TLSCA-SVM Fault Diagnosis Optimization Method Based on Transfer Learning

Abstract

In fault-diagnosis classification, a pressing issue is the lack of target-fault samples. Obtaining fault data requires a great amount of time, energy and financial resources. These factors affect the accuracy of diagnosis. To address this problem, a novel fault-diagnosis-classification optimization method, namely TLSCA-SVM, which combines the sine cosine algorithm and support vector machine (SCA-SVM) with transfer learning, is proposed here. Considering the availability of fault data, this thesis uses the data generated by analog circuits from different faults for analysis. Firstly, the data signal is collected from different faults of the analog circuit, and then the characteristic data are extracted from the data signals by the wavelet packets. Secondly, to employ the principal component analysis (PCA) reduces the feature-value dimension. Lastly, as an auxiliary condition, the error-penalty item is added to the objective function of the SCA-SVM classifier to construct an innovative fault-diagnosis model namely TLSCA-SVM. Among them, the Sallen–Key bandpass filter circuit and the CSTV filter circuit are used to provide the data for horizontal- and vertical-contrast classification results. Comparing the SCA with the five optimization algorithms, it is concluded that the performance of SCA optimization parameters has certain advantages in the classification accuracy and speed. Additionally, to prove the superiority of the SCA-SVM classification algorithm, the five classification algorithms are compared with the SCA-SVM algorithm. Simulation results showed that the SCA-SVM classification has higher precision and a faster response time compared to the others. After adding the error penalty term to SCA-SVM, TLSCA-SVM requires fewer fault samples to process fault diagnosis. Ultimately, the method which is proposed could not only perform fault diagnosis effectively and quickly, but also could run effectively to achieve the effect of transfer learning in the case of less failure data.

Time-domain waveforms and phase-domain trajectories of fault recording and mathematical models of phase-domain trajectories

“…Three-phase SCCs pre and post a three-phase short circuit in an infinite power supply system are shown in Equation (9), where there are only the decaying dc components (DDCCs) with the decay time constant (DTC) of 0.05s in addition to the power frequency fundamental components (PFFCs). It is supposed that the system frequency remains constantly 50 Hz pre and post the short circuit, and the angular frequencies of the SSCs are all 100π in the normal steady state (NSS) and the short-circuit steady state (SCSS).…”

“…When fault diagnosis algorithms combining fault analysis and artificial intelligence (e.g., principal component analysis, neural network [8], and support vector machine [9] etc.) are applied in power systems, the shortcomings of great calculation amounts, many setting parameters, and complexity in calculation processes limit their depth and breadth application.…”

Study on Algorithm of fault recording analysis combining its time-domain waveforms and phase-domain trajectories

“…The sizes of a PFW and a HPFW are 360° and 180° respectively, and both the values are fixed. However, as can be seen from Equation (22), the size of a SW varies with the changes of the system frequency and the sample frequency, i.e., any change in the system frequency or sample frequency impacts on its size. When the sample frequency remains unchanged, the size of the SW is needed to be constantly adjusted with the change of the system frequency.…”

“…Only when this condition is met can the TDW be reformulated from the parameters of each MP in the PDT. As can be seen from Equation (22), which is the definition of a SW, when sample frequency fs is fixed, SW Δφ varies with system frequency f0. To ensure that the angular frequency of the PDT is equal to that of the corresponded TDW and keep their periods to be matched, and reduce the errors between the TDW reformulated from its PDT and the original one, it is necessary to adjust SW Δφ with the variation of the system frequency and to calibrate its PDT.…”

“…In addition to the above-mentioned analysis methods, fault diagnosis is accurately achieved when fault analysis is combined with artificial intelligence (AI) algorithms, such as Kalman filter [16], expert system reasoning [17], Bayes estimation [18], petri nets [19], cluster analysis [20], and classical reasoning, etc. AI algorithms: neural network [21] and support vector machine [22] etc., fuzzy set theory [23], and rough set theory [24] etc. are also widely applied to the area of fault diagnosis using multi-source information fusion, and they are researched in a good prospect.…”