The Study On The Prediction Of Aluminium Electrolysis Anode Effect Based On The Signal Analysis

Anode effect is a common phenomenon in the process of aluminum electrolysis. Its occurrence has a serious impact on the whole series of electrolysis of aluminum, including influencing each technical specification of electrolysis, reducing the quantity and quality of aluminum, destroying the stationary power supply of the whole system. During the running of aluminum reduction cell, various physical parameters about running status will present certain rules as time changes, such as kinds of electrical signals, alumina density, temperature of aluminum reduction cell, air bubble vibration, and so on. These signals always contain important information which are helpful to do the equipment condition recognition and diagnosis of the devices. Analyzing and processing these information effectively, as well as establishing relation between these information and aluminum electrolysis running status are the foundation of fault diagnosis of aluminum reduction cell.In aluminum electrolysis production process, characteristic sound is followed when electrolyte and aluminum liquid circularly flow in aluminum reduction cell, contact surface undulates, anode gas discharges and anode effect occurs. By detecting and analyzing these characteristic sound signals, the running status of aluminum reduction cell can be judged. Xue and Oye have proposed one anode effect detecting method, which surveys the vibration of anode leader to obtain the air bubble state around the anode in the electrolyte, in order to detect the anode effect. However, there is great difference of the aluminum electrolysis producing environment, which adds the difficulties of analyzing signals. This thesis seeks for appropriate signal processing methods to achieve the anode effect locating detection, so as to be prepared in theory method and technology for the locating detection and the forecast of anode effect in the industry production.Wavelet packet and spectral estimation is combined together to extract frequency characteristics of specific frequency bands of signals, in order to detect and locate the anode effect. Firstly, when anode effect occurs in aluminum electrolysis, wavelet packet is adopted to decompose anode leader vibration signals to multilayer and reconstruct the signals. Secondly, bispectrum estimation method is adopted to extract the characteristics of anode vibration signals. This method detects and forecasts the anode effect, so as to control it before occurrence. The forecast of aluminum electrolysis anode effect is helpful to carry on the production smoothly and reduce the production cost.