,

1)Research Institute,Baoshan Iron & Steel Co.,Ltd.,Shanghai 201999,China 2)Products & Technique Management Department,Baoshan Iron & Steel Co.,Ltd.,Shanghai 201900,China

Abstract: To solve the problems in the quality control and improvement of coiled tubing steel strips production,such as scattered and inefficient production data,difficult performance fluctuation factor analysis,complex multivariate statistical analysis,and low accuracy and difficulty in mechanical property prediction,an industrial data analysis platform for coiled tubing steel strips production has been preliminarily developed.As the premise and foundation of analysis,industrial data collection,storage,and utilization are realized by using multiple big data technologies.With Django as the agile development framework,data visualization and comprehensive analyses are achieved.The platform has functions including overview survey,stability analysis,comprehensive analysis (such as exploratory data analysis,correlation analysis,and multivariate statistics),precise steel strength prediction,and skin-passing process recommendation.The platform is helpful for production overviewing and prompt responding,laying a foundation for an in-depth understanding of product characteristics and improving product performance stability.

Key words: coiled tubing steel strips; industrial big data; data analysis platform; prediction

1 Introduction

Steel production is one of the industries with the longest process and the highest complexity;the amount of annual data precipitated by a steel plant can reach several trillion bytes.The manufacturing of steel products has relied on experience while nor-mally neglecting these enormous data.With the accel-erating increase in demand for quality and high-performance steel products,the traditional straight-forward technology and management,depending more on experience,are hardly competent[1-2].Instead,the steel industry is embracing fast-developing information technologies (such as big data and programming) to break its bottleneck in the traditional economy[3].The steel industry can be not only a source of big data but also a typical area of big data applications[4],through which the efficient collection,storage,and use of industrial data can be realized,and data visualization and comprehensive analysis can be fully utilized,realizing more accurate monitoring,analysis,and prediction of production control,with great potential and advantage for promoting precise decision-making and improving product quality.

POSCO has supposedly taken the lead in the practice of digitization and intellectualization of steel production since 2015.It has developed Pos-Frame[5],a smart platform for the continuous manufacturing process,specifically for thick plate production at Gwangyang Steelworks,which can not only analyze the industrial data but also enable the development of automation models.Simultaneous-ly,the Baosteel 5 m heavy plate production line performed digital and intelligent manufacturing,building a factory-level data center based on the in-frastructure and data resources of the original four-level control system[6],and developed the process intelligence and data application system to enhance process control ability and help improve the quality of thick plate products[7].The abovementioned practices are grand architecture systems for heavy plate production with the integration of traditional industry and the technologies of the internet of things,big data,and artificial intelligence,providing many functions and permeating nearly all aspects of mass production,making a case for the steel industry to transition to intelligent manufacturing.

The coiled tubing steel strip is a hot-rolled strip with the characteristics of high strength,thin thick-ness,and high sensitivity to the fluctuations in com-position,process,and even ambient temperature,therefore the information must be tracked during the stages of production to master the stability and ad-just operational parameters promptly.Moreover,the integration and utilization efficiency of production data is normally low,and the conventional process of data collection and statistical analysis is time-consuming and laborious.With the help of big data technology,the data analysis platform for coiled tubing steel strips is preliminarily researched and developed,focusing on the relatively simple busi-ness of data analysis and process improvement for the performance of coiled tubing steel strips.This platform can realize the tracking and monitoring of the entire manufacturing process,effectively dis-pensing tedious manual data analytical processes and making a timely forecast of mechanical pro-perties,laying the foundation for production over-view mastery and providing more valuable insights into the product characteristics,thereby guaranteeing stable production and improving the quality to a certain extent.

2 Platform design

2.1 Overall design

The data analysis platform is designed for data visualization and comprehensive analysis application of production data of coiled tubing steel strips.To solve the problems of quality management and improvement of coiled tubing steel strips,such as uncertainty of performance,difficulty in the analysis of the influence factors of strength fluctuation,and difficulty in predicting the property trends,the first premise is to identify the functional requirement and determine the main functions and interface design based on the demand of manufacturing managers and product engineers.The manufacturing managers hope to be informed of practical production over-view,mechanical properties distribution,time series,and so on,focusing on the characteristics of con-venience,intuitiveness,accuracy,and practicability.Product engineers are more concerned with variable relationships,abnormality analysis,and mathema-tical models,paying special attention to the data analysis method and functional integrity.

According to the actual need of users,the data is divided into composition,process,performance,specification,and other information;the interface design is mainly divided into a guide bar,filter zone,and result display zone;the analysis objects are divided into the semi-manufactured and finished strips,including the main functions of overall infor-mation,statistical analysis,and predictive model-ing.

