M. Saimurugan

The accurate prediction of the Remaining Useful Life (RUL) of machinery components, especially bearings, is essential for optimizing maintenance strategies, improving system reliability, and enhancing operational efficiency in industrial applications. This study investigates the potential of physics-based features to overcome the limitations of traditional data driven features-based approaches. The data driven methods often face challenges such as poor adaptability to varying operational conditions and a lack of physical interpretability necessary for understanding degradation processes. By embedding domain-specific physical principles into machine learning models, physics-based features enhance the accuracy and robustness of RUL predictions, delivering reliable results across diverse scenarios. This research utilizes vibration signal data from the FEMTO bearing dataset to evaluate Long Short-Term Memory (LSTM) networks using physics-based features, comparing their performance to Support Vector Regression (SVR), Regression Trees (RT), and models based on statistical features. Results demonstrate that physics informed features using LSTM models outperform other approaches, excelling at capturing temporal degradation patterns and delivering robust, interpretable RUL predictions. It is observed that results of LSTM with the involvement of physical insights have achieved greater results when validated against recent studies based on the error measuring indicators, Root Means Square Error (RMSE) and Mean Absolute Error (MAE). This deduces that the excellence of integrating physical insights with LSTM enriches the predictive maintenance, prognostics and health management (PHM) practices.

Rotational machines under variable speed conditions often confront challenges to diagnose various component faults as an existing problem with less exploration of sound signals in this subject area. This premise motivates the paper to contribute the prodigious sound signal features to vibration-based multicomponent fault recognition at speeds ranging from 200 to 2200 rpm. With this novelty, as an added innovation, how each signal-based metric is influencing the classification of different faults in each component is examined. To begin with, speed synchronizing instantaneous frequency (IF) with two signal envelopes of vibration signa ls are individually fed to the machine learning (ML) cla ssifier, such a s Decision Tree (DT), Support Vector Machine-Radial Basis Function (SVM-RBF), and Artificial Neural Network (ANN), to verify the model performance. It is realized that the shaft is highly misclassified by fusing these vibration signal features. To enhance the ML effects and also to reduce the shaft’s misclassification, sound signal features such as Mel Frequency Cepstral Coefficients (MFCCs), spectral centroid, and zero-crossing a re incorporated to evidence its significance. Besides this feature fusion, the comparison of vibration and sound data properties for bearing, gear, and shaft provestheeffectiveness of sound-based parameters towards shaft and gear classification but less impact on bearings. Further reassessment of sound spectral attributes in differentiating gear faults clarifies its utility through SHAP, cross-domain adaptation, covariance, and correlation analysis. This imparts an interpretation of integrating vibration-a nd sound-based features, which attains 83.84% classification accuracy for ANN with a better F1 score than vibration-based features. Thus, distinctively integrating the sound spectral attributes to the vibration signal features under varying speeds demonstrates an improvement in multi-component fault prediction.

Most of the research studies are focused on detecting dynamic changes in non-linear time series with the entropy-based methods and intensifying the fault diagnostic performance in rotational machines. However, the investigations in multi-signal fusion and multi-component fault detection with reduced computation complexity are still unexplored with the assistance of entropy-focused feature-level fusion. Therefore, the paper proposes multi-feature fusion of three reputed entropy measures with instantaneous frequency (IF) and two signal envelopes (lower and upper envelope) of vibration and sound signals to identify bearing, gear, and shaft faults in rotational machinery under speed variation. To begin with, concatenated IF and amplitude demodulated features are fed into three machine learning algorithms (SVM-RBF, DT, and ANN), and this yields moderate results. Recognizing the significance of entropy in quantifying the non-linear systems, Shannon entropy, permutation entropy, and Renyi entropy is coupled with IF and the envelope of time-varying signals for enriching the model assessment. But in consideration of redundant feature removal, the Random Forest (RF) algorithm is utilized to rank these attributes. Prudently ordered IF, Shannon entropy, and lower envelope are more effectuated as three prevailing signal traits in identifying multiple faults with 93.11% accuracy of SVM-RBF. This model is extended by embedding it with BILSTM as a hybrid model, which results in 94.45%. Though this appears to be a desirable outcome, but accounting for the decision-making runtime, BILSTM-CNN with faster computation time of 0.106 seconds, is preferred to be the best fit that delivers 96.36% in potentially recognizing multi-component faults under variable speed condition.

Many research studies have been performed to extract non-stationary signal properties, such as the instantaneous frequency (IF) and envelope, to diagnose machinery faults under time-varying speed conditions. However, accurate IF extraction is difficult from such complex fluctuating signals to identify faults efficiently. Therefore, this paper proposes the statistical complexity measure (SCM) as a feature that evaluates the randomness and complexity without any additional requirement, providing an outstanding result in multi-component fault detection under speed fluctuations. Primarily, the influence of SCM is proved by ranking the features with the help of the random forest algorithm. The fault classification of individual signal characteristics confirms its ranking order as well. Then, it is targeted toward the efficacy of the signal fusion approach in multi-component fault diagnosis. Focusing on the essence of SCM, it is merged with the other three signal properties based on the ranking technique to investigate which one of the feature compositions is more advantageous. The machine learning classifier, SVM-RBF, achieves 94.7%, which proves the potential effects of SCM features in industrial rotational machinery components under variable speed conditions.

Most well-known challenging effects of identifying multiple defects in a rotational machine when speed is varied, are illustriously inspected by many researchers. However, different fusion logics are evolved in the series of research attempt, but not paid much attention in interrogating it for unsteady speed signals.Addressing this literature void, this paper focusses on multi-level fusion strategy with the help of sensor-centric feature integration and Dempster Shafer's (D-S) theory of evidence for uncovering multi-faults in bearings, shafts and gears of rotational machines under speed variational condition. Primarily, instantaneous frequency and envelope from the acquired vibration and sound signals of complex system parts are evaluated and fed to Machine learning (ML) such as Support vector machine (SVM) and K-nearest neighbors (KNN) which verifies the classification performance. The propriety of D-S combination rule is credited by comparing it with the results of other renowned decision-based fusion such as Structural causal model (SCM) and weighted voting method (WVM). This enlightens the effect of imposing DS theory for combining the ML results of integrated vibration and sound signals which is symbolized as multi-level aggregation with 85.76 % accuracy. To illuminate more of this proposed method, a profound investigation over misclassified shaft classes for vibration signals using SVM-RBF are discovered and verified for single-level and multilevel fusion. This yields promising results for multistage fusion approach in rotational machinery fault diagnosis at varying speed rate.

Rotational machines in industries often encounter uncertainties during operation, are monitored and diagnosed through machine condition monitoring. Particularly when speed varies, Artificial Intelligence (AI) provides a great deal of support in recognizing machine faults efficiently. This paper reviews various articles that focused on fault identification in industrial rotating machines under varying speed condition using AI techniques. It investigates various machinery faults, experimental datasets and reputed institutions datasets, complex signals from faulty components and their processing method. Further, the involvement of intelligent based techniques in extracting and selecting suitable features to classify defects are comprehensively reviewed. In addition, as multi-signal fusion strategy is evolving recently for fault diagnosis, the articles comprised of this tactic are also discussed. The objective of this article is to provide a comprehensive understanding of these scenarios to the institutions, industries and researchers, thereby facilitating further exploration.

The statistical relationship between sensor signature features and lubricant solid particle contamination conditions in a spherical roller bearing has been investigated in this study. The influence of particle size and concentration of solid contaminants in lubricant on the RMS parameter of time-domain acoustic emission, vibration, and sound sensor signals are examined. Machine learning algorithms are trained with time domain statistical features derived from sensor signatures to predict the lubricant conditions. Decision trees, bagging tree ensembles, and support vector machines are used to build ML models. Decision Tree models are built using classification and regression tree algorithms with three distinct split criteria, namely gini, towing, and maximum deviance. A bagged tree ensemble model is constructed using the decision tree as a base learner. In the support vector machine, kernel tricking is done to optimize the classification boundaries. Models built using Acoustic emission signature features predict lubricant conditions with better accuracy compared to models constructed using sound and vibration signature features. Feature-level fusion approach is implemented by combining the vibration, sound, and acoustic emission features at the feature level to improve the prediction power of machine learning models. The bagged tree ensemble and support vector machine models, which are trained using fused features, predict lubricant conditions in spherical roller bearings with an accuracy of around 99%.

The predictive maintenance approach in the fault diagnosis of rotating machines is becoming important in industries in order to minimize operational cost and to improve the reliability of machines. Recent studies have focused on developing an effective prediction model to detect machinery faults at various speeds using machine learning techniques. In this research study, machine-learning-based models are developed to detect multi-component faults such as gear faults, bearing faults, and unbalanced shafts, by training the four speed data sets using decision trees, support vector machines and artificial neural networks. The machine learning models are improvised by fusing the vibration signals (X, Y and Z) and sound signals with the feature selection algorithm, minimal redundancy-maximum relevance (mRMR). The research work is extended by training a model with any three speeds and testing the trained model with the remaining speed. The state-of-the-art algorithms used to train all four speeds yield poor performance in the prediction of faults at testing speeds. A linear and quadratic discriminant analysis (QDA) is chosen based on its multivariate discrimination capability for better fault prediction at testing speed. The proposed combination of QDA with mRMR selection of fused vibration and sound feature set perform well in the detection of multiple faults at variable speeds.

A gear box is widely employed in automobiles and industrial machines for transmitting power and torque. It operates under various working conditions for prolonged hours increasing the chance of gearbox failure. Major faults in gear systems are caused due to wear, scoring, pitting, tooth fracture, etc. Gear box failure leads to increases in machine downtime and maintenance costs. The nature and location of such failures can be identified with precision using condition monitoring techniques. In this study, machine condition data are acquired from the gear box using a vibration accelerometer, microphone and acoustic emission sensors under different operating conditions, such as three loading conditions (0 N, 5 N, 10 N) and three rpm variations (500, 750, 1000). The wavelet features are extracted from the acquired vibration, sound and acoustic emission signal, and prominent features are identified. To automate the process of fault diagnosis, machine learning algorithms (artificial neural network, support vector machine, and proximal support vector machine) are utilized. Dual and multi-sensor fusion is implemented with the help of prominent features, to intensify the classification accuracy. The performance of the individual signals, and dual and multi-sensor fused models in gearbox fault diagnosis are compared and discussed in detail.