.
An efficient knowledge-based algorithm for the flexible job shop scheduling problem
Knowl.-Based Syst.
(2012)
F. Tao et al.
A quantum multi-agent evolutionary algorithm for selection of partners in a virtual enterprise
CIRP Ann.-Manuf. Technol.
(2010)
J.Z. Wu et al.
Modeling semiconductor testing job scheduling and dynamic testing machine configuration
Expert. Syst. Appl.
(2008)
W.L. Pearn et al.
A case study on the multistage IC final testing scheduling problem with reentry
Int. J. Prod. Econ.
(2004)
X.L. Zheng et al.
A novel fruit fly optimization algorithm for the semiconductor final testing scheduling problem
Knowl.-Based Syst.
(2014)
X.C. Hao et al.
The cooperative estimation of distribution algorithm: a novel approach for semiconductor final test scheduling problems
J. Intell. Manuf.
(2014)
H. Jabeen et al.
Review of classification using genetic programming
Int. J. Eng. Sci. Technol.
(2010)
H. Jabeen et al.
Opposition based initialization in particle swarm optimization (O-PSO)
L. Jiao et al.
Cooperative Evolutionary Computation and Multi-Agent System
(2006)

W. Zhong et al.

A multiagent genetic algorithm for global numerical optimization

IEEE Trans. Syst. Man. Cybernet. B

(2004)

J. Liu et al.

A multiagent evolutionary algorithm for constraint satisfaction problems

IEEE Trans. Syst. Man. Cybernet. B

(2006)

J. Liu et al.

A multiagent evolutionary algorithm for combinatorial optimization problems

IEEE Trans. Syst. Man. Cybernet. B

(2010)

Cited by (29)