2.1.1 Main architecture

This platform mainly consists of four layers,namely a function layer,kernel layer,data layer,and infrastructure layer,as shown in Fig.1.The infras-tructure layer contains a web server,data base,and data processing script.The data layer generally includes smelting,rolling,performance,order,and other information and involves data acquisition and operation.The kernel layer consists of a web frame-work,mathematical model,and data synchroniza-tion.In addition,the function layer focuses on data visualization and analysis,including descriptive statistics,exploratory data analysis,correlation and multivariate analysis,a customizable pivot table,per-formance prediction and process recommendation,and data export.The platform covers aspects of pro-duction,processing,and export.Through three modules of basic information,comprehensive analysis,and multiple regression analysis,the data and information are organized and displayed effectively to provide ser-vice for manufacturing managers and product engineers.

Fig.1 Architecture chart of the data analysis platform for coiled tubing steel strips

2.1.2 Key technology

The obtained industry data has the characteristics of large capacity,diversity,high redundancy,and so on.They are filtered and stored into an established local database using PostgreSQL,which is an object-relational data base management system (ORDBMS).In addition,to achieve high speed and high performance of the overall data query and analysis,and to ensure concurrency performance of multiuser queries simultaneously and stable user experience of interactive operation,the Django framework is used to develop the platform,adhering to the model-view-controller framework,the core software design of which consists of data access logic,business logic,and presentation logic,laying the foundation to pursue sustainable progress based on its stable and mature framework into the medium and long term.

2.2 Implementation

According to the data and technical requirements,the first premise was to construct a local database,through which the high-value data of the entire process was integrated to realize the systematic management and efficient use of the data.Then,the web framework was built and connected to the local database to realize the basic functions of data statis-tics and visualization,including overall descriptive statistics,a centralized profile interface,key index tracking,an abnormal information summary,and a customizable pivot table.Then,by integrating diversified data analytical approaches,a compre-hensive analysis and application were realized with a series of functions,such as correlation analysis,exploratory data analysis,multivariable analysis,and user-defined data export.In addition,using mathe-matical analysis,the performance forecasting model was established to achieve performance prediction and intelligent process recommendation.Finally,through the server construction,network deploy-ment,and system optimization,the normal access to the platform and the corresponding accounts management with classified permissions were realized.

3 Data analysis platform for coiled tubing steel strips

3.1 Functional module division

The data analysis platform for coiled tubing steel strips adopts the open-source web framework and constructs the local database,which realizes effi-cient data storage and use,further laying the foundation for data analysis and modeling.On the basis of the data analysis platform,three modules of basic information,comprehensive analysis,and multiple regression analysis are constructed,which realize the data visualization and comprehensive analysis application of big data in coiled tubing steel strips production.

3.1.1 Basic information module

The basic information module consists of five parts:overall descriptive statistics,a centralized pro-file interface,key index tracking,an abnormal information summary,and a customizable pivot table.On the one hand,an overview of the con-cerned data distribution and a centralized display of selected samples are available,which dispenses with the complicated manual process of data preparation and improves the analysis efficiency.On the other hand,key performance indices or other information can be tracked on different time scales,and abnor-mal information of production data is automatically summarized according to the set rules,which helps achieve accurate production stability monitoring and fast abnormality identification,thereby accelerating the decision-making and response.

3.1.2 Comprehensive analysis module

The comprehensive analysis module covers the function of correlation and multivariate analysis,which is useful for indicating a predictive relation-ship that can be exploited in practice.This module consists of exploratory data analysis,correlation analysis,and multivariate statistics for more than one outcome variable,which is significant for pro-viding initial exploratory information of data and determining key influencing factors,further explor-ing the composition and process characteristics of products to formulate an objective and reasonable optimization strategy.

3.1.3 Multiple regression analysis module

Batch performance stability has always been a diffi-culty in the production of high-strength steel strips.It is sensitive to changes in the composition and process of production,raising the necessity of law analysis of relevant variables.On the basis of big data and the classical regression method,this multiple regression analysis module establishes prediction models of performance.The models can be used to predict strength according to the hot-rolled coil information.This module can not only help anticipate and respond to changes in production but also guide subsequent process adjustment,which is significant for short-term treatment and long-term optimization of the process.

