Data Failure Detection Based On Data

With the rapid development of semiconductor technology and manufactur-ing technology, also mechanics of communication and the network technology, the automation level of modern industrial production and the complexity of con-trol system are increasing day by day. The accurate measured values of the in-strument in processing of production have a very vital role to the safety of indus-trial production process. It also affects the reliability of the control system and it can ensure the quality of industry products. Industrial instrument as one of the key components of the control system, due to its material technology, manufac-turing technology and working environment and other factors, in the whole con-trol system, industrial instrumentation failure occurs more easily. Fast and accu-rate fault detection of instrument failure and taking the right fault strategy is the key and has the very vital significance to ensuring the stable operation of control system and eliminating of the production safety.This paper according to the specific situation in production of industrial process instruments, adopt the fault detection method based on data driving. Re-alize the goal of instruments fault detection for industrial process. And, the re-search methods are used in Tennessee Eastman simulation process with simulat-ing different instrument fault signals and carries out comparative analysis of test results. The main work of this paper is as follows:1) For the industrial production data is not consistent with the distribution of the Gauss distribution, and combined with the production process of multi in-strument condition, the fault detection method based on independent component analysis is adopted. Extract and separate the source information of the historical normal data, and then built a fault detection model. At last, determine the control limits by kernel density estimation. The multi instrument fault detection outcome can be realized. Compared with the fault detection result based on principal component analysis method, the fault detection method based on independent component analysis is more suitable for instrument fault detection in industrial production processes.2) The Gauss source information and non Gauss source information exists in the industrial production data. Extract and separate different information in the industrial production process data by using the theory of principal component analysis and independent component analysis method. The corresponding analy-sis method respectively, a fault detection model is not the same. The simulation results show that, compared with the fault detection method based on independ-ent component analysis single, combined with independent component analysis and principal component analysis method has better testing effect.3) Improve the model of fault detection in each independent component subspace based on the contribution of independent component subspace theory method. Apply it to solve the problem that it is difficult to detect small fault of the instrument. And provide integrated fault detection strategy according to the actual needs of the different condition. The simulation result shows that inde-pendent component subspace method and improved fault detection model im-proves detection effect instrument of small faults, different integration strategies can be more flexible and have a better applicability.In this paper, different fault detection theories are introduced to detect faults of multi instruments in industry process. Anew idea of instrument fault detection in industrial process can be provide by analysis the test results,...