Research On Machine Identification Technology For Potential Process Failure Modes

Systematic,accurate and efficient identification of potential process failure modes is the basis and prerequisite for process failure analysis.In mass stable production mode,failure analysts can identify process failure modes by brainstorming based on previous failure data and knowledge experience.For the small batch customization production mode,the failure data and experience can be used for reference are less.and the traditional failure modes identification method is no longer applicable.To solve this problem,some scholars have proposed a method of identifying potential process failure modes based on the general model of process elements,which avoids the influence of production modes and has universal applicability.However,the elements of process are complex,especially for some complex process,the identification of process failure modes only depends on human resources,which has some shortcomings,such as slow identification speed and unstable identification quality.Machine identification has the advantages of high identification efficiency and stable identification process,which can effectively improve the efficiency and quality of process failure modes identification.To this end,the paper systematically carries out the following related work:Firstly,the process theory is applied to analyze the simple process,describing the one-to-one relationship between the steps and its elements,and BP neural network is used to fit the relationship to establish the generation model of the elements of the simple process.Secondly,the process theory is used to further analyze the complex process,describing the one-to-many relationship between process and step,step and its elements,and introducing the concept of content planning in the classical pipeline theory of natural language generation,guiding the training of seq2seq model,and establishing the step and its elements generation model based on planning.Then,the semantic relationship between step elementts and 13 failure criteria is analyzed,and a machine identification model of process failure modes is established,which integrates semantic correlation.Finally,an application analysis is performed.The application results of the simple process element generation model show that the accuracy of the results of each element is above 87%,and it is pointed out that the increase of the word vector dimension makes the relative error rate of the result decrease first and then increase;The steps and its element generation model and the process failure mode machine identification model are applied to the process failure mode identification of the rudder steering assembly and the active hatch assembly of a certain aircraft.The results show that the effect of normalized element generation is better than that of non-normalized element generation.When the learning rate is 0.01,the increase of training times also has a effect show a trend from positive to negative.At the same time,the failure mode recognition effect is relatively good for the existing elements in the training set.This paper combines natural language processing technology with intelligent algorithms such as neural network,and proposes a machine identification technology with process failure modes.The preliminary exploratory research on the technology laid the foundation for the further realization of the process failure mode machine identification and the improvement of identification accuracy and quality.