Wear Condition Monitoring Of Diamond Tools Based On Deep Learning

In recent years,with the improvement of the scientific and technological innovation capacity,the production mode of manufacturing industry has been constantly improving,moving towards the industrial modernization.With the advent of 5G technology and the progress of intelligent technology,the speed of mobile phone replacement is getting faster,and the requirements for the backboard materials of mobile phone are much higher.Zirconia ceramics have received a lot of attention due to the good non-electromagnetic shielding performance.However,the high hardness and brittleness of zirconia ceramic makes the diamond tools for processing this material prone to wear.The wearing of diamond tools will directly affect the processing efficiency and forming accuracy of zirconia ceramic.Therefore,it is of practical significance and value to study the detection of diamond wear status during diamond tool grinding and processing.Wear monitoring of diamond tools is a key technology to achieve intelligent manufacturing.In this paper,a diamond tool grinding test platform for collecting vibration signals and sound signals is built.The diamond tool wearing recognition model based on particle swarm optimization and long short-term memory recurrent neural network is established.The main work in this paper is as follows:(1)Beginning from the current research status in terms of classification of tool wear detection methods,feature extraction and recognition of tool wear state based on deep learning,the results of wear recognition are further verified by the change of grinding force during machining,using the surface morphology of diamond tools under 3D microscope,the surface roughness of zirconia ceramics under 3D optical profiler and the change of grinding force during machining to obtain the indirect evaluation criteria for diamond tools(2)Acceleration sensors and sound sensors were selected to acquire signals during the machining process.Signals of diamond tools with different wear degrees were obtained based on the same experiments.These signals were processed and analyzed and multiple feature values related to diamond tool wear were extracted.The principal component analysis(PCA)was used to feature dimensionality reduction of the feature values to further improve the training accuracy of the model.(3)PSO algorithm is used to optimize the parameters in LSTM model,and a tool wear condition monitoring method based on LSTM theory is proposed.The established LSTM model is trained and tested using different features.The actual results showed that: The recognition rate of multiple signal inputs is higher than that of a single signal input;the recognition rate of multi-domain feature extraction is higher than that of single domain feature extraction;and the recognition rate of multiple signal and multi-domain input after PCA dimensionality reduction optimization is higher than that before optimization.