Research On Abnormal Sound Detection Method Of Chemical Plant Machinery And Equipment Based On Integrated Learning And Attention Mechanism

Pump equipment provides power for the entire chemical plant assembly line system,and is also one of the most vulnerable and important equipment in the chemical plant assembly line.It has been widely used in the chemical industry field.However,due to the fact that pump equipment often works under harsh conditions such as heavy loads and frequent machine collisions for a long time,this can lead to various failures of pump equipment components,which can lead to abnormalities,unbalanced output,and even downtime within the entire pump equipment,which can affect production efficiency and even lead to major production accidents.If the situation is serious,it can also endanger the safety of personnel.Therefore,it is particularly important to accurately and efficiently detect abnormalities in pump equipment.Currently,the sound abnormality detection technology based on machine learning mainly collects normal or abnormal sound signals from the running sound of pump equipment,and classifies them to extract abnormal information from the running sound of pump equipment,thereby identifying abnormal information during the operation of pump equipment.Due to their limitations,such algorithms have low prediction accuracy.In response to the above issues,this article will improve the existing abnormal sound detection methods for pump equipment during operation based on the idea of attention mechanism and integrated learning,so as to achieve abnormal sound detection for pump equipment and further improve its accuracy.The specific work is as follows:(1)Currently,pump equipment anomaly detection is based on a deep learning model for pump equipment sound anomaly detection,but there are few pump equipment network models that handle noisy data.In order to solve the problem of abnormal detection of pump equipment with noise data,this paper proposes a method for detecting pump equipment sound abnormalities based on attention mechanism.This method adopts the mechanism of attention mechanism,applies the attention mechanism to the residual network,classifies the sounds of pump equipment in two ways,and realizes the detection of abnormal sounds of pump equipment.Its model performance and accuracy have been greatly improved.(2)To address the difficulty of accurately modeling the sound of pump operation in various scenarios using a single model,this paper proposes an ensemble learning based method for detecting abnormal sound in pump equipment.This method collects the sound signal of the pump equipment,annotates and preprocesses it;Extract timedomain and frequency-domain features from the preprocessed dataset based on the different characteristics of the sound signal in the time-domain and frequency-domain when the pump equipment is abnormal;Obtain the time-domain and frequency-domain feature sets respectively through sampling;For the time-domain feature set,residual contraction networks using the time-domain attention mechanism are trained separately to obtain N different base learners;For frequency domain feature sets,residual contraction networks using channel domain attention mechanism are trained separately to obtain M different base learners;Weighted fusion of the N+M different base learners mentioned above to obtain the final classification model;Using a classification model to detect anomalies in pump equipment and determine if there are any faults.This method has achieved abnormal sound detection for pump equipment,and its model performance and accuracy have been greatly improved.