Small-Sample-Oriented Multicondition Fault Diagnosis Framework Based on Classifier- Free Denoising Diffusion Implicit Model With Multiclass Contrastive Learning

Buildings are subject to faults in their heating, ventilation and air-conditioning (HVAC) systems that can lead to excessive energy wastage, poor indoor climate, equipment failures and high maintenance costs. Field studies have shown that employing fault detection, diagnosis and prognosis (FDDP) tools followed up with equipment services and corrections can help achieve up to 40% of energy savings within the HVAC system and improve indoor climate, increase equipment lifecycle and reduce maintenance costs. The increasing adoption of building automation systems (BAS), Internet of Things (IoT) and other smart technologies in recent years have allowed large amounts of data to be continuously collected from building systems. This data-rich environment, along with the surge in data analytics and machine learning tools, has made cost-effective data-driven FDDP strategies possible. Compared to purely physics-based methods, data-driven methods require less explicit knowledge of the underlying physical system, thus are often easier to develop, and can learn certain intricate relationships that exist among data. Within the reported data-driven FDDP methods, there exists a few research gaps: 1) data imputation methods that leverage mutual information from correlated measurements to defy poor data quality from BAS have not been utilized efficiently; 2) there lacks a systematic and scalable fault diagnosis framework that incorporates probabilistic temporal relationships to track fault evolution; 3) existing fault diagnosis strategies typically focus on traditional rule-based control strategies and their scalability for advanced control strategies such as Guideline 36 have not been explored yet; 4) active threats information such as cyber-attacks, are typically not incorporated in an FDDP framework; 5) fault prognosis strategies to preemptively identify gradual faults for predictive maintenance have rarely been studied. This research attempts to address the above-mentioned research gaps through the following: Data Imputation: reported data imputation methods that are suitable for handling and repairing multi-source BAS data are evaluated. Data collected from a medium-sized, mixed-use institution building situated in Stockholm, Sweden and a small commercial building simulated in a laboratory setup is used to evaluate five different data imputation methods. Results demonstrate that incorporating time-lagged cross-correlations within the k-Nearest Neighbor (kNN) method helps to significantly improve the imputation accuracy and minimize the impact of repaired data on data-driven algorithms. Dynamic Bayesian Network (DBN)-based Framework for Cyber-Physical Fault Diagnosis: a DBN framework with discretized conditional probabilities parameters to represent the temporal relationships among building measurements is developed. Both domain knowledge and machine learning methods are used to develop the DBN structure and parameter model. The developed framework is evaluated for both traditional rule-based and Guideline 36 controls using datasets from a real building, a laboratory building, and a virtual testbed. Results show that the developed DBN framework is effective in diagnosing and isolating faults in systems even with different control strategies. The framework also successfully distinguishes whether system abnormalities originate from cyber-attacks or naturally occurring physical faults. Potential future direction to improve fault isolation using modified DBN topological structure is also reported in this study. DBN-based Framework for Fault Prognosis: an extension of the DBN framework in conjunction with Robust Multivariate Temporal (RMT) variate selection is proposed for fault prognosis. The RMT variate selection is used to extract localized temporal features from high dimensional datasets to determine the best inputs for training forecasting models. The expected fault-free behavior of multiple target variates, selected using domain knowledge, is forecasted using incoming data. The prediction errors generated from the forecasting phase are used as evidences in the DBN inference to estimate future fault probabilities. Gradual faults simulated in the virtual testbed are used to evaluate the prognosis framework. Results show that the developed framework is effective in prognosing gradual faults by leveraging the trending growth on the prediction errors. The research presented in this thesis contributes to the overall objective of developing a robust and cost-effective DBN-based framework for fault diagnosis and prognosis of building HVAC systems. Potential solutions to other existing challenges of implementing data-driven FDDP strategies, such as obtaining high-quality datasets, handling and repairing missing data, establishing a baseline model for detecting abnormalities despite other disturbances such as weather and internal conditions changes, and extracting temporal features from timeseries data are also examined.

