Fault Prognosis Research Articles

A Comparison of Signal Analysis Techniques for the Diagnostics of the IMS Rolling Element Bearing Dataset

In this paper, a comparison of signal analysis techniques for the diagnostics of rolling element bearings is carried out. Specifically, the comparison is performed in terms of fault detection, diagnosis and prognosis techniques with regards to the first rolling element bearing dataset released by NASA IMS Center in 2014. As for fault detection, it is obtained that RMS value, Kurtosis and Detectivity, as statistical parameters, are able to properly detect the arising of the fault on the defective bearings. Then, several signal processing techniques, such as deterministic/random signal separation, time-frequency and cyclostationary analyses are applied to perform fault diagnosis. Among these techniques, it is found that the combination of Cepstrum Pre-Whitening and Squared Envelope Spectrum, and Improved Envelope Spectrum, allow the faults to be correctly identified on specific bearing components. Finally, the Correlation, Monotonicity and Robustness of the previous statistical parameters are computed to identify the most accurate tools for bearing fault prognosis.

Exploring the potential of time-series transformers for process modeling and control in chemical systems: An inevitable paradigm shift?

The last two years have seen groundbreaking advances in natural language processing (NLP) with the advent of applications like ChatGPT, Codex, and ChatSonic. This revolution is powered by the development of cutting-edge transformer models that leverage multiheaded attention mechanisms, positional encoding, and highly efficient transfer learning capabilities. Despite these remarkable advances, there is still work to be done to fully realize the practical applicability of transformers in chemical systems. Thus, we are excited to present our latest work, which highlights the immense potential of transformers for non-trivial multivariate time-series prediction tasks with high-value implications in process monitoring, control, and optimization. Specifically, impressive prediction capabilities of first-generation time-series transformers (TSTs) were demonstrated by developing, testing, and comparing TSTs with existing models. Further, the practical applicability of TSTs was highlighted by developing a first-of-a-kind TST-based model predictive controller (MPC). More importantly, the current work provides a concrete foundation for exploring promising new directions, such as the development of large-scale TSTs leveraging transfer learning for modeling of new process equipment, and plant-level multisource aggregative cognitive models for fault prognosis and prevention. We are excited to see what the future holds as we continue to push the boundaries of what is possible with these ‘transformer-tive’ technologies.

Intelligent Fault Prognosis Method Based on Stacked Autoencoder and Continuous Deep Belief Network

Mechanical fault prediction is one of the main problems in condition-based maintenance, and its purpose is to predict the future working status of the machine based on the collected status information of the machine. However, on one hand, the model health indices based on the information collected by the sensors will directly affect the evaluation results of the system. On the other hand, because the model health index is a continuous time series, the effect of feature learning on continuous data also affects the results of fault prognosis. This paper makes full use of the autonomous information fusion capability of the stacked autoencoder and the strong feature learning capability of continuous deep belief networks for continuous data, and proposes a novel fault prognosis method. Firstly, a stacked autoencoder is used to construct the model health index through the feature learning and information fusion of the vibration signals collected by the sensors. To solve the local fluctuations in the health indices, the exponentially weighted moving average method is used to smooth the index data to reduce the impact of noise. Then, a continuous deep belief network is used to perform feature learning on the constructed health index to predict future performance changes in the model. Finally, a fault prognosis experiment based on bearing data was performed. The experimental results show that the method combines the advantages of stacked autoencoders and continuous deep belief networks, and has a lower prediction error than traditional intelligent fault prognosis methods.

A State Space Approach to Decentralized Fault SE-Coprognosability of Partially Observed Discrete Event Systems.

The problem of fault prognosis in the context of discrete event systems (DESs) is a crucial subject to study the security and maintenance of cyber-physical systems. In this article, the decentralized fault prognosis of partially observed DESs is analyzed with a universal state-estimate-based protocol. It follows (M,K) as the performance bound of any expected decentralized prognosers, where any fault can be predicted K steps before its occurrence and the fault is guaranteed to occur within M steps once a corresponding fault alarm is issued. To determine whether expected decentralized prognosers exist, the notion of state-estimate-coprognosability (SE-coprognosability) under the case of one fault type is proposed. Compared with existing other kinds of coprognosability, SE-coprognosability is a more generalized concept. Meanwhile, combining the formal method and algebraic state space approach, a novel state estimation algorithm is presented and based on which, the verification of SE-coprognosability is also solved.

