Monitoring Signal Feature Fusion Based On Bi-spectrum And Wavelet Packet Method For Drilling Quality Classification Evaluation

Drilling process is in closed or half-closed environment, so it is difficult to acquirequality characteristics data directly and realize quality control in processing. At present,for batch drilling quality inspection, it is realized by traditional manual samplinginspection after processing in most cases, but some drillings are undetected by this way,thus, it will lead to hidden quality trouble, and couldn’t control quality problem inprocessing at the same time. Therefore, it has important practical significance to developand research inspection method and the theory of quality consistency for batch drillingin the process. Batch drilling process monitoring signal is acquired by sensors in thisdissertation, and then we study the consistency batch drilling quality by analyzing theconsistency of sensor detection signal in batch processing. The main research work is asfollows:1. Aim at the features that drilling processes is in closed or half-closed environmentand the characteristics of different sensors in sensitivity、installation mode and signalanalysis method are various, an acoustic emission sensor and a triaxial accelerationsensor were used to acquire drilling process monitoring signal.2. According to the irregular and non-stationary characteristics of drilling processsensor signal, basis for multi-scale decomposition property of wavelet packet analysismethod and the bis-spectrum analysis method’s ability of suppressing gaussian noise,combining two kinds of analysis methods to achieve the purpose of double denoising,extracting wavelet packet bi-spectrum slice features of sensor signal, establishing themapping relationship between wavelet packet bi-spectrum slice features and the drillingquality.3. First, the extracted features are normalized processed, and given different weightsby ReliefF algorithm according to the similarity among them, then the weighted featuresare clustered by Fuzzy clustering algorithm, finally the accuracy of clustering results areevaluated by the F-measure method.Analysis results show that the method proposed in this dissertation can effectivelyrealize the classification detection of drilling work step quality, and has a higher detection rate than using bi-spectrum slice method and the traditional fuzzy clusteringmethod.