Abstract:Remaining useful life (RUL) prediction, as an essential task for prediction and health management (PHM), can improve the reliability of degradation systems. In this paper, multi-dimensional mechanical monitoring readings are treated as natural language sequences and inputted to a natural language processing (NLP) model, Transformer, to extract semantic compression vectors. The health indicators (HI) are constructed from the changes of compressed semantic vectors during the mechanical system operation, reflecti… Show more
“…The condition-based * Author to whom any correspondence should be addressed. maintenance (CBM) method needs to measure information related to the failure and reflects the machine's health status at all times [2]. Obviously, remaining useful life (RUL) prediction is a significant technology for application of CBM.…”
Bearing remaining useful life (RUL) prediction has always been a central topic in the industry field, which is aimed to optimize system safety and sustainability. The validity of the prediction model and the accuracy of the prediction results are affected by mid-term singularities and terminal mutations, under the time-domain bearing vibration information. In this paper, a network structure-Cascaded dilated convolution vision informer (CDC-Vii) is put forward to precisely forecast the RUL of bearings, which uses the time-frequency fault features as input. CDC-Vii breaks the limitation of the Original Informer which is only sensitive to time series information. An adaptive fault frequency band selection (AFFBS) algorithm is proposed, which can reduce training time while utilizing rich time-frequency information. Based on the Informer architecture, the attention mechanism is improved to form vision subsampling ProbSparse self-attention(VSPS). VSPS can precisely assign spatial attention weights and reduce computational complexity. At the same time, a truncated relative position encoding (TRPE) technique is proposed to strengthen the position dependence between attention information. Moreover, Cascaded dilated convolution (CDC) enhances the image contrast of faulty frequency bands while enlarging the use of receptive field. Experiments on extensively utilized two bearing datasets reveal that CDC-Vii surpasses the advanced RUL prediction models.
“…The condition-based * Author to whom any correspondence should be addressed. maintenance (CBM) method needs to measure information related to the failure and reflects the machine's health status at all times [2]. Obviously, remaining useful life (RUL) prediction is a significant technology for application of CBM.…”
Bearing remaining useful life (RUL) prediction has always been a central topic in the industry field, which is aimed to optimize system safety and sustainability. The validity of the prediction model and the accuracy of the prediction results are affected by mid-term singularities and terminal mutations, under the time-domain bearing vibration information. In this paper, a network structure-Cascaded dilated convolution vision informer (CDC-Vii) is put forward to precisely forecast the RUL of bearings, which uses the time-frequency fault features as input. CDC-Vii breaks the limitation of the Original Informer which is only sensitive to time series information. An adaptive fault frequency band selection (AFFBS) algorithm is proposed, which can reduce training time while utilizing rich time-frequency information. Based on the Informer architecture, the attention mechanism is improved to form vision subsampling ProbSparse self-attention(VSPS). VSPS can precisely assign spatial attention weights and reduce computational complexity. At the same time, a truncated relative position encoding (TRPE) technique is proposed to strengthen the position dependence between attention information. Moreover, Cascaded dilated convolution (CDC) enhances the image contrast of faulty frequency bands while enlarging the use of receptive field. Experiments on extensively utilized two bearing datasets reveal that CDC-Vii surpasses the advanced RUL prediction models.
“…By constructing MFF-HI with an MFF depth network and leveraging WTCN with a new loss function, the proposed method achieved accurate RUL prediction for bearings with higher prediction accuracy across various datasets. Duan et al [41] introduced a method for RUL prediction by treating mechanical monitoring readings as natural language sequences, utilizing a transformer model to extract semantic compression vectors for constructing health indicators (HIs) and achieving RUL prediction through weighted summation. Chen et al [42] developed a novel spatial attention-based convolutional transformer for precise bearing RUL prediction without the need for feature engineering.…”
Artificial intelligence (AI) has achieved significant progress in recent years and its applications cover a wide range of fields such as computer vision, natural language processing, autonomous driving and medical diagnosis. In industry, the rapid development of real-time-sensor measurement techniques promotes equipment surveillance and maintenance into the era of big data. It is still challenging to manually analyze this big data and establish general physical modelling by using the information hidden in these data. To this end, AI technology, such as machine learning and neural networks, has emerged as a promising tool to extract useful knowledge from measured data, and to tackle the real-time monitoring, diagnosis problems as well as enhancing the health management and reliability of modern industrial equipment. With the vision of establishing a strong link between AI and industrial equipment surveillance and maintenance, this special feature is designated to select, organize and exhibit the latest research progress on the cutting-edge research topics relevant to AI-Enabled Industrial Equipment Monitoring, Diagnosis and Health Management. Submissions for this topic in Measurement Science and Technology are open from 23 March 2022 to 30 September 2022 and contains 42 outstanding papers in above research fields.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.