China Mechanical Engineering

Previous Articles     Next Articles

Research on Remanufacturing Cost Prediction Model of Used Parts Considering Failure Features

ZHANG Xugang1,2;AO Xiuyi1,2;ZHANG Hua1,2;JIANG Zhigang1,2   

  1. 1.Key Laboratory of Metallurgical Equipment and Control Technology,Ministry of Education,Wuhan University of Science and Technology,Wuhan,430081
    2.Hubei Key Laboratory of Mechanical Transmission and Manufacturing Engineering,Wuhan University of Science and Technology,Wuhan,430081
  • Online:2018-11-07 Published:2018-11-07

[机械装备再制造]考虑失效特征的废旧零部件再制造成本预测模型研究

张旭刚1,2;敖秀奕1,2;张华1,2;江志刚1,2   

  1. 1.武汉科技大学冶金装备及控制教育部重点实验室,武汉,430081
    2.武汉科技大学机械传动与制造工程湖北省重点实验室,武汉,430081
  • 基金资助:
    国家自然科学基金资助项目(51605347)

Abstract: The impacts of failure features of used parts on remanufacturing cost were analyzed, and a predictive model for the remanufacturing cost for used parts considering failure features was established. The model combined with semi-supervised learning and LS-SVR algorithm, realized the prediction of remanufacturing costs of used parts under the conditions of only a few completed remanufactured parts and a large number of unknown remanufacturing cost parts. The algorithm incorporated a k-nearest neighbor (kNN) learner, supplemented by kNN, and the LS-SVR was used to evaluate the confidence level of untagged samples, adding the best unlabeled samples from each phase stepwise to the labeled ones. The gradual update of the prediction model might effectively reduce noise and improve model accuracy. Case verifications show that the proposed algorithm has good ability of regression prediction and generalization.

Key words: failure feature, remanufacturing, cost prediction, semi-supervised learning, least squares support vector regression(LS-SVR)

摘要: 分析了废旧零部件失效特征对其再制造成本的影响,建立了一种基于失效特征的废旧零部件再制造成本预测模型,该模型将半监督学习与最小二乘支持向量机回归(LS-SVR)算法相结合,实现了在只有少量已完成再制造的废旧零部件样本和大量未知再制造成本的零部件样本的条件下,对废旧零件再制造成本的预测。该算法加入了k最近邻(kNN)算法,以kNN为辅、LS-SVR为主对未标记的样本进行置信度评估,将各阶段最优未标记样本逐步添加进有标记的样本集中,逐步更新预测模型,能够有效降低噪声,提高模型精度。经案例验证,提出的算法具有良好的回归预测能力和泛化能力。

关键词: 失效特征, 再制造, 成本预测, 半监督学习, 最小二乘支持向量机回归(LS-SVR)

CLC Number: