Research On Adaptive Scheduling For Hybrid Flow Shop Problem In The Internet Of Manufacturing Things

Hybrid Flow Shop (HFS) manufacturing is a kind of customer-demand-oriented production mode which is commonly seen in mass customization production nowadays. Its production orders features multi-species, periodicity and quantity variety. But its resource demand model during the production is relatively stable. Inevitably, there are many uncertain dynamic events in the procedure of HFS manufacturing, such as equipment breakdowns, emergency orders, and quality accidents. And because of that, the pre-scheduled plan cannot execute as its wish. So dynamic scheduling mechanism is desired to eliminate the influence of dynamic events and maintain the stability of the manufacturing process.However, most of the researches on dynamic scheduling are based on an ideal production model with its random events following some kinds of distribution, without considering the problem of workshop info-feedback failure in acutual shop floor. So it is hard to cope with practical workshop situation. Especially with the rapid development of the Internet of Things (IOT), it opens the way to real-time manufacturing environment, which makes the realtime adaptive scheduling possible.This thesis is supported by the NSFC ’Research on multi-stages bi-level realtime dynamic scheduling for OKP based on RFID (61074146)’. The research is about adaptive scheduling method for HFS manufacturing in real-time feedback situation based on the Internet of Manufacturing Things (IOMT).The specific research contents of the thesis are as follows:1) To solve the information gap problem in shop floor, a real-time manufacturing system framework is proposed based on the technology of IOMT which provide the technical basis for the follow-up study. In this framework, RFID middleware with unified interface specification is programed to realize the plug-and-play access of RFID object under Multi-Agent encapsulation mode. Then a RFID-Bus is designed to process the real-time manufacturing information in a unified feedback mechanism. Based on this RFID-Bus, a real-time manufacturing system environment and its deployment method in the shop floor are constructed.2) To solve the problem of large model scale in actual enterprises, a MPN (Manufacturing Petri Net) modeling mechanism based on ROPN (Resource Oriented Petri Net) is proposed to reduce the model scale. A MPN model is defined to do this by modeling the equal-parallel-machine into one resource node, in which elaborated the knowledge function and the path search method which is mapped in the model by the intelligent token.3) After the analysis of the nonequal-parallel-machine problem in MPN, an offline scheduling algorithm based on MMAS (Max-Min Ant System) is designed. In this algorithm, a transition fire sequence is optimized instead of the token flow sequence. And Taguchi method is used to find the best parameters configurations. By this algorithm, an optimized baseline plan for the MPN can be produced.4) For execution of the baseline plan in MPN, after analyzing the characteristics of dynamic events in HFS in the real-time situation, the thesis maps the dynamic events into capacity-disable events in the MPN. According to the features of real-time manufacturing environment, using the modification strategy, an online adaptive scheduling framework based on interacting of the plan and execution is proposed. In this framework, a decision tree is structured to monitor the executing deviation of each job in the real-time MPN and makes adaptive scheduling decisions. Then, using the decision tree structure, an online scheduling algorithm based on level-feedback is designed.5) According to the above researches, an adaptive scheduling system is realized to simulate the operation of MPN system and deduct the plan execution to verify the effectiveness of the algorithm.A large number of tests and results show that the adaptive scheduling method in IOMT environment proposed by this thesis is able to eliminate the influence by some dynamic events and improve the efficiency of production.The research of this thesis still has many deficiencies need to improve in further study.