A greedy-based crow search algorithm for semiconductor final testing scheduling problem
2023, Computers and Industrial Engineering
The semiconductor final testing scheduling problem (SFTSP) is of great importance to the efficiency of integrated circuit firms and has been widely investigated in the field of intelligent optimization. In this paper, a greedy-based crow search algorithm (GCSA) is presented for solving the SFTSP. According to the characteristics of SFTSP, new encoding and decoding strategies are proposed to link the feasible solutions to the scheduling schemes. The search operations are performed only in the operation sequence space, and a corresponding machine allocation vector is generated for each operation sequence vector based on the greed mechanism. Two crow position update strategies named track and hover are redesigned and the improved crow search algorithm is utilized to search the operation sequence space efficiently in order that the GCSA can adapt the SFTSP and make full use of the information obtained during the search process. Moreover, the effect of parameters is investigated based on a multi-factor analysis of variance (ANOVA) approach. Finally, extensive computations and comparisons on ten test instances derived from the practical production demonstrate that the proposed GCSA outperforms the state-of-the-art methods in the literature to solve the SFTSP.
Semiconductor final-test scheduling under setup operator constraints
2022, Computers and Operations Research
We consider a semiconductor final-test scheduling problem that aims at minimizing the total weighted tardiness. In contrast to previous studies on this problem, we explicitly take account of the need to assign human operators to setup operations. We present decomposition-based heuristic solution approaches and a mixed integer program. In a computational study based on real-world problem instances that mimic settings at our industry partner, we show that our heuristics clearly outperform a standard solver when computational time is limited. Based on this result, we provide decision support for managers by analyzing the capability and effect of rescheduling jobs in the presence of a highly dynamic environment with frequently changing customer requests and common test machine failures.
Semiconductor final testing scheduling using Q-learning based hyper-heuristic
2022, Expert Systems with Applications
Semiconductor final testing scheduling problem (SFTSP) has extensively been studied in advanced manufacturing and intelligent scheduling fields. This paper presents a Q-learning based hyper-heuristic (QHH) algorithm to address the SFTSP with makespan criterion. The structure of QHH employs the Q-learning algorithm as the high-level strategy to autonomously select a heuristic from a pre-designed low-level heuristic set. The selected heuristic in different stages of the optimization process is recognized as the executable action and performed on the solution space for better results. An efficient encoding and decoding pair is presented to generate feasible schedules, and a left-shift scheme is embedded into the decoding process for improving resources utilization. Additionally, the design-of-experiment method is implemented to investigate the effect of parameters setting. Both computational simulation and comparison are finally carried out on a benchmark set and the results demonstrate the effectiveness and efficiency of the proposed QHH.
An effective invasive weed optimization algorithm for scheduling semiconductor final testing problem
2018, Swarm and Evolutionary Computation
Citation Excerpt :
We can also see that our implements of the HEDA and KMEA yield slightly better results than their original versions in the literature, while the nFOA performs slightly worse than its original version. The superiority of the HEDA and KMEA over nFOA has already been demonstrated by Wang et al. [15] and Wang and Wang [14]. So, the superiority of the presented CCIWO algorithm over the other competing algorithms is still sound.
In this paper, we address a semiconductor final testing problem from the semiconductor manufacturing process. We aim to determine both the assignment of machines and the sequence of operations on all the machines so as to minimize makespan. We present a cooperative co-evolutionary invasive weed optimization (CCIWO) algorithm which iterates with two coupled colonies, one of which addresses the machine assignment problem and the other deals with the operation sequence problem. To well balance the search capability of the two colonies, we adopt independent size setting for each colony. We design the reproduction and spatial dispersal methods for both the colonies by taking advantage of the information collected during the search process and problem-specific knowledge. Extensive experiments and comparison show that the proposed CCIWO algorithm performs much better than the state-of-the-art algorithms in the literature for solving the semiconductor final testing scheduling problem with makespan criteria.
A shuffled multi-swarm micro-migrating birds optimizer for a multi-resource-constrained flexible job shop scheduling problem
2016, Information Sciences
Citation Excerpt :
This improvement suggests the effectiveness of the shuffled multi-swarm strategy compared with the MBO algorithm. Interestingly, SM2-MBO1 with θ ∈ [9, 21] yields an ARPI value of less than zero, which indicates that the average makespan over the 30 replications found by SM2-MBO1 is better than the best solution reported in the literature [45]. Recently, Zheng et al. [53], Wang et al. [46], and Wang et al. [45] addressed the MRC-FJSSP with the makespan criterion.
Scheduling problems with resource constraints have been a new research trend in recent years. This paper addresses a multi-resource-constrained flexible job shop scheduling problem that is very common in semiconductor manufacturing, precision engineering, and many other modern industries. To address this important problem, a novel algorithm called the shuffled multi-swarm micro-migrating birds optimization (SM²-MBO) algorithm is presented with a two-vector representation. The SM²-MBO forms a number of micro-swarms, each of which performs its own MBO independently. A random shuffle process applied to the entire population is invoked periodically to propagate the good information that is found in some of the micro-swarms. A diverse controlling strategy based on the aging phenomenon of life is proposed to diversify the population. An adaptive search operator based on a problem-specific crossover and a two-vector crossover helps to balance exploitation and exploration. Numerical experiments and comparisons are conducted against the best performing algorithms reported in the literature for the considered problem. The results demonstrate that the proposed SM²-MBO performs significantly better than the existing algorithms in solving the multi-resource-constrained flexible job shop scheduling problem with the makespan criterion. Furthermore, the proposed SM²-MBO can improve 9 out of 10 best known solutions for the benchmark instances in the literature.
Application of a genetic algorithm in solving the capacity allocation problem with machine dedication in the photolithography area
2016, Journal of Manufacturing Systems
Wafer fabrication is a complicated manufacturing process with high process capability. Hence, maximizing machine capacity to meet customer deadlines is a very important issue in this field. This study proposes an integer programming model and a heuristic algorithm approach to solve the loading balance problem for the photolithography area in the semiconductor manufacturing industry. Considering process capability, machine dedication, and reticle constraints, we aim to minimize the difference in loading between machines. Process capability means that each product must be processed in machines that meet the process specification. Machine dedication means that if the first critical layer of a wafer is assigned to a certain machine, then the following critical layers of such wafer must be processed in this certain machine to ensure wafer quality. This research compares the results of two methods and finds the best parameter settings of the genetic algorithm (GA). The computational performance results of the GA shows that we can find the near-optimal solution within a reasonable amount of time. Finally, this research analyzes machine capability and reticle flexibility to determine the best percentage that can be used as reference for application in the semiconductor industry.

View all citing articles on Scopus

View full text

A knowledge-based multi-agent evolutionary algorithm for semiconductor final testing scheduling problem

Abstract

Introduction

Section snippets

Literature review

Problem description

Knowledge-based multi-agent evolutionary algorithm

Numerical results and comparisons

Conclusions

Acknowledgments

Knowl.-Based Syst.

Knowl.-Based Syst.

CIRP Ann.-Manuf. Technol.

Expert. Syst. Appl.

Int. J. Prod. Econ.

Knowl.-Based Syst.

The cooperative estimation of distribution algorithm: a novel approach for semiconductor final test scheduling problems

J. Intell. Manuf.

Review of classification using genetic programming

Int. J. Eng. Sci. Technol.

Opposition based initialization in particle swarm optimization (O-PSO)

Cooperative Evolutionary Computation and Multi-Agent System

A multiagent genetic algorithm for global numerical optimization

IEEE Trans. Syst. Man. Cybernet. B

A multiagent evolutionary algorithm for constraint satisfaction problems

IEEE Trans. Syst. Man. Cybernet. B

A multiagent evolutionary algorithm for combinatorial optimization problems

IEEE Trans. Syst. Man. Cybernet. B