On-line part deformation prediction based on deep learning

Abstract

Deformation prediction is the basis of deformation control in manufacturing process planning. This paper presents an on-line part deformation prediction method using a deep learning model during numerical control machining process, which is different from traditional methods based on finite element simulation of stress release prior to the actual machining process. A fourth-order tensor model is proposed to represent the continuous part geometric information, process information, and monitoring information, which is used as the input to the deep learning model. A deep learning framework with a Conventional Neural Network and a Recurrent Neural Network has been constructed and trained by monitored deformation data and process information associated with interim part geometric information. The proposed method can be generalised for different parts with certain similarities and has the potential to provide a reference for an adaptive machining control strategy for reducing part deformation. The proposed method was validated by actual machining experiments, and the results show that the prediction accuracy has been improved compared with existing methods. Furthermore, this paper shifts the difficult problem of residual stress measurement and off-line deformation prediction to the solution of on-line deformation prediction based on deformation monitoring data.

Recent Advances on Machine Learning Applications in Machining Processes

“…Misaka et al [89] considered Neural Networks, under the form of CNN, based on camera images of the metal cutting processing for machining parameters extraction, obtaining a model accuracy of 85.5% and a precision of 92.9%. In the same way, a Convolutional Neural Network structure was implemented with ResNet configuration, considering the responsive fixtures and process data as input, in order to allow the Bidirectional LSTM model to predict the part deformation with an error equal to 10.61% [29]. LSTM was also applied to model the dependency between the deviation, tensile force and eccentricity of low-rigidity shaft machining with a MSE of 1.5456 × 10 −5 [90].…”

“…These available datasets allow further extensions, such as Deep Learning (DL) applications that, differently from ML, do not obtain a limited learning rate due to the amount of data. DL applications are used as feature extraction and selection methods (4 to 5) for ML applications, proving to be a reliable and efficient application based on the obtained results, as shown in References [29,30]. Moreover, it allows the implementation of simpler ML methods, such as k-nearest neighbors (k-NN) or simple ANNs or a Decision Tree (DT).…”

Milling Tool Wear Prediction Method Based on Deep Learning Under Variable Working Conditions

“…Wang X [34] uses the LSTM network to predict the fault time sequence based on the historical fault data of the complex system. Zhao Z [35] makes use of CNN's advantages in image information processing and RNN's advantages in dealing with time sequence correlation, and finally, realize the prediction of part deformation.…”

Spark Analysis Based on the CNN-GRU Model for WEDM Process

“…However, they ignored the essential phenomenon in the process of WEMD (or EDM): the generation of electric spark. Although some research gradually begins to apply ANN [46] and other intelligent algorithms to the research of control system and on-line prediction [47,48], there is still high potential for improvement. In other words, the methods based on electrical parameters and traditional intelligent algorithms encounter a bottleneck effect due to the limitations we have mentioned above.…”