A novel deep capsule neural network for remaining useful life estimation

Abstract

With the availability of cheaper multi-sensor systems, one has access to massive and multi-dimensional sensor data for fault diagnostics and prognostics. However, from a time, engineering and computational perspective, it is often cost prohibitive to manually extract useful features and to label all the data. To address these challenges, deep learning techniques have been used in the recent years. Within these, convolutional neural networks have shown remarkable performance in fault diagnostics and prognostics. However, this model present limitations from a prognostics and health management perspective: to improve its feature extraction generalization capabilities and reduce computation time, ill-based pooling operations are employed, which require sub-sampling of the data, thus loosing potentially valuable information regarding an asset’s degradation process. Capsule neural networks have been recently proposed to address these problems with strong results in computer vision–related classification tasks. This has motivated us to extend capsule neural networks for fault prognostics and, in particular, remaining useful life estimation. The proposed model, architecture and algorithm are tested and compared to other state-of-the art deep learning models on the benchmark Commercial Modular Aero Propulsion System Simulation turbofans data set. The results indicate that the proposed capsule neural networks are a promising approach for remaining useful life prognostics from multi-dimensional sensor data.

Towards Interpretable Deep Learning: A Feature Selection Framework for Prognostics and Health Management Using Deep Neural Networks

“…The second case study used in this work corresponds to the NASA Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) datasets [ 77 ], widely used as benchmark for RUL prediction methods [ 29 , 30 , 31 , 78 , 79 , 80 , 81 ]. They are four datasets containing simulated data of turbofans degradation time series.…”

“…Each sensor measurement is intentionally polluted with noise in order to emulate a real-case scenario. To generate the labels, the same procedure shown in [ 29 ] is used. Thus, degradation is represented as a piecewise linear function of the time cycles, with a maximum RUL number of 125.…”

“…For example, in healthcare, DL models are used for diagnosing and predicting diseases [8][9][10][13][14][15][16][17]. In prognostics and health management (PHM), much like in healthcare, ML and DL models have been used for diagnostics [18][19][20][21][22][23][24][25], prognostics [26][27][28][29][30][31][32] and anomaly detection [33,34] in machinery, showing promising results. In [25], the author proposes a technique called BCAoMID-F (Binarized Common Areas of Maximum Image Differences-Fusion) for feature extraction using thermal images.…”

Using temporal convolution network for remaining useful lifetime prediction

“…Using the pooling operation in CNNs will loss important information related with RUL, so CapsNets model is designed for RUL estimation 39 …”

An Improved PF Remaining Useful Life Prediction Method Based on Quantum Genetics and LSTM

“…A framework for estimating the RUL of mechanical systems is proposed, which is composed of the multi-layer perceptron and multilayer perceptron and evolutionary algorithm for optimizing parameters [27]. Besides, there are many other machine learning algorithms, such as neural networks [28]- [30], capsule neural networks [31], dynamic Bayesian networks [32] and so on.…”