Earthquake-induced liquefaction is an unpredicted phenomenon that causes catastrophic damages and devastation to the environment, structures, and human life. The assessment of soil liquefaction behavior is a decisive work for geotechnical engineers especially during the designing phase of any civil engineering projects. These decisions implicate tedious and costly experimental procedures and extensive evaluation. Considering these facts, the present study aims to simplify the process of evaluating soil’s liquefaction behavior in a broader domain involving the least experimental datasets. Three PCA (principal component analysis)-based advanced hybrid computational models, namely PCA-ANN, PCA-ANFIS, and PCA-ELM were developed to predict the liquefaction behavior of soils. The dimension reduction technique, i.e. PCA, was used to avoid the multicollinearity effect during the course of the development of the said models. Geotechnical parameters, namely plasticity index, SPT blow count, water content to liquid limit ratio, bulk density, total stress, effective stress, and fine content along with other seismic input variables, such as the ratio of peak ground acceleration and acceleration due to gravity, and magnitude of an earthquake were used to develop the predictive models. The predictive accuracy of the proposed models was evaluated via several fitness parameters. In the end, the best predictive model was determined using a novel tool called Rank Analysis. Based on the results, it has been established that the PCA-ELM hybrid computational model can be considered as a new alternative tool to assist geotechnical engineers in the task of assessing the liquefaction potential of soil during the preliminary design stage in any engineering project.

地震引起的液化是一种不可预测的现象，会对环境、结构和人类生活造成灾难性的破坏和破坏。土壤液化行为的评估对于岩土工程师来说是一项决定性的工作，尤其是在任何土木工程项目的设计阶段。这些决定涉及繁琐且昂贵的实验程序和广泛的评估。考虑到这些事实，本研究旨在简化在涉及最少实验数据集的更广泛领域中评估土壤液化行为的过程。开发了三种基于 PCA（主成分分析）的高级混合计算模型，即 PCA-ANN、PCA-ANFIS 和 PCA-ELM，用于预测土壤的液化行为。降维技术，即PCA，用于避免上述模型开发过程中的多重共线性效应。岩土参数，即塑性指数、SPT 冲击计数、含水量与液限比、容重、总应力、有效应力和细粒含量以及其他地震输入变量，例如峰值地面加速度与重力加速度的比值和地震的震级被用来开发预测模型。所提出模型的预测准确性通过几个适应度参数进行评估。最后，使用一种称为等级分析的新工具确定了最佳预测模型。根据结果​​，