Proactive-Aware-Event-Driven Job-Shop-Scheduling Problems In The Internet Of Manufacturing Things

With the popularity of sensing devices and the development of Internet of Things(Io T),the combination of new-generation information/artificial intelligence technologies such as cloud computing,cyber physical systems,big data,deep learning and fog computing,with advanced manufacturing technology,promotes the continuous development of the Internet of Manufacturing Things(Io MT)model.This study is to meet the urgent demand of cloud-fog fusion for such development of Io MT.On the one hand,the fog layer is used to collect and timely process the real-time data from the sensing equipment layer,including the monitoring data of the job processing status based on RFID,and the sensing data of vibration and force in the process of tool processing,as well as resulting in the aware-events for the job processing status,the tool remaining useful life(RUL)and the tool wear status after such data timely processed.On the other hand,the cloud layer is used to calculate large-scale data and train the two models for the tool wear level prediction and the tool wear status classification in the fog layer.The aware-events are used to drive job shop scheduling and monitor the production process in real time,and form a proactive-aware-event-driven dynamic job shop scheduling method.The details are as follows:(1)Cloud-fog integrated Io MT architecture and experimental platform construction.The Io T architecture is analyzed,the related concepts and characteristics of cloud computing and fog computing are expounded,the Io MT architecture with complementary advantages of cloud and fog is constructed,and then the proactive-aware-event-driven job shop scheduling framework model and experimental platform with the fusion of cloud and fog under Io MT are built.(2)Job processing status monitoring based on RFID complex event processing.The RFID-based sensing environment of processing workshop is built,and an RFID-based proactive aware complex event processing system is proposed for job processing.The experiments have verified that the system can effectively deal with the RFID complex events,monitor the whole process of job processing,and feedback the processing status of job in real time.(3)Tool RUL prediction based on multi-sensor data fusion.A prediction system of the tool RUL based on multi-sensor data fusion is designed.In this system,the vibration and force sensor signals in the process of tool machining are collected,and the signal denoising,feature extraction and feature selection are researched.On these bases,an improved artificial bee colony-back propagation neural network optimized model is established to fuse the selected features and predict tool wear level,and polynomial curve fitting is adopted for predicting the tool RUL.The reliability and stability of the prediction model is verified using historical and real-time data,and the proposed model has better prediction performance by comparison and analysis with several traditional neural networks.(4)The tool wear status monitoring based on multi-channel Convolution Neural Network(CNN).Taking the improved residual network of Res Net50 as the body structure,the advantages of depthwise separable convolution,improved channel attention mechanism and channel shuffle method are combined to build a multi-channel CNN for tool wear condition monitoring model.The spectrograms generated by multi-sensor signals in the process of tool machining are used as the training data set of the model,which is further used to monitor the tool wear status in real time and verified with comparison of different types of CNN and traditional neural network models.(5)Job shop scheduling based on hybrid whale optimization algorithm.Lévy flight(LF)operator,differential evolution(DE)strategy and the whale optimization algorithm(WOA)are combined to propose a hybrid whale optimization algorithm,called WOA-LFDE.By solving the same job shop scheduling problems,it is verified that the performance improvement of the hybrid algorithm is improved by LF operator and DE strategy.Compared with 9 representative intelligent algorithms to solve the same job shop scheduling problems,the proposed algorithm performs better.(6)Proactive-aware-event-driven job-shop scheduling problem in Io MT.An event driven job shop scheduling approach is formed,in which AGV is used as workpiece distribution,WOA-LFDE as the scheduling algorithm,and job processing status,tool RUL and tool wear status as disturbance events,and verified by experiments.