Industrial fault diagnosis based on active learning and semi-supervised learning using small training set

Failures of rotating machinery, such as bearings and gears, are a critical concern in industrial systems, leading to significant operational downtime and economic losses. A primary research challenge is achieving accurate fault diagnosis under complex industrial noise, where weak fault signatures are often masked by interference signals. This problem is particularly acute in demanding applications like offshore wind turbines, where harsh operating conditions and high maintenance costs necessitate highly robust and reliable diagnostic methods. To address this challenge, this paper proposes a novel Multi-Scale Domain Convolutional Attention Network (MSDCAN). The method integrates enhanced adaptive multi-domain feature extraction with a hybrid attention mechanism, combining information from the time, frequency, wavelet, and cyclic spectral domains with domain-specific attention weighting. A core innovation is the hybrid attention fusion mechanism, which enables cross-modal interaction between deep convolutional features and domain-specific features, enhanced by channel attention modules. The model’s effectiveness is validated on two public benchmark datasets for key rotating components. On the Case Western Reserve University (CWRU) bearing dataset, the MSDCAN achieves accuracies of 97.3% under clean conditions, 96.6% at 15 dB signal-to-noise ratio (SNR), 94.4% at 10 dB SNR, and a robust 85.5% under severe 5 dB SNR. To further validate its generalization, on the Xi’an Jiaotong University (XJTU) gear dataset, the model attains accuracies of 94.8% under clean conditions, 95.0% at 15 dB SNR, 83.6% at 10 dB SNR, and 63.8% at 5 dB SNR. These comprehensive results quantitatively validate the model’s superior diagnostic accuracy and exceptional noise robustness for rotating machinery, establishing a strong foundation for its application in reliable condition monitoring for complex systems, including wind turbines.

To find out the hidden danger in the industrial production process in time, it is necessary to monitor the health condition of the key components of the mechanical system in operation. However, traditional fault diagnosis methods usually adopt manual feature extraction, which not only costs expensively and depends on prior knowledge. Therefore, it is of great significance to study the application of automatic fault identification based on the original vibration signals. Recently, existing studies have shown that most of fault diagnoses are implemented by using deep neural network. Although these methods have achieved satisfactory performances, there are obvious limitations in real applications, that is, the complexity of deep neural network requires a lot of hardware computing resources. This hinders the development of online fault diagnosis tools. To solve this problem, this paper proposes a fault diagnosis model based on lightweight convolutional neural network MobileNet, and realizes an end-to-end intelligent fault classification and diagnosis application. We evaluated the proposed method with the rolling bearing dataset from Western Reserve University. The best average precision, recall and F1 score of ten different bearing health conditions are about 96%, 82% and 88%, respectively. In addition, we also compare the accuracy of the rolling bearing fault diagnosis classification model under the standard ReLU and the improved ReLU. Experimental results show that both obtain good performances, but the improved ReLU reaches the over 96% accuracy more rapidly.

Industrial fault diagnosis often requires open-set domain adaptation, in which target data may contain fault types absent from the source. An enhanced filtering-dynamic matching-coordinated interaction (EFDMCI) framework is introduced to address three practical obstacles: (i) reliance on manually tuned thresholds for unknown detection; (ii) negative transfer caused by indiscriminate alignment; and (iii) weak coupling among modules. Within EFDMCI, enhanced filtering (EF) is used to perform threshold-free, rank-based sample selection, and a binary gate is trained to route unknown samples. Dynamic matching (DM) is used to align marginal and conditional distributions with reliability-aware, dynamically updated weights. Optimization is stabilized by a coordinated interaction scheme that alternates EF and DM updates. Across intra- and cross-machine transfers, mean accuracies of 97.45% and 95.02% were obtained, respectively, together with strong open-set robustness (OS * 97.44%, OS 90.36%, unknown-class accuracy 71.55%, H-score 82.48%), as detailed in the results. These findings indicate that reliable recognition of known and unknown faults can be achieved without manual thresholds, enabling online monitoring and cross-facility deployment.

Achieving asymptotic and concurrent fault diagnosis in hydraulic system remains a challenging endeavor due to the inherent attributes of the hidden occurrence, simultaneous manifestation, coupling, and limited sample size. To address the above issues, this paper proposes a hierarchical multi-output fault detection and diagnosis framework, namely, HMDF, based on a hierarchical learning strategy to leverage an improved convolutional neural network (CNN) and support vector machine (SVM). Both a multi-channel CNN and a multi-branch CNN are employed to extract and downscale features collected by the sensors at diverse sampling frequencies first, and then, such features are subsequently subjected to classification using SVM. The hierarchical learning strategy enables the identification of different fault states, both at the component and the intra-component level. Additionally, a modified whale optimization algorithm is also utilized to optimize the classification process of SVM. Extensive experiments are conducted to test the proposed HMDF with the hydraulic system datasets. Results show that HMDF achieves a diagnostic accuracy of up to 98.9% for the dataset, surpassing traditional methods reliant on manual extraction of time–frequency features, and it also exhibits superior classification performances with a small sample size. The HMDF is expected to offer a generalized framework for the multi-output fault detection and diagnosis in hydraulic systems and other complex components.

