ABSTRACT

The controlled laser air-force detection (CLAFD) technique was developed to explore the feasibility of multi-mechanical property detection of polyurethane by single excitation mode. The adhesiveness, elastic modulus, hardness, resilience, and cohesiveness of polyurethane were predicted by the global variable partial least squares regression (Gv-PLSR) algorithm. Different preprocessing methods were used to preprocess the original laser data. The interval partial least squares regression (I-PLSR) algorithm was used to decrease the influences of the multicollinearity of the global laser variable and increase the stability of the multi-mechanical property prediction models. To further improve the prediction accuracy of the modeling of I-PLSR algorithm, the synergy interval PLSR (Si-PLSR) algorithm was used to combine the intervals with the higher evaluation index root-mean-square error of cross-validation (RMSECV) to predict the multi-properties. The results demonstrated that as a novel mechanical property detection technique, the CLAFD technique predicts in an efficient way. A suitable preprocessing method for the original laser data could greatly improve the effectiveness of prediction. The I-PLSR algorithm was used to improve the model’s stability significantly. Nevertheless, the prediction accuracy decreased. Comparing the I-PLSR algorithm with the Si-PLSR algorithm, the prediction accuracy and the modal stability were optimized by the latter. However, the accuracy was still lower than the Gv-PLSR algorithm. Therefore, the Gv-PLSR was the best algorithm to establish the multi-mechanical properties prediction model. This study provided a new comprehensive, nondestructive, and cross-contamination-free method to evaluate the comprehensive mechanical properties (adhesiveness, elastic modulus, hardness, resilience, and cohesiveness) of materials efficiently, especially for the soft materials such as biomaterial and food material.

摘要

研制了可控激光气压检测（CLAFD）技术，探讨了通过单激发模式对聚氨酯进行多机械性能检测的可行性。通过全局变量偏最小二乘回归（Gv-PLSR）算法预测聚氨酯的粘合性，弹性模量，硬度，回弹性和内聚性。使用不同的预处理方法对原始激光数据进行预处理。使用区间偏最小二乘回归（I-PLSR）算法来减少全局激光变量的多重共线性的影响，并提高多机械性能预测模型的稳定性。为了进一步提高I-PLSR算法建模的预测精度，采用协同区间PLSR（Si-PLSR）算法将区间与交叉验证的较高评价指数均方根误差（RMSECV）组合在一起，以预测多属性。结果表明，作为一种新颖的机械性能检测技术，CLAFD技术可以进行有效的预测。针对原始激光数据的合适预处理方法可以大大提高预测的有效性。I-PLSR算法用于显着提高模型的稳定性。然而，预测精度下降。将I-PLSR算法与Si-PLSR算法进行比较，通过后者优化了预测精度和模态稳定性。但是，准确性仍低于Gv-PLSR算法。因此，Gv-PLSR是建立多力学性能预测模型的最佳算法。这项研究提供了一种新的，全面的，无损的，无交叉污染的方法，可以有效地评估材料的综合机械性能（粘附性，弹性模量，硬度，弹性和内聚性），尤其是对于生物材料和食品材料等柔软材料。