Thermal fault prognosis of lithium-ion batteries in real-world electric vehicles using self-attention mechanism networks

Battery temperature is directly related to various thermal faults and thermal runaway of a battery system in electric vehicles. Its prediction is critical to the safety and durability of the battery system. This paper presents a novel clustering-based data partitioning method and a partitioned all-season coverage model established by the spiral self-attention neural network to perform real-time multi-forward-step temperature prediction for battery systems. A yearlong real-world operational dataset of an electric taxi is deployed. The proposed series-temperature method can generate a generic and non-interfering model for all temperature probes. The results demonstrate that the partitioned training model has excellent online prediction performance for non-differentiated real-world vehicle operation in all climates. The prognosis feasibility and reliability of the model for various temperature anomalies are also validated, providing a reliable tool for battery safety by determining the advent of thermal faults and mitigating the risk of thermal runaway. Furthermore, the flexible and adjustable predicted horizons could provide sufficient processing time for predicting and preventing battery anomaly/fault. This method's superiority, stability, flexibility, and robustness are verified under different hyperparameters and operating patterns.

Fault Prognosis and Isolation of Lithium-Ion Batteries in Electric Vehicles Considering Real-Scenario Thermal Runaway Risks

Advanced safe battery storage systems with health prognostic performance are vital for electric vehicles. Various faults of lithium-ion batteries are usually undetectable in their early stage due to their concealment and graduality. This article presents a real-time fault diagnosis and isolation scheme for real-scenario batteries using the normalized discrete wavelet decomposition. The early frequency-domain features of the fault signals are extracted utilizing the high-frequency detail wavelet components, and a multilevel fault prognosis strategy is developed considering complex charging/driving characteristics under real-vehicle operating conditions. The verification results, implemented on loose wire connection batteries and real-scenario thermal runaway batteries, demonstrate that the proposed method can accurately extract and locate the hidden fault signals even under small magnitudes and effectively detecting and isolating battery faults before thermal runaway. Furthermore, significant reliability and stability of the proposed method are verified on more real-vehicle operation data, enabling online monitorable and traceable of battery faults before triggering thermal runaway, safeguarding drivers and passengers in real-world vehicular operation.

Detecting bearing failures in wind energy parks: A main bearing early damage detection method using SCADA data and a convolutional autoencoder

AbstractWind energy maintenance and operation costs can total millions of dollars each year in an average industrial‐size wind park. Therefore, moving from preventive and corrective maintenance to predictive maintenance is imperative in the wind energy sector. This paper contributes to this challenge by providing a main bearing early damage detection technique that exclusively uses standard supervisory control and data acquisition (SCADA) data (10‐min average) and a convolutional autoencoder with the following contributions. (i) Entirely semisupervised (not requiring the labeling of data through work order logs and avoiding the problem of data imbalance between classes) based only on healthy data, thus expanding its range of application (even when the failure of interest has never occurred in the park before). (ii) Validated using real‐world SCADA data and shown to be resistant to seasonality, and operational and environmental conditions. (iii) Reliable predictions with minimum false alarms thanks to specially designed fault prognosis indicators based on the image mean square error metric. (vi) The early warning is achieved months in advance, thus providing adequate time for plant operators to plan properly. (v) The main use of exogenous variables in the model (variables that are not affected by other variables, e.g., wind speed, wind turbulence, and ambient temperature) guarantees the detection of damage directly related only to the low‐speed shaft temperature (the only nonexogenous variable used by the stated model). (vi) Finally, the proposed strategy is validated in a wind park made up of 12 wind turbines.

Establishing a Real-Time Multi-Step Ahead Forecasting Model of Unbalance Fault in a Rotor-Bearing System

Recently, prognostics and health management (PHM) has garnered a lot of attention in the industrial sector for its cost-effective maintenance and safe operation of the system. In this regard, vibration-based predictive maintenance using sensors plays a significant role in the diagnosis and prognosis of various faults. The need of the hour is to know when and which part must be replaced in advance for efficient and reliable operation. Unbalance is one major fault acting on any rotary system leading to excessive vibration and causing various other faults developing early failure in components directly or indirectly. In this paper, we show how a prognostic model can be built for the identification of future unbalance trend of a rotor-bearing system with the aid of a mathematical model of the system and statistical/machine learning methods. The prognostic model developed is used to forecast the unbalance time-series data of an industrial turbine rotor in real-time which forecasts the month ahead unbalance values. The proposed model is verified for prognostic analysis using datasets from a local plastic company. After careful examination of the results, it is concluded that the proposed model can aid in precisely detecting future system unbalance.

