Exploratory study on sound quality evaluation and prediction for engineering machinery cabins

The noise produced by engineering machinery in harsh environments poses risks to operators’ health and reduces driving comfort, making sound quality a critical concern. However, most of the current research on sound quality evaluation of engineering machinery is based on the automotive field. However, due to the significant differences in the external environment and cab structure between engineering machinery and automobiles, the evaluation methods in the field of direct transplantation of automobiles have great limitations. To address this issue, a five-level subjective evaluation system was developed, combining the ranking scale method with semantic segmentation. Standardized processing was used to minimize variations caused by inconsistencies in evaluators’ scoring. A dual-ear synchronized measurement technique was applied to collect noise data, addressing the asymmetry of sound sources inside the cabin. Correlation analysis between subjective scores and extracted objective parameters identified key factors affecting cabin sound quality. An optimal parameter combination was determined, and a prediction model based on particle swarm optimization-based random forest (PSO-RF) was proposed. Compared to random forest, support vector regression, and genetic algorithm optimization-based random forest models, the PSO-RF model showed superior accuracy (root mean square error = 0.51) and generalization (mean relative error = 6.61 %). This study introduces an effective method for evaluating and predicting sound quality in engineering machinery cabins. The approach can be applied to other products, supporting the improvement of equipment comfort and market competitiveness.