Investigation Of Aluminum Ingot Demoulding Fault Diagnosis Based On Acoustic Signal Analysis

Most of the aluminum is produced by the method of electrolysis. Aluminum ingot is forged by Aluminum Continuous Casting Production Line with high temperature of liquid aluminum. Aluminum ingot mold is a container loading the high temperature liquid aluminum. After high temperature liquid aluminum is poured to the mold and transmitted by chain. When aluminum liquid is solidified into solid aluminum ingot,demoulding working device knocks mould and aluminum ingot is demoulded. The condition of mold running is bad and all sorts of reasons lead to aluminum ingot is not demoulded. It is very important to detect whether aluminum ingot is demoulded. If aluminum ingot is not demoulded, control equipment must deal with that mold. The weight of aluminum ingot is high, inertia is big and the temperature is high in the working environment. The commonly used detection methods are mechanical sensors and photoelectric sensors, but there are some limitations. Mould tapping acoustic signal contains its abundant state information. Acoustic fault diagnosis methods has advantages of high reliability, non-contact, on-line operation and so on, so it has a good application prospect. So, the proposed fault diagnosis method uses mould tapping acoustic signal to diagnose.The mould tapping acoustic signal collected at the scene contains a lot of complicated background noise. Denoising is a key step in fault diagnosis method,because noise seriously affects the fault recognition rate. Since the problem of de-noising is difficult in fault diagnosis of aluminum ingot mold sound, an improved wavelet packet multi-threshold de-noising method was proposed. Firstly, mold signal and noise signal are decomposed with same wavelet basis and decomposed layer;Secondly, the standard deviation of the noise signal in different frequency is introduced to new threshold function and came into being adaptive multi-threshold.Mold signal in corresponding frequency can be disposed; Finally, de-noising signals are reconstructed with the inverse transformation of wavelet packet. The root mean square error and signal to noise ratio of three wavelet packet de-noising methods are contrasted and the superiority of this threshold function is verified. The effectiveness for new method is demonstrated by the scene of the de-noising experiments. If aluminum ingots have demoulding after first hit, there is obvious pitch frequency difference to the first percussion sound and the second knock sound. Pitch frequencydifference as the fault feature parameter and setting a threshold value. It can be diagnosed whether aluminum ingots demould. The effectiveness for new method is demonstrated by the acoustic processing experiments.