A Data-Driven Framework for Small Hydroelectric Plant Prognosis Using Tsfresh and Machine Learning Survival Models.

Maintenance in small hydroelectric plants (SHPs) is essential for securing the expansion of clean energy sources and supplying the energy estimated to be required for the coming years. Identifying failures in SHPs before they happen is crucial for allowing better management of asset maintenance, lowering operating costs, and enabling the expansion of renewable energy sources. Most fault prognosis models proposed thus far for hydroelectric generating units are based on signal decomposition and regression models. In the specific case of SHPs, there is a high occurrence of data being censored, since the operation is not consistently steady and can be repeatedly interrupted due to transmission problems or scarcity of water resources. To overcome this, we propose a two-step, data-driven framework for SHP prognosis based on time series feature engineering and survival modeling. We compared two different strategies for feature engineering: one using higher-order statistics and the other using the Tsfresh algorithm. We adjusted three machine learning survival models-CoxNet, survival random forests, and gradient boosting survival analysis-for estimating the concordance index of these approaches. The best model presented a significant concordance index of 77.44%. We further investigated and discussed the importance of the monitored sensors and the feature extraction aggregations. The kurtosis and variance were the most relevant aggregations in the higher-order statistics domain, while the fast Fourier transform and continuous wavelet transform were the most frequent transformations when using Tsfresh. The most important sensors were related to the temperature at several points, such as the bearing generator, oil hydraulic unit, and turbine radial bushing.

Semi-Supervised Framework with Autoencoder-Based Neural Networks for Fault Prognosis

This paper presents a generic framework for fault prognosis using autoencoder-based deep learning methods. The proposed approach relies upon a semi-supervised extrapolation of autoencoder reconstruction errors, which can deal with the unbalanced proportion between faulty and non-faulty data in an industrial context to improve systems' safety and reliability. In contrast to supervised methods, the approach requires less manual data labeling and can find previously unknown patterns in data. The technique focuses on detecting and isolating possible measurement divergences and tracking their growth to signalize a fault's occurrence while individually evaluating each monitored variable to provide fault detection and prognosis. Additionally, the paper also provides an appropriate set of metrics to measure the accuracy of the models, which is a common disadvantage of unsupervised methods due to the lack of predefined answers during training. Computational results using the Commercial Modular Aero Propulsion System Simulation (CMAPSS) monitoring data show the effectiveness of the proposed framework.

Bearing Fault Prognosis Method Based on Priori Knowledge-Enhanced Particle Filter

A priori knowledge-enhanced particle filter (PKE-PF) method is proposed to solve the problem of particle impoverishment in bearing fault prognosis with incomplete data. Based on the existing bearing life degradation data and the parameter transfer method in the transfer learning theory, particle initialization optimization, which is very important in the PF method, is carried out to effectively improve particle effectiveness and avoid the problem of premature particle exhaustion. Based on the whole-life degradation experiment of rolling bearings, the validation results show that the traditional PF method and its improved method are prone to particle exhaustion, which seriously affects the fault prediction results. The PKE-PF method proposed in this paper can effectively avoid the problem of premature particle depletion and obtain a more ideal fault prognosis results.

A novel transformer-based multi-variable multi-step prediction method for chemical process fault prognosis

As the digitalization of process industry deepens, process fault detection and diagnosis (FDD) is an essential tool to ensure safe production in chemical industries. However, FDD may have a long detection delay for some chemical faults. Process fault prognosis methods could predict the occurrence of faults in advance, which would give operators more time and reduce the impact of faults. Nevertheless, many fault prognosis methods still suffer from fixed or insufficient prediction time ahead, which greatly confines their usage in critical scenarios. In this paper, we propose a novel Transformer-based multi-variable multi-step (TMM) prediction method for chemical process fault prognosis. Specifically, Transformer models are trained to predict the change of process variables at the next step, and iterative forecasting is used to predict multi-step changes of process variables. Finally, extensive evaluation of applications in a continuous stirred tank heater (CSTH) system and the Tennessee Eastman process (TEP) demonstrates that the proposed TMM prediction method shows high prediction accuracy and early fault prognosis, compared with representative statistical methods and other advanced deep learning methods.

