Table 5. Scheduling problem in AI scheduling approaches.
Scheduling Number of Reference number Period
problem publications of publications
Parts dispatching16 [65], [92], [95– 1987–2000
98], [100], [101],
[103], [106–110],
[112], [113]
Machine 3 [97], [99], [105] 1989–1997
selection
AGV scheduling 3 [104], [105], 1996–1999
[111]
criteria problem and proposed a way to use fuzzy logic in order to build aggregated rules and obtain a compromise between the satisfaction of several criteria.
4.2 Statistics on Scheduling Problems and Methodologies
4.2.1 Scheduling Problems
Table 5 showed the scheduling problems found in this category. Parts dispatching was the most popular scheduling problem for research. This result is consistent with the one found in general FMS scheduling studies and multi-criteria approaches.
4.2.2 Methodologies
Table 6 summarised the AI scheduling methodologies that were used in the reported research in this section. A fuzzy approach and an expert system were the most frequently employed methods. However, expert systems have not been investigated since 1994. Other fields, such as generic algorithms and neural networks, are potential research areas.
4.3 Review of Related Publications
Karwowski and Evans [94] illustrated the potential applications of fuzzy methodologies to various areas of production manage- ment, including new product development, facilities planning, human product management, production scheduling, and inven- tory control.
Table 6. Methodologies in AI scheduling approaches.
Methodologies Number of Reference number Period publications of publications
Fuzzy 7 [94], [97], [98], 1986–1999
[101], [104],
[105], [112]
Expert system 5 [65], [95], [96], 1987–1994
[99], [102]
Generic 3 [106], [107], 1997–2000
algorithm [113]
Neural network 4 [103], [109–111] 1995–1999
Others 3 [92], [100], [108] 1992–1998
Schnur [95] discussed the use of “what if” analysis as a decision support tool for manufacturing systems. The appli- cation of simulation in the decision-making process by man- agers, using artificially intelligent knowledge-based expert sys- tems, was discussed as well. However, the application of artificial intelligence in the dispatching of parts was not demon- strated.
Wu and Wysk [65,96] described a multi-pass real-time scheduling algorithm, in which discrete simulation in combi- nation with an expert system and straightforward part dis- patching rules in a dynamic fashion, was employed. The algor- ithm used a constant short time window for each scheduling interval. The logic of the algorithm was as follows. A dis- patching rule that performed the best, according to the selected performance criteria, was applied in successive short-term scheduling intervals. A learning module in the expert system learnt from previous decisions and then generated the candidate set based on the current shop floor status. A simulation model was constructed according to the collected data. A series of simulation runs were carried out starting from the current state using each of the candidate dispatching rules for the next short planning horizon (At) introduced by the user. The rule that had the best-simulated performance in the time period was used to generate a series of commands to the real-time control system of the FMS. The evaluation/application process was then carried on repeatedly, based on a relatively short time frame. As a result of this process, a continual alternating of different dispatching rules would be carried out automatically (for example, in time period one, SPT was selected; in time period two, WINQ was selected; in time period three, FIFO rule was selected, etc.). Consequently, in the long run, this process resulted in a combination of different dispatching rules based on their performance in each short time period. The use of different dispatching rules at different times was designed to overcome the weakness of any single rule. It was reported that an improvement of 2.3% to 29.30% was achieved under three different simulation windows (At) and measures of performance. However, Wu and Wysk did not consider any alternative routeings or operations, nor AGVs dispatching rules. Hintz and Zimmermann [97] developed a simulation model for the purpose of a multi-objective study. They proposed a fuzzy linear programming model to provide a master schedule. For the parts release scheduling and machine scheduling, a multi-criteria decision-making approach using a knowledge- based machine was developed. The simulation model consisted of several workstations, tooling, transportation and storage facilities, pallets, and fixture units. The decision of the knowledge-based machine scheduling was compared with some priority rules. The comparison showed that the knowledge- 柔性制造系统调度的仿真英文文献和中文翻译(19):http://www.youerw.com/fanyi/lunwen_79959.html