Prediction Of Queuing Time Of Raw Material Transportation Vehicles In Steel Logistics Scenarios

As an important part of steel logistics,the transportation of raw materials is responsible for transporting raw materials such as scrap steel and coal to steel production enterprises for use.Due to production efficiency,warehouse capacity and other reasons,the vehicles transporting raw materials are scheduled to queue when they are near the steel plant and wait for notification to enter the plant to unload.At present,the steel logistics field is in the early stage of information transformation.The lack of data and complex business logic make it difficult to estimate the waiting time for vehicles,which greatly reduces the driver's service experience.Due to the long waiting time,the drivers missed the order and did not enter the factory in time.This not only affects subsequent vehicles en-tering the plant and increases the management difficulty of the steel plant,but also affects the production efficiency of the steel plant.Currently,most steel mills estimate the time based on the vehicle queuing number,which has a low reference value.At the same time,due to the particularity of steel logistics,there are very few studies involving the queuing time prediction problem in this scenario.In summary,facing the steel logistics scene,it is urgent to design a method to accurately predict the queue time in order to optimize the raw material transportation.Different from the traditional time prediction problem,the queuing scene of raw material vehicles under steel logistics is complex.Effective feature generation,applicable prediction models,appropriate data selection and model update have become the challenges of this research.In order to solve the above problems,this paper designs a three-stage raw material vehicle queue time prediction framework.First of all,combine the scene characteristics of the raw material transportation business to deeply explore the characteristics that affect the queuing time of vehicles,as the basis for constructing the prediction model.Secondly,according to whether raw materials are frequently transported,different time prediction methods are used,and a high-quality training set is constructed through data selection to improve the accuracy of prediction.Finally,considering that the phenomenon of concept drift often occurs in the actual environment,in order to alleviate the impact of data changes,the model is updated in a timely manner by monitoring data distribution and changes in model performance.The main work of this paper is as follows:· Queuing scene introduction and data analysis This paper introduces the queuing business of raw material transportation vehicles.It combines the characteristics of the steel logistics scene and integrates the data related to the queuing business to dig out important information.Finally,the feature engineering method is used to generate features as the basis of subsequent research work.· Queuing time prediction model Aiming at the characteristics of raw material vehicle queuing,this paper proposes a combined model based on LSTM and linear model.LSTM mines information from the sequence data of the most recently notified vehicle as the prior knowledge of the current prediction.The linear model combines the prior knowledge and the real-time information of the vehicle to make predictions.The combined model utilizes each part of the data through suitable models to improve the accuracy of the prediction results.· Data selection and model update Based on the principle of locality,this paper divides time series data into windows with different time spans,and clusters data in the small window through a clustering algorithm,aiming to classify these data into multiple production status categories.The data in the large window is classified according to the clustering results.It calculates the weights based on the classification result,and then uses the Ares algorithm for selection.Finally,a training set is constructed to improve the prediction accuracy of the model.After the model is online,monitor its performance and data distribution,and update the model in time to adapt to changes in the scene.In summary,based on the characteristics of the steel logistics scene,this paper designs a framework for predicting the queuing time of raw material transportation vehicles,and proposes corresponding models and algorithms.A large number of experiments on the real production data of Rizhao Steel Group support the effectiveness and efficiency of the method in this paper.