Study On Spot Welding Quality Judgment Based On Hidden Markov Model(HMM)

Spot welding is widely applied in metal plate connection process,thanks to its high efficiency,low cost and high automation.Therefore,ensuring the qualification rate of spot welding manifests a great social and economic significance for safety of industrial production and saving of production cost.The quality of spot welding,however,can only be monitored indirectly by its closely related electrical and physical parameters,which results from the non-observability and instantaneity of the nugget formation process.The relationship between process parameters and spot welding quality can hardly be described by accurate mathematical models,which is attributed to the high nonlinearity and multi-parameter coupling in spot welding process.Consequently,how to comprehensively analyze the influencing factors of spot welding quality and establish a mapping relationship between the factors and the quality is a research hotspot in this field.Based on these,this paper starts with the factors that affect the spot welding quality,and studies the mapping relationship between the factors and the quality,which provides further technical support for monitoring spot welding quality by establishing timing models.The main research work of this paper is as follows:(1)Analyzing the factors that affect spot welding quality,and constructing a mapping relationship model between the factors and the quality.The factors that affect spot welding quality are analyzed first,when main factors are summarized as the process parameters and the state changes of spot welder during process parameters execution.The monitoring waveform curves are obtained under different states of welding spot quality through experiments.The analysis shows that,the method for judging spot welding quality based on waveform curves cannot deeply analyze the quality difference between welding spots,so this paper proposes using HMM to analyze the discrepancy of welding spot quality under different states against timing.The off-line data collected are further used to build a mapping relationship model between welding spot quality and the variation law with time for signals including welding current,interelectrode voltage and electrode pressure,when the spot welder is executing process parameters,based on improved genetic algorithm(IGA)and HMM,where IGA is employed to optimize the initial model of HMM,while HMM is used as the classifier.(2)Using the timing model,and judging the processing quality based on the real-time waveform curves of welding current,interelectrode voltage,and electrode pressure when the spot welder is executing the spot welding task.The real-time trait of spot welding quality change is analyzed first,which is followed by demonstrating the defect when traditional EM algorithm is applied to the real-time judging process of spot welding quality,that is,in the iterative process,the algorithm searches the same step size along the negative gradient direction,and cannot converge to the global optimal solution at a fast speed.A method for judging the real-time information of spot welding quality based on Aitken-IEM and HMM,where Aitken-IEM is used for improving the training speed of HMM model parameters.Finally,the real-time data collected are used to verify the feasibility of the proposed algorithm by comparative analysis.(3)Monitoring whether the spot welder executes the processing technology optimally when executing spot welding tasks,through constructing timing models.The method for controlling spot welding quality is analyzed first,then it is demonstrated on how the state change impacts welding quality when the spot welder is executing process parameters.Since a algorithm needs to re-estimate model parameters when a traditional HMM incremental learning algorithm observes sequence updates,a recognition method of spot welder based on improved HMM incremental learning is proposed,by using the collected real-time data and based on the above method for constructing timing models.The effectiveness of the proposed algorithm is verified through comparative analysis.The state recognition results are applied to the spot welding quality monitoring process,and the sampling signal is recovered according to Shannon sampling theorem.The quality of spot welding is judged based on the judgment method of spot welding quality in the last chapter,and the technical support for improving the qualified rate of welding spot.