The Research And Application Of Low-Carbon Anomaly Detection Method Based On Hidden Markov Model

Both low energy efficiency and environment pollution are very serious in manufacturingindustry, to resolve these problems, in this paper we research the technology of anomalydetection for the purpose of low carbon development, and make the anomaly detectiontechnology applied to industrial production. The manufacturing industry is a high energyconsumption industry and Energy consumption is mainly used in equipment production.Modern manufacturing equipment is becoming more and more complex, sophisticated andintelligent, the impact factors for normal operation of the equipment will be increased, in thissituation we need to find a way of production and this kind of production with the purpose ofthe minimum energy consumption and the maximum profit and also ensure the performanceand quality of machinery manufacturing product. In this paper we research the process ofmanufacture tires, study on the high energy consumption process of the production, extractthe factors that have a great relationship with energy consumption to detect, use the detectionof energy consumption data to judge the abnormal reasons for production process andimprove the process to ensure the manufacturing process smoothly and to reduce energywaste and saving manpower and resources. Because of these factors data has a certain timesequence and we couldn’t see the inner state of the device to find out why theexception occurs, in this paper the HMM classification algorithm is proposed, the algorithm isa statistical classification model based on finite state, this model can carry out classificationon the data that have the spatial and temporal characteristics.As the data collection in the industrial production and manufacturing process, in thispaper we use HMM for the modeling of Vulcanization process which is the biggest energyconsumption process, with the parameter temperature, pressure and time as the hidden state ofHMM, and the possible fault as the observation data of HMM. For every possible faultestablish the corresponding model respectively, then when the new observed data arrives,input the data to the models, and the model has the maximum similarity is the stateof the vulcanizing machine based on the the theory of probability. When the test results wereabnormal, make corrections according to the reason.The innovation of this paper lies in the application of anomaly detection technology inthe production process to achieve the purpose of energy saving, controlling the workingcondition of the equipment to ensure the effective use of energy, and putting forward the improvement based on K-means algorithm for the method initializing according to theinadequacies of the model. According to the experimental results, the improved modelhas better stability.