Research On The Auditory Perception Used For Fault Diagnosis Of Electric Impact Drill

Since intelligent manufacturing was listed as the main direction of "Made in China2025" in 2015,the intelligent detection of mechanical equipment has become one of the research focuses.The detection of parts defects of rotating machinery has also attracted more attention.The defect detection of mechanical equipment parts is mainly to monitor and diagnose the running state of various parts in the equipment and study the fault diagnosis.In recent years,the conventional fault diagnosis technology is mainly based on current signal,vibration signal,and acoustic emission signal for detecting and analysis.However,considering the inconvenience of vibration sensor and acoustic emission sensor installation in specific environments,the traditional mechanical fault diagnosis technology is greatly restricted.The sound signal of the parts and components in the mechanical equipment contains the abundant automatic state information.Therefore,the acoustic signal can also detect and diagnose the state of the mechanical equipment.Acoustic fault diagnosis is a noncontact measurement method and has a great development space in diagnosis.Simultaneously,there are also many limitations in acoustic fault diagnosis.The testing environment is more stringent,usually in the semi-noise chamber or in the environment without noise pollution.To obtain a better sound signal,it is also essential to extract and analyze its features.The research shows that acoustic characteristics play an important role in fault diagnosis,and its identifiability and distinguishability directly affect the performance of fault diagnosis.Therefore,to improve acoustic fault diagnosis technology's performance,it is necessary to invest a lot of research work in the extraction of acoustic signal characteristics.The traditional method uses vibration signal analysis and processing technology,and statistical characteristics to extract the acoustic signal features.These methods are effective for simple fault diagnosis of some equipment,but it is challenging to work in the case of complex faults or noise interference.This paper studies the physiological and psychoacoustic characteristics of the human auditory system for the electric impact drills' gearbox,proposes a variety of fault feature extraction methods based on the human auditory signal processing mechanism in order to further improve the robustness of acoustic signal features in the noise interference environment,and comprehensively considers the characteristics of the electric wrench parts with few sound signal samples,SVM and LR,which are widely used in industry,are selected to classify the sound features of electric impact drills.The main research work of this paper is as follows:1.To explore the acoustic factors of human ear perception fault information,this thesis proposes a method to extract the human ear's physiological characteristics based on the nonlinear mechanism of the human cochlear in signal processing.The technique is in line with the process of human ear recognition of acoustic signals.It can extract the inherent attribute characteristics of the electric wrench parts' defects and improve the diagnosis and recognition rate.2.The psychoacoustic characteristics reflecting the subjective auditory feeling of human ears are given.Combined with the particularity of the electric wrench's defects,a timevarying loudness based on the model of the average peak to peak ratio of the time-varying loudness spectrum is proposed,and the more recognizable acoustic characteristics are extracted.The experimental results show that the accuracy of the electric wrench parts' defect identification can be improved effectively based on the psychoacoustic factors and the time-varying loudness spectrum peak to peak ratio model.3.Aiming at the engineering of feature selection for fault recognition,a hybrid feature selection method is proposed,which integrates the physiological acoustic features,psychoacoustic features,and time-varying loudness spectrum features of human ear acoustic features.SVM and LR are used to classify the selected feature sets.The results of classification prove that the combination of auditory physiological acoustic features and auditory psychoacoustic features can improve the performance of fault recognition.The hybrid feature selection method based on time-varying loudness spectrum feature can improve the overall defect-recognition rate and the operation efficiency of the diagnosis system.4.The acoustic signal acquisition and fault diagnosis system of electric wrench based on Lab VIEW is designed,which mainly includes the acoustic signal acquisition,acoustic signal analysis,and fault identification test software of electric wrench.The diagnosis system integrates the acoustic feature extraction method and fault classification algorithm mentioned in this thesis.