论文部分内容阅读
针对水下机器人的特性和传统的主元分析方法中主元个数选取方法的不足,提出了一种基于平均特征值的累计贡献率法来计算主元得分,以降低累计方差贡献率法的主观性.通过对水下机器人系统中变量的协方差矩阵进行特征值分解,推导了基于主元分析的水下机器人故障检测和故障识别的具体方法.考虑到传统的主元分析法进行数据重构时可能夹带估计误差,提出了一种迭代的故障传感器数据重构方法,以减小估计误差.“海狸”号水下机器人的实验结果验证了该方法的可行性和有效性.
In view of the characteristics of underwater robots and the traditional method of principal component analysis, this paper proposes a method of calculating the principal component score based on the cumulative contribution rate method of average eigenvalue to reduce the cumulative variance contribution rate method Subjectivity.Based on the eigenvalue decomposition of the covariance matrix of variables in underwater robotic system, a detailed method of underwater robot fault detection and fault identification based on principal component analysis is deduced.According to the traditional principal component analysis An estimation error may be entrained in the structure, and an iterative reconstruction method of fault sensor data is proposed to reduce the estimation error. Experimental results of Beaver ’s underwater robot verify the feasibility and effectiveness of this method.