论文部分内容阅读
提出了一种基于非线性核判别分析的悬跨管道状态识别方法。首先分析了悬跨管道状态识别问题的主要特点,根据悬跨管道状态变化频率结构的内部特征,构造由自然频率、归一化频率、频率变化率等特征参数组成的状态矢量。然后基于统计模式识别中的核判别分析原理,建立了悬跨管道的状态识别方法。最后通过实例分析研究了该方法的有效性,并讨论了悬跨管道状态矢量选取对识别效果的影响。研究结果表明:该方法能有效识别悬跨管道的多种状态,具有较强的抗测试误差能力。
A method based on non-linear kernel discriminant analysis is proposed to identify the condition of suspension pipelines. First of all, the main features of the problem of suspended pipeline status identification are analyzed. According to the internal characteristics of the frequency structure of the suspended pipeline, a state vector composed of characteristic parameters such as natural frequency, normalized frequency and frequency change rate is constructed. Then, based on the principle of nuclear discriminant analysis in statistical pattern recognition, a method of state recognition of suspended pipeline is established. Finally, the validity of this method is analyzed through an example, and the influence of vector selection of the suspended pipeline on the recognition effect is discussed. The results show that this method can effectively identify a variety of state of the suspended pipelines and has a strong ability of resistance to test errors.