Demand Forecast Of General Parts For Small Batch Customization Products Based On Virtual BOM

With the development of the production mode of small batch customization,the production mode of enterprises has gradually transformed into the mixed mode of make to order(MTO)and make to stock(MTS).Customized parts need to be determined according to the actual needs of customers,usually according to the order to arrange production;while general parts have certain versatility in the same or even multiple types of products.Due to the different structure of customized products,the demand and lead time of general parts in different orders are not the same,so it is unreliable to directly predict the historical data of general parts.In this paper,an indirect prediction method based on virtual BOM is proposed to solve the problem that it is difficult to predict the demand of general parts in small batch customization.First of all,this paper introduces the background and significance of the research,investigates and summarizes the requirements of general parts of small batch customization products,then puts forward a construction method of virtual products and BOM corresponding to small batch customization products,compares and analyzes the characteristics of general parts corresponding to virtual products,and lists several general parts categories with different characteristics.Then,based on the direct prediction of historical data and the indirect prediction of virtual product,the prediction models of common parts are established respectively.The comparison of the direct prediction method and indirect prediction method under different demand and lead time settings under a single virtual product,and the comparison of the prediction results between direct prediction and indirect prediction verify that the indirect prediction method.The applicability of general parts demand forecast in small batch customization production.On this basis,further analysis of the actual production of common parts,found that the demand trend and randomness of common parts among different virtual products will be more complex.Comparing the direct prediction with the indirect prediction,the indirect prediction has a better prediction effect than the direct prediction,which fully shows that the indirect prediction has a wide range of applicability.Finally,considering that the single prediction solution method has its own characteristics,can not avoid its inherent defects,and reflects the different prediction accuracy,so the combined prediction method based on indirect demand prediction is used to predict the demand of general parts.For this reason,combined with indirect demand prediction of general components,the prediction of three methods,adaptive differential moving average(ARIMA),multi-layer perceptron(MLP)and cyclic neural network(RNN),are compared and analyzed.Then,based on the combined forecasting model,the three methods are combined to build a feasible demand forecasting model of general components,and the model is analyzed and verified feasibility and practicability.