PCM mechanical properties

Polymer composite materials (PCMs) are multi-component materials whose reinforcing and binding fillers possess significantly different physical and chemical properties. Mechanical properties are one of the most important indicators that can affect the entire life cycle of PCMs. Experimental measurements are the most common and classical approach to studying the mechanical properties of PCMs; however, they require significant time and money. Considering the resource intensity and duration of the empirical measurement processes of PCM properties, as well as the necessity to determine such properties to meet the high requirements of PCM production, the aim of this work was to create a method for predicting the main properties of PCMs based on machine learning algorithms.
During the study, a dataset was obtained, including information about empirically measured mechanical properties of PCMs, manufacturing technology, binding properties, main fabric properties, main thread properties (warp), and main thread properties (weft). Using the obtained dataset, a machine learning model based on the LASSO algorithm was trained to predict the mechanical properties of PCM using information about manufacturing technology, binding properties, main fabric properties, main thread properties (warp), and main thread properties (weft). The LASSO algorithm demonstrated a high R2 value, reaching 96%. These results indicate the potential application of this algorithm in further research and its important practical significance for improving the processes and quality of manufacturing new PCMs.

聚合物复合材料（Polymer Composite Materials，PCMs）属于多组分材料，其增强填料与粘结填料具有显著差异的物理与化学性质。力学性能是影响聚合物复合材料全生命周期的核心指标之一。实验测试是研究聚合物复合材料力学性能最常用且经典的手段，但该方法需耗费大量时间与资金成本。鉴于聚合物复合材料性能的实验表征过程资源消耗大、周期长，且为满足聚合物复合材料规模化生产的高标准要求，需对其性能进行精准测定，因此本研究旨在构建一种基于机器学习算法的聚合物复合材料主要性能预测方法。
研究过程中构建了一套数据集，涵盖聚合物复合材料力学性能的实验测试数据、制备工艺参数、基体粘结性能、织物基体基本性能、经纱核心性能以及纬纱核心性能等信息。基于该数据集，研究人员训练了一种基于LASSO算法的机器学习模型，通过输入制备工艺参数、基体粘结性能、织物基体基本性能、经纱核心性能以及纬纱核心性能等信息，实现对聚合物复合材料力学性能的预测。该LASSO算法展现出优异的拟合性能，决定系数（R²）可达96%。上述结果表明，该算法在后续研究中具备广阔的应用潜力，同时对优化新型聚合物复合材料的制备工艺与产品质量具有重要的实际应用价值。