Constrained Independent Component Analysis And Its Application To Rolling Element Bearing Fault Diagnosis

With the scientific and technological progress and the development of industry requirement, modern mechanical equipments are being developed towards the direction of complexity, high speed, high efficiency, hugeness, automation, and at the same time, their running condition is more and more rigorous. Once there is something wrong with them, the whole production efficiency will fall or machine sets halt, even catastrophic accidents will occur. Therefore, a important part of machine fault diagnostics is how to enrich and improve the machine fault diagnosis technology. This dissertation addresses the fault diagnosis of rolling element bearing, with the purpose of enriching machine fault diagnostics and requirements of engineering application of fault diagnosis of the key equipment in mechanical engineering, by means of constrained independent component analysis methods. It is necessary and important to diagnose machine fault accurately and effectively, so as to provide maintenance strategy and deduce economic losses. It is not only of great theoretical significance, but also of great engineering value.It is studied that constrained independent component analysis(CICA) based on the pulse method, constrained independent component analysis based on model and frequency domain constrained independent component analysis in turn in the dissertation. It mainly includes the following research contents in the dissertation.The background and significance of the dissertation are elucidated from the viewpoint of theoretical analysis and engineering application. The related literatures are reviewed which consists of the relationship between blind source separation (BSS), independent component analysis (ICA) and constrained independent component analysis (CICA). Basic principle and procedure of BSS, and the advantages of CICA are introduced. Difficulties of the BSS implicated to machine fault diagnosis are analyzed. The research contents and technical route are finally described.Mathematics background, basic concepts and principles associated with BSS are introduced. Then model, identifiability, measurement of non-gaussianity, and basic principles of CICA are studied according to historical development of ICA and relationship between ICA and CICA. The advantages, constrained type, principle and evaluation index of CICA are finally studied.Constrained independent component analysis based on pulse algorithm is established and applied to fault diagnosis of the rolling element bearing. CICA algorithm is derived and proved. The condition and establishing method of CICA are analyzed and induced. The relation between the parameters and the performance of CICA, the conditions of convolutive mixing model of CICA replacing by instantaneous mixing model are studied. CICA based on pulse algorithm with reference of pulse sequence whose frequency equals to characteristic frequency of rolling element bearing will converge towards the direction of reference signal. The fault signal will be extracted correctly by CICA based on pulse algorithm if the fault signal in sensors is consistent with reference signal, otherwise the fault signal will not be extracted correctly.A constrained independent component analysis based on fault model (CICA based on model algorithm) is presented and it can be applied to fault diagnosis of rolling element bearing. CICA based on model algorithm whose reference signal comes from fault model of rolling element bearing converges to reference signal, and the extraction of fault signal have basically nothing to do with model error. Simulations and experiments showed that the fault signal will be extracted correctly by CICA based on model algorithm if the fault signals in sensors are consistent with reference signal, otherwise the fault signal will not be extracted correctly. The calculation speed of the algorithm is improved compared with CICA based pulse algorithm. CICA based on model algorithm can diagnose rapidly and accurately the fault of rolling element bearing and it provides new method and tool for the intelligent rapid diagnosis of rolling element bearing.A frequency domain constrained independent component analysis (FCICA) algorithm is presented and can be applied to frequency bandwidth or the interesting frequency bandwidth known. The signals in the frequency bandwidth are extracted by CICA algorithm based on fast fixed point ICA algorithm embed a filter as a constraint. The algorithm can not only extract the signals rightly whose frequencies are overlapped but also have noise cancellation ability. The effectiveness and validity of the algorithm are finally demonstrated by simulations and experiments.Research work and the main innovation of the paper are summed up, and further research contents are proposed.