Industrial system monitoring includes fault diagnosis and anomaly detection, which have received extensive attention, since they can recognize the fault types and detect unknown anomalies. However, a separate fault diagnosis method or anomaly detection method cannot identify unknown faults and distinguish between different fault types simultaneously; thus, it is difficult to meet the increasing demand for safety and reliability of industrial systems. Besides, the actual system often operates in varying working conditions and is disturbed by the noise, which results in the intraclass variance of the raw data and degrades the performance of industrial system monitoring. To solve these problems, a metric learning-based fault diagnosis and anomaly detection method is proposed. Fault diagnosis and anomaly detection are adaptively fused in the proposed end-to-end model, where anomaly detection can prevent the model from misjudging the unknown anomaly as the known type, while fault diagnosis can identify the specific type of system fault. In addition, a novel multicenter loss is introduced to restrain the intraclass variance. Compared with manual feature extraction that can only extract suboptimal features, it can learn discriminant features automatically for both fault diagnosis and anomaly detection tasks. Experiments on three-phase flow (TPF) facility and Case Western Reserve University (CWRU) bearing have demonstrated that the proposed method can avoid the interference of intraclass variances and learn features that are effective for identifying tasks. Moreover, it achieves the best performance in both fault diagnosis and anomaly detection.

Bearings are the most common mechanical components in machines. Once a bearing fails (or components in it), other adjacent components or the machine itself are effected up to failure. Therefore, bearing health condition is of great interest in practice. Several benchmark datasets are developed to evaluate development in bearings health state (diagnosis) and remaining useful lifetime (prognosis). Among these datasets, Case Western Reserve University (CWRU) dataset is one of the most cited ones used to validate the performance of different diagnostic approaches. Over recent years, a significant amount of research approaches are developed using CWRU data. Most approaches are focused on specific performance parameters like detection rate or accuracy etc. The main problems in connection with CWRU dataset use are: no overview about latest results is available. Furthermore several results published are not complete, for example published accuracies rate without false alarm rates.In this contribution an overview about the development change over the last years, the approaches applied, and specifically the results obtained will be given. Additionally, the new approaches emerging in recent years like deep learning (DL) also in combination with fusion methods and related performance will be given in comparison with conventional machine learning (ML) methods. Special care will be given to the completeness of published results also in combination with shown robustness. As outcome of this contribution the newest and best results are noted, furthermore a recommendation how to complete research work using benchmark dataset will be given. Although most approaches using CWRU dataset as benchmark get high accuracy, for further bearing fault diagnosis research, more and more suitable measures as well as other datasets are needed for increased performance evaluation.

In an intelligent manufacturing context, the smooth operations of mechanical equipment in the production process of enterprises and timely fault diagnosis during operations have become increasingly important. However, the effect of traditional fault diagnosis depends on the feature extraction quality and experts’ empirical knowledge, which is inefficient and costly, and cannot match the needs of mechanical equipment fault diagnosis in intelligent manufacturing. The TSK fuzzy system has a strong approximation capability and the ability to interpret expert knowledge. The broad learning system (BLS) has strong feature extraction and fast computation capabilities. In this paper, we present a new model—the TSK fuzzy broad learning system (TSK-BLS). The model integrates the advantages of the BLS and the fuzzy system at the same time, which can be calculated quickly and accurately by pseudo-inverse and symmetry methods. On the other hand, the model is an embedded model-building mechanism, which extends the application scope of BLS theory. The model was tested on a bearing fault data set, provided by Case Western Reserve University, and the model’s fault diagnosis accuracy was as high as 0.9967. The results were compared with those of the convolutional neural network (CNN) and the BLS models, whose fault diagnosis accuracies are 0.6833 and 0.9133, respectively. Comparison showed that the proposed fault diagnosis model—TSK-BLS, achieved significant improvements.

This work presents a framework for fault detection and diagnosis of multimode industrial processes. It is shown that the proposed method to generate features for the faults provides a generalization that can allow the fault diagnosis in modes for which no training was performed. The proposed method is applied to a sophisticated simulator of a continuous stirred tank reactor (CSTR). Experimental results suggest the usefulness of this approach for the detection of faults under operating conditions not considered during training phase.