State-Partial Accurate Voltage Fault Prognosis for Lithium-Ion Batteries Based on Self-Attention Networks

Multiple faults in new energy vehicle batteries can be diagnosed using voltage. To find voltage fault information in advance and reduce battery safety risk, a state-partitioned voltage fault prognosis method based on the self-attention network is proposed. The voltage data are divided into three parts with typical characteristics according to the charging voltage curve trends under different charge states. Subsequently, a voltage prediction model based on the self-attention network is trained separately with each part of the data. The voltage fault prognosis is realized using the threshold method. The effectiveness of the method is verified using real operating data of electric vehicles (EVs). The effects of different batch sizes and window sizes on model training are analyzed, and the optimized hyperparameters are used to train the voltage prediction model. The average error of predicted voltage is less than 2 mV. Finally, the superiority and robustness of the method are verified.

Self adaptive learning scheme for Fault prognosis in oil wells and production & service lines

The production of Oil & Gas from underground reservoirs involves chemical and mechanical processes that affect well drilling and operation. Many of these processes may eventually cause a problem with the well, resulting in a decrease in production or in equipment failure. This paper deals with fault prognosis during the practical operation of Oil & Gas wells. This work focus on the remaining useful life prediction of the “Spurious Closure of the Downhole Safety Valve” fault. This paper proposes a scheme based on the use of unsupervised machine learning approach and a drift detection mechanism is employed in order to predict the time to failure, real fault scenarios data are used, the proposed scheme is evaluated using different prognosis performance metrics.

A dynamic-inner LSTM prediction method for key alarm variables forecasting in chemical process

With the increase in the complexity of industrial system, simply detecting and diagnosing a fault may be insufficient in some cases, and prognosing the fault ahead of time could have a certain necessity. Accurate prediction of key alarm variables in chemical process can indicate the possible change to reduce the probability of abnormal conditions. According to the characteristics of chemical process data, this work proposed a key alarm variables prediction model in chemical process based on dynamic-inner principal component analysis (DiPCA) and long short-term memory (LSTM). DiPCA is used to extract the most dynamic components for prediction. While LSTM is used to learn the relationship and predict the key alarm variables. This work used a simulation data set and a real hydrogenation process data set for applications and explained the model validity from the essential characteristics. Comparison of results with different models shows that our model has better prediction accuracy and performance, which can provide the basis for fault prognosis and health management.

Coupling mechanism between systematic elastic deformation and gear surface damage

Understanding the coupling mechanism between systematic elastic deformation and gear tooth surface damage is particularly helpful for us to explore the life-cycle surface damage evolution process. A deformable body dynamic model of gear transmission systems is developed to acquire the systematic elastic deformation. Then the equivalent meshing deviation parameters, which are used to quantify the systematic deformation, are imported into a hybrid finite element-analytical contact model to acquire the load distribution under non-ideal meshing conditions. Considering the concurrent effects of surface wear and pitting, a generalized surface damage model is developed based on the transient elastohydrodynamic lubrication theory. An interactive calculation scheme is built to realize the damage simulation over the life-cycle evolution process. Distinguishing from the conventional models, the proposed multi-physics model bridges the macroscopic systematic deformation and microscopic behaviours (local contact state and damage initiation mechanism). A two-staged gear test rig is utilized to verify the proposed model. Based on the proposed model, the coupling mechanism amongst the tribo-dynamic behaviours, systematic elastic deformation and life-cycle fault characteristics are analysed. It is revealed that the bending-induced asymmetric load distribution will lead to severe pitting at one edge of the gear surface; The life-cycle dynamic performance can be significantly improved by optimizing the supporting configuration. This work can provide the theoretical basis for the tooth fault prognosis, physics-based condition monitoring and configuration design.

Artificial Intelligence Enhanced Two-Stage Hybrid Fault Prognosis Methodology of PMSM

Fault prognosis based on single model is generally inaccurate due to the varying working conditions. A multistage fault prognosis methodology combining stage identification with Bayesian networks (BNs) and time series approach with particular emphasis on the autoregressive moving average (ARMA) model is proposed to solve this problem. In the first stage, degradation data are identified, and outliers are marked by the Euclidean distance. Degenerate attributes of outliers are finely identified by BNs and matched to the corresponding model. In the second stage, the ARMA model is used for prognosis according to the results of the fine identification. Subsequently, the double-precision identification and ARMA submodel prognosis are carried out alternately throughout the prognosis process. Three degradation types of permanent magnet synchronous motor are simulated to verify the applicability of the method. Result shows that it can track the changes in the degradation in time and obtains better results.