3.2 Function module application

3.2.1 Production overview and stability analysis

Process and product performance stability have always been the focus of industrialized production.Being timely informed of the manufacturing situation is of great significance for improving the product qualification rate and formulating counter-measures and adjustment measures.The factors affe-cting the performance of coiled tubing steel strips are of large complexity,including the chemical com-position,rolling and cooling process,and ambient temperature,leading to unavoidable performance fluctuations.Given this scenario,the data analysis platform has realized the overview survey and stability analysis of industrial data of coiled tubing steel strips.Fig.2 shows the results of descriptive statistics and a frequency distribution histogram of the main variables under specific filter criteria.Fig.3 illustrates the monthly boxplot series of key variables for a specific period,collects the items where the performance results are beyond the con-trol limits,and provides a customizable pivot table according to the target field.An index boxplot series in different time dimensions is an intuitive way to understand the overall situation and the changing trend.A summary of abnormal information can help quickly and accurately identify the steel strips with performance problems,thereby expediting the pro-cess of analysis and response.In addition,the pivot table is an interactive way to summarize,analyze,explore,and present summary results of a large amount of data,allowing users to easily see pat-terns,trends,and comparisons.

Fig.2 Descriptive statistics and histogram charts

Fig.3 Stability and perspective analysis

3.2.2 Correlation and multivariate analysis

Data analysis includes univariate characteristic analysis,bivariate correlation analysis,and multi-variate statistical analysis,which can help clarify the key variables and explain the interaction among the variables to obtain a deep understanding of the pro-cess and performance characteristics of the products.Although many factors play an important part in the properties of coiled tubing steel strips,some of them are normally interference or irrelevant terms during a specific period.Through exploratory analysis,the structure and law of original data can be explored using mapping,tabulation,and so on,under as few a priori assumptions as possible,and the relationship among concerned variables can be further displayed with a correlation matrix plot,as shown in Fig.4.Fig.5 illustrates the statistical chart with multivariate variables and characteristics of dynamic adjustment,through which the complex interrelationship is shown in graphs.The customizable and interactive analysis chart facilitates the optimization and fine control of the production process.

Fig.4 Correlation analysis

Fig.5 Multivariate analysis

3.2.3 Performance forecast and process recommend-ation

Skin-pass rolling is a process in coiled tubing steel strip production.The main objective in skin-pass rolling is not only to obtain a certain surface roughness profile and better shape but also to improve the performance stability to a certain extent.For steel strip production using the skin-passing process,the strength property can be adjusted by the process of skin passing and relative to the composition,thickness,rolling and cooling temperatures,and the rolling force of skin passing.Appropriate skin-pass rolling force implementation can improve the performance stability along the width of the strip and,to a certain extent,the perfor-mance difference between different strips.To ach-ieve precise control,the influence of composition,rolling temperature,and cooling temperature on the performance of steel is studied.Meanwhile,the variation in strength before and after the skin-passing process,and the relationship between the change value of strength and its influencing factors,such as the initial strength,thickness,and skin-pass rolling force,are further investigated.On the basis of the analyzed data,the platform uses an estab-lished mathematical model to realize the accurate prediction of the strength of specific steel strips after various skin-passing processes,further suggest-ing the optimal rolling force of this process,as shown in Fig.6.

Fig.6 Process recommendation and performance prediction results

4 Conclusions

Engineers face many problems in coiled tubing steel strip production,such as data dispersion and inefficiency,difficulty in analyzing performance fluctuation influencing factors,trend variability in key index tracking,inconsistent multivariate relationships in different periods,and lack of accurate prediction of mechanical properties.To cope with these diffi-culties,a data analysis platform is preliminarily developed based on the big data service concept.The goal is to provide data visualization and com-prehensive analysis applications.

(1) The platform assembles the features of an over-view survey and stability analysis for coiled tubing steel strips in five major parts,namely,overall descriptive statistics,a centralized profile interface,key index tracking,an abnormal information summary,and a customizable pivot table,considering the usage requirements of manufacturing managers and product engineers.This approach realizes accurate tracking and monitoring of production information and dispenses with tedious manual data preparation and analytical processing,forming the basis for prompt responses and effective decisions.

(2) The platform implements a comprehensive analysis of production data of coiled tubing steel strips.On the basis of the function and module designs,the platform provides the functions of exp-loratory data analysis,correlation analysis,and multivariate statistics,facilitating the initial data exploration inspiration and diagnosis of crucial influencing factors and laying the foundation for deeply exploring product characteristics and formu-lating scientific and reasonable optimization strategies.

(3) On the basis of data accumulation and full analysis of the factors affecting the mechanical properties of coiled tubing steel strips,the platform was used to study the regulation of relevant variables,such as composition and process in produ-ction,and to realize the precise prediction of steel strength under conditions of different rolling forces after the skin-passing process.Thereby,the platform provides the recommended optimal skin-pass rolling force,enhancing the foresight and effectiveness of production decision-making and providing theoret-ical guidance for stable performance.