Implementation Of Fault Prediction Algorithm Based On Data-driven For Analog Circuit And Software Development

According to statistics, more than 80% failures are from the analog circuits in electronic system equipment. Therefore, the analog circuits fault prediction technology is the key to enhance the reliability of electronic equipment. In the field of fault prediction, fault prediction method based on data-driven is widely used. It mainly uses the equipment simulation datas, fault injection datas to make trend prediction through data analysis and processing algorithm. Autoregression prediction model has the advantages of modeling simply and computing quickly. Meanwhile, grey model requires less modeling datas and the prediction accuracy is high. Based on the above reasons, we do a thorough research in software architecture design and key modules design for fault prediction software, major works in this paper are summarized as follows:1. The architecture of the prediction software is designed. The prediction software is divided into four functional modules i.e. fault prediction, database management, interface service, user rights management. Using VC++6.0 as the development platform, we have designed the data flow of the software. In addition, we have used the property sheet and property page as the integrated way of functional modules.2. Fault prediction algorithm based on the data-driven is implemented. Fault prediction algorithm is the core of the software, according to the modeling steps of autoregression prediction model, grey prediction model and particle swarm optimization algorithm(PSO), we have designed the software implementation process and function interface of these algorithms using standard C++ library, during the process of algorithm implementation, function libraries which are relevant with MFC are not called, this method makes the call of prediction algorithms implemented in this thesis are not limited to the specific operating system, to some extent the portability of the prediction algorithm between different operating system is increased and the reusability of the prediction algorithm in different environments is strengthened.3. Fault prediction software platform development is implemented. Firstly, according to the requirement of third normal form, server-side database is designed. Software uses ADO connection way to connect the database to realize the data access of fault information, prediction results and so on. Secondly, the software has designed and packaged the methods and classes of each functional module with the idea of object-oriented. The software is based on MFC dialog box and makes full use of the MFC controls and Visio ActiveX graphical development controls to achieve the goal of friendly human-computer interaction.4. The prediction software has been tested and verified. IGBT(Insulated Gate Bipolar Transistor) measured datas and Sallen-key analog circuit simulation datas are tested using prediction software. By comparing practical circuit Remain Useful Lifetime(RUL) with predicted RUL, the results show that the software has the very high accuracy for circuit RUL and can provide real-time and reliable theory basis for the health management of the circuit system.