, ,

Research Institute,Baoshan Iron & Steel Co.,Ltd.,Shanghai 201999,China

Abstract: Digital transformation is the process of an industry moving to a digital business of continuous growth and improvement with digital technologies.It is an important path for the manufacturing industry to reconstruct and upgrade the existing operation mode to achieve efficiency,effectiveness,and product quality improvement.In this study,we analyze why the successful digital transformation practice is scarcely found in the core business of process manufacturing industries,such as the steel and cement industries.By introducing the development and application of the Process Intelligent Data Application System (PIDAS),which is the digital platform of heavy plates in Baosteel,the concepts,ideas,and development experience of the digital transformation of the process manufacturing are summarized.Moreover,the core issues worthy of attention are pointed out.Finally,a feasible route for the digital transformation of the process manufacturing is proposed.

Key words: process manufacturing; digital transformation; steel industry; PIDAS

1 Introduction

The recent years have witnessed the fast progress in the field of information and communication technology (ICT) and the quick changes in the daily lives and workplaces of individuals.All these deve-lopments stem from the popularity of various sen-sors;diffusion of the Internet at an extraordinary speed;explosion of wireless communication;incre-ment of mass data storage technology;continuous deve-lopment of distributed computation;wide applica-tion of machine learning and optimization algorithm supported with interpreted,object-oriented,and high-level programming languages,such as Python;and realization of high computing performance with the help of the graphics processing unit[1-3].In other words,the progress in ICT improves data availabi-lity,universal connection,high computing power,and effective realization of computing algorithms.

Nowadays,data from multiple sources are of high feasibility to be collected,transmitted,stored,and treated and are consequently monetizable.Owing to the Internet,whether wireless or not,the world has been tightly connected,and the universal connection is not utopian anymore.Data could be synergi-stically processed without a spatiotemporal barrier,and the knowledge and wisdom generated from them could be easily shared.The transformation from data to information,then to knowledge,and finally to wisdom is effectively implemented with traditional or contemporary computing algorithms with high computing power.Normally,the data people use on a daily basis could hardly be called big data,which are commonly defined with the features of 3Vs (i.e.,volume,velocity,and variety) or 5Vs (volume,velocity,variety,variability/veracity,and value).However,they are of massive size,quickly accumulated,in various formats,coming with different qualities,and,after all,very valuable indeed.

Manufacturing is also profitable from such prominent technological progress.Actually,the so-called discrete manufacturing has achieved a big leap in terms of production and market efficiency,effectiveness,and product quality when performing digitization,digitalization,and digital transfor-mation.Digitization,digitalization,and digital transformation are commonly defined as the three digital stages of industries.Digitization means con-verting non-digital items into their digital represent-ation (zeros and ones) by facilitating computer storage,processing,and transmission.Digitalization is the use of digitized data and analytics tech-nologies to improve and optimize operations.Digital transformation is the process of an industry moving to a digital business of continuous growth and improvement with digital technologies,such that the company changes its operating method and employs such changes in its working process.

In discrete manufacturing,a product is assembled on the production line from parts listed in the bill of materials.There is no geometrically,physically,or chemically change in the parts during production.As a result,the features of parts could be digitized,and the value of digitized data could be extracted to improve the product quality,reduce the management and material costs,optimize the production,deliver planning,and customer experience.In the process manufacturing,quite to the contrary,products are made with recipes containing different composi-tions,and the workpiece changes geometrically,physically,or chemically during the manufacturing process.To understand such process-dependent changes,quite complicated professional knowledge of specialists with many years of experience in the field is needed in normal circumstances.That is,the features of the workpiece and the relationship among the workpiece features and process are very difficult to digitize and digitalize without ICT and special professional knowledge.These factors explain why the successful digital transformation practice is scarcely found in the core business of process industries,such as the steel and cement industries.

In the plate mills of Baosteel,a platform called the Process Intelligent Data Application System (PIDAS) was built for digital transformation pra-ctices by a joint team with members who possess ICT or metallurgical technology backgrounds[4-5].The basic structure of the PIDAS,its main function,and development experience are summarized and introduced in this paper as an example of the digital transformation of the process manufacturing.

