基于SVDD的顶锤裂纹故障预测方法研究

发布时间：2018-05-31 07:23

本文选题：顶锤 + 故障诊断　；参考：《北京邮电大学》2015年硕士论文

【摘要】：金刚石生产过程中,压机顶锤承受交变的高温高压,容易发生断裂,如果裂锤继续使用,可能会出现塌锤事故,导致同一压机的另外五个顶锤报废,对金刚石生产压机的安全稳定运行有着极大的危害。因此,对压机顶锤断裂进行及时有效地识别和诊断对于保证金刚石生产的安全有着重大意义。本文以支持向量数据描述为基础,对金刚石顶锤裂纹故障诊断展开以下研究： 1)针对采集到的顶锤裂纹信号背景噪声大的问题,通过对比分析顶锤故障信号与背景噪声的能量分布特点,运用高通滤波器去除背景噪声,并根据信号的能量利用阈值将故障信号准确提取。最后通过实验验证了该方法的有效性。 2)通过大量实验发现,传统参数分析方法无法准确地区分顶锤故障和正常信号。因此本文通过分析,提取了能够表征故障脉冲的过零率、线性倒谱和功率谱密度三种特征参数,实现了两者的有效区分。经实验验证,该方法可准确识别顶锤故障脉冲。 3)通过分析故障识别的限制因素,本文引入特征参数优化方法。将连续特征数字量化,然后利用信息增益进行特征选择。通过实验验证,在建立分类器之前,对特征参数进行优选,使得利用特征参数构建的分类模型更加精确。 4)随着数据的不断采集更新,新增数据集将极大丰富训练集,并且新增数据蕴含的信息和知识具有很大的潜在价值。因此,本文引入增量学习算法,并通过实验验证增量学习能缩短分类时间,提高分类效率。
[Abstract]:In the process of diamond production, the top hammer of the press is subjected to alternating high temperature and high pressure, so it is easy to break. If the hammer continues to be used, there may be a collapse hammer accident, resulting in the other five top hammers of the same press being scrapped. It has great harm to the safe and stable operation of diamond production press. Therefore, it is of great significance to identify and diagnose the breakage of press head hammer in time and effectively to ensure the safety of diamond production. In this paper, based on the support vector data description, the following research on crack diagnosis of diamond top hammer is carried out: 1) aiming at the problem that the background noise of the crack signal of the top hammer is large, the characteristics of energy distribution between the fault signal and the background noise of the top hammer are compared and analyzed, and the background noise is removed by using the high-pass filter. The fault signal is extracted accurately according to the energy threshold of the signal. Finally, the effectiveness of the method is verified by experiments. 2) through a large number of experiments, it is found that the traditional parameter analysis method can not accurately distinguish the malfunction of the top hammer from the normal signal. Therefore, through analysis, three characteristic parameters, which can represent the zero-crossing rate of fault pulse, linear cepstrum and power spectral density, are extracted, and the effective distinction between the two parameters is realized. The experimental results show that the method can accurately identify the top hammer fault pulse. 3) by analyzing the limiting factors of fault identification, the method of feature parameter optimization is introduced in this paper. The continuous feature number is quantized and the information gain is used for feature selection. The experimental results show that the feature parameters are optimized before the classifier is established, which makes the classification model constructed by the feature parameters more accurate. 4) with the continuous data collection and updating, the new data set will greatly enrich the training set, and the information and knowledge contained in the new data has great potential value. Therefore, the incremental learning algorithm is introduced in this paper, and experimental results show that incremental learning can shorten the classification time and improve the classification efficiency.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TQ163

【参考文献】