The piston ring nitriding process involves strong parameter coupling, pronounced nonlinearity, and inherent uncertainties, which hinder accurate prediction of nitrided layer hardness. To address this issue, an uncertainty quality prediction method for engineering applications is proposed. Kernel principal component analysis is applied to reduce the dimensionality of process parameters, where the selected principal components achieve a cumulative variance contribution of 85.23 %. Wavelet output-hidden-input feedback Elman neural network is then constructed, and the chicken swarm optimization (CSO) algorithm is employed to optimize network weights and thresholds, improving prediction accuracy. In addition, uncertainty analysis is introduced by transforming process parameters into interval forms using pan-grey number theory, enabling interval prediction of nitrided layer hardness. Validation using 111 sets of real production data demonstrates that the CSO-optimized model reduces the maximum relative error from 10.90 % to 6.97 %, with all predicted values falling within the corresponding prediction intervals. The proposed method enhances both prediction accuracy and reliability, providing a practical tool for quality prediction and process optimization in piston ring nitriding.