Study On Hierarchical Modeling Method And Key Technologies Of Parallel Multi-task Automatic Test System

Along with the development that the serial test has been substituted by the parallel multi-task test in Automatic Test System(ATS), The parallel multi-task ATS's(PATS's) structure design, task scheduling and construction have been studied in academic field, and many first fruits have been achieved. The PATS has the flexibility and extension, and its structure is complex and huge. The process of its construction and design is more complicated than the serial ATS, and it can not be completed depending on the traditional method only based on the people's experiences because it will bring the deadlock, resouces confliction, hunger and so on. So the formal modeling and verification of the PATS are very important to the construction and design of test system. This dissertation bases on the test technologies, Petri net, Unified Modeling Language(UML), genetic algorithm and neural network. The PATS's modeling and verification, deadlock avoidance, task scheduling optimization and fault diagnosis are studied deeply in the dissertation. Author's main work is depicted as follows:1. A new hierarchical system modeling method is proposed for engineer instruction in getting a good project design of the PATS at short time.At first, studying on the problems produced by the project design method only based on the people's experiences, combine the comparative analysis between UML and Petri net, use the UML to depict system function, build initial domain model. Then, the mapping method from the UML model to Petri net model is proposed. Get system Petri net model with strict formal semantic for performance analysis and evaluation, the advantage of UML and Petri net is combined consequently.2. A hierarchical modeling instance of the radar ATS is designed, and then through it the system's performance evaluation is studied.Firstly, the execution process of the radar ATS hierarchical modeling method is introduced, the consistency of behavior sematics between UML model and Petri net model got by UML models maping to is analysed, the soft "UML2Petri" is developed and the automatic map pattern is realized. Secondly, The system performance is evaluated qualitatively base on Petri net theory, is compared with the system designed only by people's experiencs to prove the advantage of Petri net. Thirdly, after the time factor is introduced, the system's Petri net model is extended to Generalized Colored Stochastic Petri net model, then the system performance is evaluated quantitatively about the system test time and the resource utilization, and is compared with the system designed by TestStand software, the results proved the validity of this hierarchical modeling method.3. Studying on the deadlock in the PATS, a deadlock avoidance policy based on the Petri net and genetic algorithm (GA).Firstly, the deadlock problem in the PATS is discussed, and some typical deadlock avoidance methods are compared. Secondly, the deadlock problem in the PATS is elaborated based on Petri net, a new deadlock avoidance policy is proposed by combined with the GA, this policy is validated by a parallel test experiment about dual channels amplifier. Lastly, practicality of this policy is emobied with comparing between GA and heuristic algorithm using a parallel test simulation system about radar reciever.4. Studying on the PATS's task scheduling, the optimal method based on the Time Petri net and the Simulated Annealing and Genetic Algorithm (SAGA) is proposed.Firstly, the PATS's task scheduling problem is discussed, introduce the time factor into system Petri model to get system Timed Petri net model. Secondly, in order to acquire the optimal task schedule sequence, the SAGA is combined to design the optimal algorithm of task scheduling. Thirdly, via a parallel test experiment, advantage of this optimal algorithm is embodied by a parallel test experiment. Lastly, the search efficiency is compared between SAGA and GA methods in a radar receiver PATS.5. Studying on the fault diagnosis problem in the ATS, a fault diagnosis mothed based on the Adaptive Fuzzy Petri net (AFPN) is proposed.Firstly, the fault diagnosis problem in the ATS is discussed, according to the defect of tradition expert system based on production rule and fault tree, the Fuzzy Petri net is used to present knowledge and reason process. Secondly, the AFPN is created by combined neural to adjust the weight. Lastly, this method is used to diagnose the fault in the phased array radar system.