Predicting piezoelectric ceramic quality from indirect sintering indicators using an integrated Bayesian optimization and LightGBM framework

In order to improve the accuracy and timeliness quality prediction, this paper proposes a model for predicting the quality of piezoelectric ceramic sintering specifically for tunnel kiln sintering. This method involves analyzing the relationship between indirect quality indicators during the sintering process and the final quality indicators. Subsequently, a quality prediction model based on indirect quality indicators is established to predict the direct quality indicators. Initially, the input parameters of the model are determined using Pearson correlation coefficients. Then, the quality prediction model is built using the gradient boosting decision tree (GBDT) algorithm from the ensemble learning method called LightGBM. To enhance the prediction accuracy, an algorithm optimization is performed using the leaf-wise leaf growth strategy with depth constraints. Additionally, the BOA algorithm is employed for hyperparameter optimization of the LightGBM model. Through a comparison with the XGBoost algorithm, CatBoost, and Random Forest, the results demonstrate that the proposed BOA-LightGBM model exhibits higher prediction accuracy and generalization ability for direct quality indicators in piezoelectric ceramic quality prediction.