Remaining Useful Life Prediction from 3D Scan Data with Genetically Optimized Convolutional Neural Networks

Abstract

In the current industrial landscape, increasingly pervaded by technological innovations, the adoption of optimized strategies for asset management is becoming a critical key success factor. Among the various strategies available, the “Prognostics and Health Management” strategy is able to support maintenance management decisions more accurately, through continuous monitoring of equipment health and “Remaining Useful Life” forecasting. In the present study, convolutional neural network-based deep neural network techniques are investigated for the remaining useful life prediction of a punch tool, whose degradation is caused by working surface deformations during the machining process. Surface deformation is determined using a 3D scanning sensor capable of returning point clouds with micrometric accuracy during the operation of the punching machine, avoiding both downtime and human intervention. The 3D point clouds thus obtained are transformed into bidimensional image-type maps, i.e., maps of depths and normal vectors, to fully exploit the potential of convolutional neural networks for extracting features. Such maps are then processed by comparing 15 genetically optimized architectures with the transfer learning of 19 pretrained models, using a classic machine learning approach, i.e., support vector regression, as a benchmark. The achieved results clearly show that, in this specific case, optimized architectures provide performance far superior (MAPE = 0.058) to that of transfer learning, which, instead, remains at a lower or slightly higher level (MAPE = 0.416) than support vector regression (MAPE = 0.857).

Behavioral Change Prediction from Physiological Signals Using Deep Learned Features

“…As reported in Figure 6, the performances exhibited by the joint use of BLSTMAE and TCN with a window of only 10 seconds can be bought with those provided by OCSVM and by BLSTMAE alone with windows of 70 seconds. The architectures of the BLSTMAE and TCN networks have been specifically optimized to work together using the algorithm suggested in [58]. The performance was significantly lower without joint optimization, i.e., using the individually optimized BLSTMAE network.…”

“…The effectiveness of the AE in learning features lies in constraining the latent space to be smaller than the input ( < ), which forcing the neural network to learn the most salient features of the time series data . The network parameters of the BLSTM-AE architecture, whose overview is shown in Figure 4, are optimized using the genetic approach presented by Diraco et al in [58]. For this purpose, a variable number of blocks is considered ranging from 3 to 5, two external and one more internal, each block consisting of BLSTM, fully-connected, Rectified linear unit (ReLU) and dropout layers, where the last two layers are optional.…”

“…As for the BLSTMAE network, also in the case of the TCN network the parameters have been optimized using the genetic approach presented in [58]. The corresponding optimized parameters, i.e., the numbers of convolutional filters , the filter sizes , and the drop out percentages , are reported in Table 4.…”

“…In the joint architecture, the parameters of the BLSTMAE network are optimized again on the basis of the activations extracted from the TCN network, and by following the approach presented in Diraco et al [58]. The optimized architecture is provided in Figure 7, and optimized network parameters, i.e., number of hidden units , output size , and dropping out probability , are reported in Table 5.…”

A Semi-Supervised Approach with Monotonic Constraints for Improved Remaining Useful Life Estimation

“…More recently, deep learning techniques that make use of spatial information, such as Convolutional Neural Networks (CNN), or time dependencies, such as Recursive Neural Networks (RNN) and Long Short-Term Memory (LSTM), have been used. Pretrained CNN are adapted using a genetic algorithms approach for the RUL estimation of punch tools [14]. The LSTM architecture is used to predict a HI and RUL [15].…”