Aiming at the problems of fixed convolution kernel size and poor targeting of extracted features in the application of deep learning in fault diagnosis, this paper develops a fault diagnosis framework called adaptive signal diagnosis network (ASDN), in which the EEMD of multi-scale signal decomposition is improved in the adaptive preprocessing stage to adaptively capture the transient changes of signals. Meanwhile, SSA is improved to further extract fault trends and periodic components to optimize the signal representation. In the adaptive deep learning stage, an innovative temporal convolutional network (TCN) with a dynamic adjustment mechanism was developed to enable the neural network to adjust its convolutional kernel size according to different frequency components so as to accurately process signals of different frequencies. Validation on datasets from Case Western Reserve University and Xi'an Jiaotong University shows the superior performance of the proposed method, with diagnostic accuracies of 100% and 97.42% on these two public datasets, respectively.

CFI-LFENet: Infusing cross-domain fusion image and lightweight feature enhanced network for fault diagnosis

A key part of CNC machine tools is the rolling bearing, and thus, it is vital to employ a data-driven approach for fault diagnosis. This paper proposes a two-stage fusion sparse learning algorithm for fault data processing that can identify and diagnose the fault types of rolling bearings based on sensor measurement data. During the feature extraction phase, temporal features of sequential data within the big data are extracted using a Long Short - Term Memory (LSTM) network. Moreover, the classification learning stage contains a new sparse learning algorithm, which applies regularization on stochastic configuration networks (SCN). The iterative learning formula combines the alternating direction method of multipliers (ADMM) with the analysis of the quadratic equations theory. Simultaneously, the model’s inequality supervision mechanism is updated based on convergence analysis. This developed algorithm incorporates the benefits of LSTM in extracting temporal data characteristics, along with the sparsity, ease of convergence, and lightweight nature of SCN. Consequently, it mitigates the shortcomings of deep models in end-to-end applications, particularly in terms of interpretability and structural redundancy, thus making it suitable for deployment on edge devices. Finally, a fusion sparse learning model (LSTM--SCN) is introduced based on the two-stage learning algorithm for rolling bearing fault diagnosis. In the experiments on the benchmark dataset, the optimal sparsity degree of this algorithm for the Sparse Coding Network (SCN) reached 76.66%, which was 30% higher than that of the Pooling-based Sparse Coding Network (PSCN). Moreover, in the experiments based on the dataset of Case Western Reserve University (CWRU), the optimal test classification accuracy achieved was 97.51%, and the optimal sparsity degree for SCN reached 29.39%. These results verify that the proposed algorithm exhibits sparsity, demonstrates effectiveness, and is capable of identifying faults in rolling bearings.

Rolling bearings are vital components in rotary machines where even minor faults can lead to the collapse of the entire mechanical system, resulting in significant economic losses. The condition monitoring and fault diagnosis of rolling bearings are crucial for safe operation of complex industrial systems. Traditional fault diagnosis methods mainly depend on expert experience, which leads to obvious deficiencies in diagnostic efficiency, especially in terms of accuracy. To address the issues of traditional bearing fault diagnosis relying on expert experience and low efficiency and accuracy of fault diagnosis, a bearing fault diagnosis model combining synchrosqueezed wavelet transform (SWT) and SE-ResNet is proposed. Firstly, the one-dimensional non-stationary bearing vibration signal is converted into a high-frequency two-dimensional time-frequency image through SWT, which serves as the input for the convolutional neural network; Secondly, an attention mechanism is introduced to construct an SE-ResNet model for extracting two-dimensional image features, which enhances the attention of important features and thus enhances the network's representation ability; Finally, the results of fault diagnosis are output through Softmax layer. To validate the capacity of the proposed model, experiments with the bearing dataset collected from Case Western Reserve University (CWRU) were conducted. The results show that the transfer learning accuracy of the CWRU dataset is 86.25% to 99%. This method has good feasibility, good convergence performance, and has high practical application value.

In the application of deep learning-based fault diagnosis, more often than not, the network model could perform better with a balanced dataset input, whereby the number of fault data is equivalent to that of normal data. However, in the context of real-world applications, the number of fault data is generally insufficient compared to the normal data. In this study, a new approach for fault diagnosis in unbalanced data sets is proposed using the Gramian angular field (GAF) method. Firstly, the GAF method is employed to convert one-dimensional data into two-dimensional data, which enhances the feature extraction process. Secondly, to balance the sample distribution, fault data is generated using Generative Adversarial Networks (GANs). Finally, the Residual neural network (ResNet) with an attention mechanism is utilized to improve the accuracy of fault diagnosis. The proposed method is experimentally validated using open-source bearing datasets that are published by Case Western Reserve University and the University of Ottawa. The experimental results show that the proposed method has greatly improved fault diagnosis performance in cases of data distribution imbalance, surpassing that of the compared methods.