The density of methyl diethanolamine (MDEA), governed by temperature and pressure, plays a crucial role in optimizing chemical engineering processes, demanding precise predictive models. This research utilizes a gradient boosting decision tree (GBDT) framework, enhanced by four advanced optimization techniques: batch Bayesian optimization (BBO), evolution strategies (ES), Bayesian probability improvement (BPI), and Gaussian processes optimization (GPO). The algorithm is developed utilizing a dataset of 160 experimental samples, with 90% designated for training and 10% for testing, employing temperature and pressure as input parameters to estimate MDEA density. To counter overfitting, k-fold cross-validation is implemented throughout training procedure. The effectiveness of every optimization method is evaluated through computational runtime and performance indicators, including R-squared (R²), mean squared error (MSE), and average absolute relative error (AARE%). Correlation analysis indicates a significant negative relationship between temperature and MDEA density (correlation coefficient: –0.71), while pressure shows a moderate positive association (0.56). Among the optimization strategies, GBDT-BBO yields the highest precision, with an R² of 0.996223 for the training set and 0.982056 for the test set, outperforming other methods. In terms of computational efficiency, GPO proves the fastest, requiring 147.9 seconds, while BBO is the slowest at 199.5 seconds. Sensitivity analysis highlights the effect of every input variable on MDEA density, demonstrating the robustness of data-driven approaches in tackling complex systems. These models provide dependable tools for predicting MDEA density, minimizing requirement for expensive, time-intensive, and labor-heavy experimental methods.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号