Abstract:As the key equipment for the production of aluminum, the stability and reliability of aluminum extrusion machine is fundamental for normal production. Traditional equipment anomaly detection method belongs to intrusion detection, which requires other instrument to be embedded in production equipment for analysis, with the disadvantages of highcost and low adaptivecapacity. As a new nonintrusion detection method, energy consumption anomaly detection can reflect operating conditions of equipment and find out anomaly situation timely. Therefore, support vector regression (SVR) and genetic algorithm (GA) were combined to provide an anomaly detection model. Through analyzing the key factors of extrusion energy consumption, a GA-SVR energy consumption prediction model was built, whose input was the key energy consumption factor and output was power consumption. Then, by considering the uncertainty of the anomaly point, a confidence interval of energy consumption was founded based on GA-SVR model and the confidence interval was seen as energy consumption anomaly interval. At last, an experiment was accomplished to verify the effectiveness of the model for SY1000Ton extruder. The results show that the model mentioned above can be used to find out anomaly accurately when confidence coefficient is more than 97%. So, it can be used to ensure the steady production of extruder.