基于声信号的铝锭脱模故障诊断研究

发布时间：2018-03-17 17:49

本文选题：铝锭铸模　切入点：去噪方法　出处：《兰州理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：电解法生产出的高温铝液需经过铝锭连铸生产线铸造成型,铝锭铸模是承载这些高温铝液的容器,高温铝液浇铸到铸模后由链条传送,铝液冷却凝固成固体铝锭,通过敲击设备敲击铸模,实现铝锭脱模。铸模运行环境恶劣,各种原因导致经常出现铝锭无法正常脱模,必须进行检测并控制装备对没有脱模的铝锭铸模进行处理。由于铝锭工作在高温环境,且铸锭重量高、惯性大,常见的机械式或者光电式传感器都存在着一定的局限性,而铸模敲击声音信号中蕴含着丰富的铸模状态信息,基于声信号的故障检测技术具有高可靠性、非接触性、在线工作等优势,有良好的应用前景,因此,论文提出了利用铸模敲击声音进行铝锭脱模故障诊断方法。现场采集的铸模敲击声音信号中含有大量复杂的背景噪声,严重影响着故障识别率,去噪是该故障诊断方法中的关键一步。针对铝锭铸模声音故障诊断的去噪问题,提出一种改进的小波包多级阈值去噪方法。首先,选择同种小波基和分解层数,对铸模声音信号和噪声信号分别进行小波包分解;其次,将不同频率段的噪声信号标准差引入新的阈值函数,形成自适应多级阈值,对相应频率段的铸模声音信号进行降噪处理;最后,通过小波包逆变换对信号进行重构。通过对三种小波包去噪方法的均方根误差和信噪比的对比,验证了本文阈值函数的优越性。现场的去噪实验证明了本文方法对铝锭铸模声音去噪的有效性。在对去噪后的声音信号频域分析后,发现一次敲击后如果铝锭脱模,则一次敲击声音信号和二次敲击声音信号存在明显的频率峰值差,将其作为故障特征参数,并设定阈值,对铝锭是否脱模进行故障诊断,对现场采集的声音处理实验证明了本文所提方法的有效性。
[Abstract]:The high temperature aluminum liquid produced by the electrolysis method needs to be cast and formed through the aluminum ingot continuous casting line. The aluminum ingot casting mould is the container carrying the high temperature aluminum liquid. The high temperature aluminum liquid is transferred by chain after casting to the mould, and the liquid aluminum is cooled and solidified into the solid aluminum ingot. Aluminum ingot can be demoulded by percussion equipment. The mold runs in a bad environment, and various reasons often lead to aluminum ingot can not be demoulded normally. It is necessary to test and control the equipment to deal with the casting mould of aluminum ingot without removing mould. Because the aluminum ingot works in high temperature environment, and the ingot has high weight and inertia, the common mechanical or optoelectronic sensors have some limitations. However, the sound signal of mould percussion contains abundant state information of casting mold. The fault detection technology based on acoustic signal has the advantages of high reliability, non-contact and on-line operation, so it has a good application prospect. In this paper, a fault diagnosis method for aluminum ingots is put forward by using mould percussion sound. There are a lot of complex background noise in the sound signal of mould tapping, which seriously affects the fault identification rate. Denoising is a key step in this fault diagnosis method. An improved wavelet packet multilevel threshold de-noising method is proposed for the sound fault diagnosis of aluminum ingot casting mould. Firstly, the same wavelet basis and decomposition layer number are selected. The sound signal and noise signal of casting die are decomposed by wavelet packet. Secondly, the standard deviation of noise signal in different frequency band is introduced into the new threshold function to form adaptive multilevel threshold. Finally, the signal is reconstructed by inverse wavelet packet transform. The root mean square error and signal-to-noise ratio (SNR) of the three wavelet packet denoising methods are compared. The superiority of the threshold function is verified. The field experiment proves the effectiveness of this method for the sound denoising of aluminum ingot casting mould. After the analysis of the frequency domain of the de-noised sound signal, it is found that if the aluminum ingot is demoulded after one stroke, There is obvious frequency peak difference between the first knock sound signal and the second knock sound signal, which is regarded as the fault characteristic parameter and set the threshold to diagnose whether the aluminum ingot is demoulded or not. The experimental results of sound processing in the field show that the proposed method is effective.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TG292

【相似文献】