Reliability Design And Fault Analysis Techniques Of CNC System Software

CNC machine tool is the basic equipment for processing and manufacturing industry.As the core functional component of the CNC machine tool, CNC system plays a key role tothe integrity of the CNC machine tool. Compared with the foreign products, the domesticCNC system has gradually had approximate function and accuracy level, but in reliability,there is still a large gap. China import a lot of high-precision CNC systems from foreigncountries every year, which are not only expensive, but are also limited by various factors.Therefore, it has become an inevitable road to take to promote the development ofequipment manufacturing industry of China to improve the reliability of domestic CNCsystem and make great efforts to develop the domestic CNC system.With the gradual maturity of hardware technology, software fault has become the mainreason to cause the fault of CNC system. With the continuous development of softwaretechnology, the scale of software has been larger and larger, the level of software has beenmore complex, the concurrency of system is continuously strengthened and the CNC systemsoftware fault has also been increasing. In order to avoid the corresponding loss, it is veryimportant to research the reliability of CNC system software.In combination with the research work of project “demonstration role of domesticrolling functional components in medium and high-grade CNC machines” and“demonstration role of domestic rolling functional components in medium and high-gradeCNC machines”。 The research on CNC system software reliability allocation, componentassemble model based on reliability, software fault location and reliability predictionmodeling is conducted. The main researches of the paper are as follows:(1) The system architecture-based software reliability allocation method of CNC systemis proposed.The CNC system structure and component technology are analyzed, thecomponent-based CNC system software structure is established and the systemarchitecture-based CNC system software reliability indicator allocation method is proposed.The CNC system reliability indicator is distributed to the functional component, and thereliability allocation is proposed with the system practicability as the objective function, functional component reliability and expense function as the restriction function. Therelative weight value of elements in each layer of the CNC system software architecture iscalculated with AHP method and the allocation indicator of each component is calculatedthrough cultural algorithm. Reliability allocation experiment of CNC system software isconducted to obtain the reliability allocation indicator of each functional module. The resultindicates this method guarantees that the CNC system reliability indicator meets therequirements, and meanwhile effectively saves the development cost, improves theeffectiveness of allocation method and provides basis for the development of subsystem.(2) The reliability-based CNC system component assembly model is established.The CNC system reliability evaluation indicator system is established, reliability-basedCNC system component assembly technology is proposed and the reliability-based CNCsystem component assembly model is established. The CNC system function iscomponentization and CNC system is set up with CNC system software reliability measuringparameters as the main evaluation indicator. Aimed at the characteristics of localconvergence of particle swarm, the improvement of particle swarm algorithm with inertiaweight scaling method is proposed, and the component assembly modeling solution methodbased on the improvement of particle swarm algorithm is proposed. Component assembly isconducted to the interpolation function with high requirements for timeliness of CNC system.The practice result indicates that this method does not only make the CNC system softwareflexible, easy to be modified and expanded, but can also help to assemble high-reliabilityCNC system conforming to the manufacturing demand more easily and efficiently.(3) The architecture expansion-based fault location method of CNC system software isproposed.The failure mechanism of CNC system software is analyzed, the expansion method ofCNC system software architecture is proposed, the expansion component of CNC systemsoftware architecture is established. The software data morphology information and runningpath are monitored and recorded, the failure pathway is obtained, similar path set algorithmis adopted to generate the similar pathway set of the fault path, least squares SVM-basedsuspicion model is established to determine the fault statement, eliminate fault and positionthe fault of software in the level of CNC system structure. Fault positioning experiment isconducted in the multi-axis movement control card PCI-7344. The experiment result showsthat the system structure expansion based on the overall structure characteristics of CNCsystem software and the fault positioning method through similar path and least squares SVM avoid the repeated tests and debugging of programmers. Without being limited byartificial factors and levels, it is a reliable CNC system software fault location method.(4) The reliability prediction model of CNC system software is established.The reasons for the failure of main functional component of CNC system are analyzed,the failure model and failure reason of CNC system functional component with repeatedfault frequency is analyzed and discussed, and the main reasons for the model of CNCsystem software fault are analyzed. The CNC system software of optimizing BP neuralnetwork algorithm based on particle swarm is proposed, the improved algorithm overcomesthe limitation of slow convergence speed of the traditional BP neural network and sinkinginto local weight value and effectively improves the convergence speed and prevision of BPneural network. According to the requirements for the reliability prediction of CNC systemsoftware, neural network prediction model is established. Meanwhile, the data collected inthe on-site test are counted, the training samples and test samples are established, the neuralnetwork is predicted for training, and after the network reaches the error requirementsrequired, the reliability of the next period of such model of CNC system is predicted.