Scholarlink

<p>For the coordinated on-machine measurement of curved surfaces, the accurate prediction of the number of measuring points (MPs) could provide an important theoretical time basis for machining-inspection planning, workshop production scheduling, etc. To address the problem that the prediction of the required number of MPs for curved surfaces in relation to their complexity, machining accuracy, and other ambiguous characteristics, an adaptive neuro-fuzzy inference system (ANFIS) model optimized by the improved particle swarm optimization (IPSO) algorithm is proposed. Firstly, the input and output variables of the prediction model are defined as the surface measurement descriptor (SMD) by analyzing the impact indicators when measuring complex surfaces. Then, the IPSO algorithm with adaptive weights and particle acceleration updates is applied to optimize the ANFIS model with an organic composition of fuzzy logic knowledge and neuronal network. Finally, the prediction ability of the back propagation (BP) model, ANFIS model, and several ANFIS models optimized by the grasshopper optimization algorithm (GOA), genetic algorithm (GA), differential evolution (DE), support vector machine (SVM) and particle swarm optimization (PSO) was evaluated based on 77 groups of sample data collected from the machining-inspection experiment. The empirical results indicate that the proposed IPSO-ANFIS model with an optimal root mean square error of 2.708 and determination coefficient of 0.9855 has a higher prediction ability than GOA-ANFIS, PSO-ANFIS, DE-ANFIS, GA-ANFIS, SVM-ANFIS, ANFIS and BP models, and appears to be an effective and reliable tool for estimating the number of MPs on complex curved surfaces.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Novel application of SMD-based IPSO-ANFIS model in predicting the number of measuring points on complex surfaces

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月21日

查看全文
  • Ruiping Luo,
  • Qihao Chen,
  • Gaocai Fu,
  • Shengwen Zhou,
  • Yingkang Lu

摘要

For the coordinated on-machine measurement of curved surfaces, the accurate prediction of the number of measuring points (MPs) could provide an important theoretical time basis for machining-inspection planning, workshop production scheduling, etc. To address the problem that the prediction of the required number of MPs for curved surfaces in relation to their complexity, machining accuracy, and other ambiguous characteristics, an adaptive neuro-fuzzy inference system (ANFIS) model optimized by the improved particle swarm optimization (IPSO) algorithm is proposed. Firstly, the input and output variables of the prediction model are defined as the surface measurement descriptor (SMD) by analyzing the impact indicators when measuring complex surfaces. Then, the IPSO algorithm with adaptive weights and particle acceleration updates is applied to optimize the ANFIS model with an organic composition of fuzzy logic knowledge and neuronal network. Finally, the prediction ability of the back propagation (BP) model, ANFIS model, and several ANFIS models optimized by the grasshopper optimization algorithm (GOA), genetic algorithm (GA), differential evolution (DE), support vector machine (SVM) and particle swarm optimization (PSO) was evaluated based on 77 groups of sample data collected from the machining-inspection experiment. The empirical results indicate that the proposed IPSO-ANFIS model with an optimal root mean square error of 2.708 and determination coefficient of 0.9855 has a higher prediction ability than GOA-ANFIS, PSO-ANFIS, DE-ANFIS, GA-ANFIS, SVM-ANFIS, ANFIS and BP models, and appears to be an effective and reliable tool for estimating the number of MPs on complex curved surfaces.