Research On Quality Assessment System Of Flash Welding Marine Engineering Mooring Anchor Chain

The mooring anchor chain is an important device of the ship,and its quality problem is related to the safety of the ship.Therefore,in recent years,researchers have been paying more and more attention to the quality assessment in the manufacture of mooring anchor chains.The flash welding mooring anchor chain has become the mainstream of the high quality anchor chain market in the shipbuilding industry at this stage.The traditional mooring anchor chain welding quality inspection can only be carried out by tensile test after the production,so that not only the quality of the chain can not be guaranteed,but also the repair of the fault chain requires a large cost.Therefore,the real-time quality assessment method of mooring anchor chain is crucial for timely detection and avoidance of faulty products and improvement of the quality of the chain.Firstly,a flash welding control system of mooring anchor chain based on industrial control computer and multi-function data acquisition card is developed,which can control the flash welding process mooring anchor chain in real time,control and manage welding process parameters,reduce the probability of failure chain,and collect and record sensor signals in real time during welding process.The real-time sensor signal acquired during the welding process is provided for quality assessment of flash welding mooring anchor chain.Secondly,a method for evaluating the quality of mooring chain flash welding based on improved dynamic time warping algorithm and multidimensional scaling method is proposed,which not only solves the problem of space-time asymmetry between the flash welding signal(electrode position and current signal and its combined two-dimensional signal),but also quantifies this difference,to make the mathematical description of the abnormal welding signal more accurate.According to the characteristics of the difference between normal and abnormal welding signals in the regular distance matrix,the multidimensional scaling analysis method is introduced to reduce the dimension and visualization of the regular distance matrix.Finally,the KNN classification algorithm is used to achieve quality assessment.Thirdly,for the problems of the KNN classification model,such as the poor fault tolerance of training data and the poor effect of unbalanced samples,the dirichlet process mixture model is proposed to classify the signals processed by the improved dynamic timewarping and multidimensional scaling.the dirichlet process mixture model is an unsupervised clustering.In the dirichlet process mixture model,there is no need to pre-set the number of clusters.Each cluster follows a multivariate Gaussian distribution.Each new observation is assigned to a cluster with the highest probability generated by a multivariate Gauss distribution.It is good enough to solve the shortcomings of the KNN classification algorithm for the classification accuracy and fault tolerance of the unbalanced samplesFinally,deep learning algorithms have received extensive attention in recent years and have made breakthroughs in various fields.A new recursive analysis method,namely the deep learning algorithm based on recursive graph,is proposed to further analyze the sensor signals in the collected welding process.Firstly,the recursive map of the electrode position and current signal with more welding quality information in the welding process is calculated,and then the recursive graph is directly studied by the convolutional neural network to achieve quality assessment.The experimental results show that the proposed method can effectively classify the welding quality and realize real-time quality assessment by converting the identification of faulty weldments to the analysis of sensor signals in the welding process.The accuracy indicates that the current signal is more important to determine the quality of the flash weld than the electrode position signal.Moreover,the regular distance matrix obtained by selecting a large step size dynamic time warping algorithm can better reveal the difference between normal and abnormal signals.The research methods presented in this paper show great potential and provide an effective solution for flash welding real-time quality assessment solutions.