Answers are grounded in this paper and cite their sources.
Field reliability of components in distributed electromechanical systems exhibits nonlinear, non-stationary patterns across geographic regions, which makes direct prediction difficult. To address this, taking the air-conditioning compressor as an example, a zonal prediction methodology integrating regional clustering and machine learning for the field failure rate is proposed. First, failure mode and effects analysis systematically identifies influencing factors, and a self-organizing map neural network performs regional clustering to yield homogeneous zones. Second, mutual information quantifies correlations between failure rates and influencing factors, identifying key predictors. Third, the discrete wavelet transform decomposes the failure sequence into low-frequency and high-frequency components. The low-frequency part is predicted using the GM(1,1) model based on Fourier series residual correction. For the prediction of high-frequency components, an enhanced rime optimization algorithm incorporating dynamic opposite learning and cyclone foraging is proposed. It optimizes a radial basis function neural network and an extreme learning machine for separate predictions, and a variable-weight projection-resonance method is then used to combine the results. Finally, all component predictions are summed to obtain the final forecast. Applied to the Zhejiang-Jiangsu-Shanghai region in mainland China, the method achieves mean absolute error and root mean square error values of 4.66 × 10−7 and 5.03 × 10−7, respectively, substantially outperforming benchmark models.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: