Answers are grounded in this paper and cite their sources.
The critical need for detecting multiple faults in automotive suspension systems is driven by the goal of improving ride comfort, stability, and control. This research addresses this challenge by leveraging vibration signals and machine learning techniques to develop a sophisticated fault diagnosis system. A novel stacking-based ensemble approach is employed supported by vibration data acquired using a specially designed experimental setup under constant load conditions. Seven distinct suspension faults along with healthy condition were studied. Features were extracted using statistical, histogram and auto-regressive moving average (ARMA) methods. Extracted features undergo a selection process using the J48 decision tree algorithm, identifying the most important attributes. Classification was conducted in two stages: Standalone Classifiers: Eight standalone classifiers (J48, logistic model tree, multilayer perceptron (MLP), k-nearest neighbours, Naïve Bayes, multinomial logistic regression (MLR), support vector machine and random forest) were evaluated. Stacking-Based Classification: The top four standalone classifiers were ensembled in various combinations to enhance performance. The stacking classifier trained on ARMA features achieved the highest accuracy of 100%, outperforming the best standalone classifier which achieved 99.38% accuracy using MLR with ARMA features. Stacking consistently enhanced classification accuracy across all feature sets. This study demonstrates the efficacy of stacking-based ensemble learning for robust fault diagnosis in suspension systems. Future work will focus on extending the methodology to dynamic, real-world conditions.
No takes yet. Share an insight, caveat, or question.
Raja et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: