IEEE Transactions on Industrial Informatics · 2021 · 280 citations · 32 references
ComputingComputer ScienceManufacturing EngineeringComputer VisionMachine LearningSensor FusionDeep LearningDiagnosisDiagnosticsFault DetectionFault DiagnosisImage FusionAccuracy-computational Cost TradeoffConvolutional Neural NetworksDeep Learning-based DiagnosisMultisensor DiagnosticsMultisensor FusionMultisensor Image FusionNeural Network-based DiagnosisResidual Networks
Mechanical fault diagnosis is essential for safety and cost savings, and advances in data transmission and sensor technologies enable the acquisition of large multisensor datasets. The study aims to develop an intelligent fault diagnosis approach that fuses multisensor data with a convolutional neural network. The method converts multisensor signals into RGB images via PCA, feeds them into a residual‑network CNN, and is validated on two datasets. The approach achieves higher accuracy than other deep‑learning methods.
Diagnosis of mechanical faults in manufacturing systems is critical for ensuring safety and saving costs. With the development of data transmission and sensor technologies, measuring systems can acquire massive amounts of multisensor data. Although deep learning (DL) provides an end-to-end way to address the drawbacks of traditional methods, it is necessary to do deep research on an intelligent fault diagnosis method based on multisensor data. In this article, a novel intelligent diagnosis method based on multisensor fusion (MSF) and convolutional neural network (CNN) is explored. First, a multisignals-to-RGB-image conversion method based on principal component analysis is applied to fuse multisignal data into three-channel red−green−blue (RGB) images. Then, an improved CNN with residual networks is proposed, which can balance the relationship between computational cost and accuracy. Two datasets are used to verify the effectiveness of the proposed method. The results show that the proposed method outperforms other DL-based methods in terms of accuracy.
32
Gradient-based learning applied to document recognition
Yann LeCun, Léon Bottou, Yoshua Bengio et al. · Proceedings of the IEEE · 1998 · 58.2K citations · Full text
Deep Sparse Rectifier Neural Networks
Xavier Glorot, Antoine Bordes, Yoshua Bengio · 2012 · 5.4K citations