Combining Online Diagnosis and Prognosis for Safe Controllability

In this article, we combine fault diagnosis and prognosis to generalize the notion of safe controllability of discrete-event systems. To do so, we reformulate the notions of safe diagnosability, prognosability, safe controllability by diagnosis, and safe controllability by prognosis in the context of strings. Moreover, we combine these notions to introduce the concept of safe controllability by diagnosis or prognosis, or simply DP-safe controllability. We show that a language can be DP-safe controllable even if it is not safe controllable either only by diagnosis or only by prognosis. Thus, the DP-safe controllability can be considered a generalization of the safe controllability concept found in the literature. If a DES is DP-safe controllable, to achieve fault tolerance using an active approach, reconfiguration actions could be forced based not only on online fault diagnosis, but also on online fault prognosis. Thus, our approach outperforms the previous ones, since it provides additional control options to keep the system away from forbidden zones and to switch from the nominal supervisor to a postfault-detection supervisor designed to achieve postfault performance objectives. Necessary and sufficient conditions for DP-safe controllability are presented and an example is used to illustrate the introduced concepts.

Multi-time-step and multi-parameter prediction for real-world proton exchange membrane fuel cell vehicles (PEMFCVs) toward fault prognosis and energy consumption prediction

Proton exchange membrane fuel cell vehicle (PEMFCV) is considered to be a promising way to cope with the environment pollution and energy exhaustion. But PEMFCV is also suffering from some difficulties for the real-world application, including different kinds of faults and hydrogen refueling. This raises requirement for timely and accurate fault prognosis and energy consumption prediction. However, limited and sparse parameters obtained by onboard sensors and random influencing factors during the real-world vehicular operation make the PEMFCV hardly be modelled. To cope with the issue, this study firstly puts forward a vehicle state-driving behavior factor (VDF) construction of PEMFCV to consider as comprehensive factors as possible and uses maximal information coefficient (MIC) to extract related factors for model training. Then a data-driven model is constructed to achieve synchronous multi-time-step and multi-parameter prediction for the fuel cell and hydrogen system in real-world PEMFCVs. The model considers noise denoising, spatial feature processing and temporal feature processing by combining convolutional neural network (CNN) and gated recurrent unit neural networks (GRU). To optimize numerous hyperparameters of constructed data-driven model, a “discrete gradient-based optimization” method (DGO) is first proposed to achieve local-optimized hyperparameters as well as reduce the time complexity of grid searching. Based on the predicted parameters of PEMFCVs, a scheme is designed for fault prognosis and energy consumption prediction. All the procedure is trained and verified by real-world operation data to reflect the real-world applicable conditions for different seasons and vehicles. Results show that the proposed model can achieve accurate 30-time-step synchronous fuel cell temperature, hydrogen temperature, and hydrogen pressure prediction with mean-relative-errors (MREs) of 0.54%, 0.85% and 0.71%; The fault can be accurately prognosed five minutes ahead with MRE of 0.71% to provide driver sufficient time to deal with the fault; The small MRE of 6.3% for energy consumption prediction also indicates the effectiveness of proposed model. The proposed whole method can be applied in the real-world PEMFCVs for long-time-step parameter prediction, fault prognosis, and energy consumption prediction. The proposed DGO can be used to optimize other data-driven methods.

Main Bearing Fault Prognosis in Wind Turbines based on Gated Recurrent Unit Neural Networks

The transition from onshore to offshore wind farms is an imminent fact in the future. It supposes to face hard challenges like difficulties to carry out offshore maintenance operations due to increased downtime (because of several causes like continuously bad environmental conditions) on wind farms. That is why, there is a need to improve maintenance and monitoring practices like those involved in condition-based area. This work proposes a methodology based on three key points: (i) a semi-supervised model built from a gated recurrent unit (GRU) neural network and by using only healthy real SCADA data, (ii) propose a fault prognosis indicator (FPI) to trigger warnings or fault alarms as such, and (iii) detect the main bearing fault several months in advance on a faulty wind turbine. The reported results show the excellent performance of the GRU trained model to predict the main bearing temperature as output by exploiting the capabilities of GRUs (recurrent-based neural networks) to decide what information to forget or preserve through time. In the FPI construction, the use of exponentially weighted moving average (EWMA) helps at the results to avoid the presence of false alarms that is very useful in any detection strategy. Finally, the stated methodology lets to detect the main bearing fault on a WT two months in advance at least, which contributes to plan maintenance actions ahead of time. Furthermore, in this way, the lifespan of this large component may be extended and wind turbine’s uptime may increase in a significant percentage.