Research On Low Power Real-time Scheduling Algorithm For Cyber-Physical Machine Tool System

With the development of computer system,especially the network technology,the realization form of computer integrated manufacturing system is moving from centralized system to system application platform,open system structure,manufacturing resource as a service,intelligent management decision-making and green manufacturing.High-end CNC machine tools,as master of high-end equipment manufacturing,are undoubtedly playing a unique role in Cyber-Physical Production Systems(CPPS).The necessity of advancing the machine tool to meet the industrial 4.0 concept has been recognized.Cyber Physical Machine Tools(CPMT)and Cyber Physical Machine Tool System(CPMTS)have become an important direction for the development of Industry 4.0,which are an essential element of CPPS.CPMT is the integration of machine tool,machining processes,computation and networking,where embedded computers and networks can monitor and control the machining processes,with feedback loops in which machining processes can affect computations and vice versa.CPMTS is based on CPMT and they are mutually reinforcing.However,CPMTS emphasizes networking,digitization,and systematization.Thesis focuses on the realtime low power scheduling theory to study the framework and the scheduling algorithm in CPMTS,and proposes the framework of the local cloud-end integrated cyberphysical machine tool system.Based on this framework,to study real-time scheduling methods and applications of low-power and reliability for multi-type tasks suitable in CPMTS.In addition,considering the requirements of the fusion of physical components and information models of CPMTS,thesis studies the data fusion cleanseing algorithm of physical components and information models.The feasibility and effectiveness of the method is verified by experiments.The main contents of the thesis include the following aspects:Firstly,considering the need of scheduling sporadic tasks in real-time systems.A new algorithm is proposed by the slack time to dynamically scale the sporadic task speed.According to the random arrival time characteristics of sporadic task,the algorithm defers the processor speed is scaled until the task arrives at the moment.After the task execution is completed,the remaining slack time can be reclaimed in advance to scale the speed of subsequent task.Processor's dynamic power consumption and static power consumption are considered in the processor general model.In addition,to balance dynamic power consumption and static power consumption of sporadic task,dynamic voltage scaling technology and critical speed are adopted in the low power consumption algorithm,and the dynamic power management technology is combined to further reduce the system power consumption.However,there is a balance factor between the critical speed and the traditional DVS scheduling strategy.To reduce power consumption of real-time system,a low energy consumption scheduling algorithm(LPDSAFST)based on the balance factor is proposed.Experiments show that the new algorithm can save energy better than the existing DVSST and DSTLPSA algorithms.Secondly,to study the demand of power consumption and reliability in CPMTS.By optimizing the slack time allocation strategy,it is possible to minimize system energy consumption while ensuring the reliability of CPMTS.A Low Power and Reliability Based on Sliding Window(LPRSW)algorithm is proposed.LPRSW algorithm is divided into LPRSW-H algorithm and LPRSW-A algorithm.The former makes the task fault-tolerant with the highest speed and the latter using scaled speed.Based on this,a Cooperative Optimal Scheduling Algorithm for Low Power and Reliability Based on Sliding Window(COSALPRSW)is proposed.The algorithm distributes the global slack time to backup tasks instead of assigning a backup task to each task,which can get more slack time to reduce system power consumption,while the slack time factor is able to allocate slack time to subsequent tasks for reliability and low power consumption.Finally,the feasibility and effectiveness of these two scheduling algorithms are verified by the schedulability analysis of the algorithm and simulation experiments.Thirdly,to make full use of the parallel capability of multi-core processors,the low power consumption of the system is realized while ensuring the reliability of the numerical control system.Based on coarse-grained task-level software pipeline technology,a non-recursive acyclic graph algorithm is proposed to convert the periodic dependency task to a set of independent tasks based on retiming.The system energy consumption is reduced through concurrent task execution and dynamic voltage scaling.Secondly,the algorithm is divided into two phases in ensuring system reliability: the fault-free phase and the fault-tolerant phase.In the fault-free phase,multiple copies of a task are executed simultaneously,and voting is used to determine if the task is performed correctly.When inconsistent,the task enters fault-tolerant mode.In the fault tolerant mode,the remaining copy exclusive processor completes fault tolerance with its highest speed.The experimental results show that the proposed DPDAGA algorithm can reduce the system energy consumption while ensuring the reliability of system.Compared with the existing algorithms,DPDAGA algorithm not only has higher reliability but also lower energy consumption.Fourthly,a high-performance open CPMTS framework is proposed.CNC systems are moving toward an open architecture with flexibility,adaptability,versatility and expansibility.Existing CNC systems tend to have higher power consumption.Develop a new open and high-performance CNC(OHP-CNC)cyber physical system platform using international standards and open hardware and software.With OHP-CNC,open CNC systems can be migrated to other platforms.Based on the bridged interconnects between MCUs and multiprocessors,processors that support dynamic voltage regulation can be used in CNC systems.The real-time system API encapsulation allows the CNC software to be run on different operating systems,enabling the design of lowpower,high-reliability CNC systems to be scalable vice versa.The proposed low-power scheduling algorithm is used to schedule mixed tasks including periodic and aperiodic tasks.The algorithm is divided into two phases.Firstly,slack time and utilization are calculated on each processor and tasks are assigned to the processor based on the load.Secondly,since there is a tradeoff between the response time of aperiodic tasks and energy-saving,the scheduling server is used to schedule aperiodic tasks to meet the response time constraints of aperiodic tasks.Finally,the performance of the proposed CPMTS is verified by experiments,which show that the energy-aware real-time algorithm yields high-performance and effective machining processes.Fifthly,a low-power data cleaning algorithm model based on CPMTS is proposed.CPMTS usually uses the wireless sensor network to carry on monitoring and control.However,the data collected by wireless sensor networks is often not accurate and unreliable due to battery power limitations and noise.To achieve reliable and accurate data acquisition,while ensuring low energy consumption and long lifetime of WSNs,an energy-saving data cleanseing algorithm is proposed.The data cleanseing algorithm first cleanses the data in the local sensor.With dynamic voltage scaling and dynamic power management,sensor power consumption is reduced at task scheduling level without affecting system performance.In addition,a low-power protocol for node aggregation is proposed at the network level.Based on the above techniques,a health monitoring system for a Cyber-Physical Machine Tool system(HMS-CPMST)is designed.Experiment results show that the proposed energy-saving data cleansing algorithm yields high-performance and effective monitoring.