2 Development of PIDAS in Baosteel

2.1 Requirements from the continual operation optimization

The steel industry is a typical process manu-facturing.Plate mills are important production plants characterized by varying plate steel products in multiple and complicated steel grades/specifi-cations and small production lots for engineering applications.Slabs/ingots with a designed chemical composition are reheated to elevated temperatures,normally approximately 1 200 ℃,and then undergo roughing and finishing rolling,followed by water or air cooling and hot leveling.After cooling down to room temperature,rolled plates are cut to custo-mers’ preferred sizes.Generally,a rolled mother plate is divided into several final plates for different orders.Furthermore,some plates require heat treat-ment such as quenching,tempering,and nor-malizing.When a workpiece goes through all these processes,its three dimensions and temperature,physical,microstructure,and mechanical properties,as well as surface,are continually changing;oxide scales are continuously formed and removed on its surface.Moreover,the states of tools/equipment,such as reheating furnaces,mills,cooling machines,and levelers,and process measurement facilities,such as pyrometers,thickness gauges,profile meters,and plan view measurement devices,are slowly varying.As a result,it is quite difficult to digitize the workpiece,tools/equipment,measure-ment facilities,and their changes and interrelationships.Hence,plate steel production is a typical process manufacturing,and the digital transformation method of plate mills could serve as a good example of the process manufacturing.

As shown in Fig.1,the overall control targets in plate steel production are quality,cost,delivery,development,and service (QCDDS).In the era of information and automation,which is conventionally called the Third Industrial Revolution,the overall multiple targets are divided into several sub-targets with the so-called material design and quality design on level 3 manufacturing operation and control computer system after receiving order information from level 4 business planning and logistics computer system to level 2 process control computer system that controls individual processes.The models in level 2 calculate the setup values and send them to level 1 basic automation systems,in which the most popular programmable logic controller/technology and drive control are used for execution.Level 1 systems also collect measured values from level 0 measuring devices for feedback control and send them back to level 2 for post-calculation and model optimization.Then,the level 2 system sends them back to level 3 for quality back-tracing as production records.In such a way,data with different frequencies and formats are taken as either production instructions or parameters to be executed without any change at the receiving system,information to be referred to when making adjustment decisions,or production or process records to be stored and recalled when needed for a set period at various computer layers.

Fig.1 Illustration of the plate mill control

The kernel idea of a four-layer computer control system is to control the whole manufacturing process according to the quality and material design,mainly backed up by the expert knowledge and experience of designers and statistical analysis.When performing the statistical analysis,designers hardly receive the support of sufficient systematic data,especially high-frequency process data.The data they need are normally stored in different computer systems of various platforms,neither well connected nor well sorted.Furthermore,many val-uable data are deleted when their mission as transi-tional information is completed.Consequently,it is hugely difficult for designers to reach an optimized design in time,and optimization is time-consuming work.The overall optimization can only be con-ducted offline.

Some online adjustments are performed at the level 1 and 2 systems.For example,the rolling schedule is continually optimized in the mill level 2 system during the rolling process.The rolling schedule is recalculated several times according to the measurement results after pre-calculation before rolling.Then,the post-calculation is performed when rolling passes have been performed.Another example is the automatic gauge control in level 1,where the measured deviation of the separated force or thickness of steel can be directly transformed into the adjustment of the roll gap.However,neither the optimization of the models in level 2 nor the optimization of the algorithms or logic in level 1 could be directly achieved by such automation at levels 2 and 1 because of the shortage of systematic data and models for the optimization.

2.2 Development of the PIDAS

The PIDAS has been developed and continually enhanced in the plate mill of Baosteel to lay the data foundation for quality and material design optimization in level 3,models in level 2,and algorithms or logic in level 1 and aid in various decision-making processes in level 4.

The prototype of the PIDAS was first built in 2010 when the renewed automation control system of water cooling after rolling was commissioned to observe the control effect and optimize the control system.In this process,all related data from level 1 to level 3 are collected,labeled,connected,and descriptively visualized.In 2015,the platform extended to cover all the rolling lines from reheating to hot leveling to optimize the property control of all steel plates undertaking the thermo-mechanical control process (TMCP),especially pipeline steels,with breakthroughs at the data ingestion across multi-level systems.In 2019,the PIDAS covered all the production processes in the Baoshan Works of Baosteel Plate Mills in Shanghai.Until now,the platform has been extended to connect Zhanjiang Works of Baosteel Plate Mills in Guangdong and Ezhou Works of Central South Steel Plate Mills in Hubei.

The configuration of the PIDAS is illustrated in Fig.2.For the convenience of introduction,the figure is divided into five parts,labeled with P,I,D,A,and S.

Fig.2 Illustration of the configuration of the PIDAS

Part I is the infrastructure of the PIDAS.The data are collected from very-high-frequency ones gen-erated in level 0,level 1,and level 2 systems through quite-low-frequency ones generated in level 3 and 4 systems in the format of time series,pic-tures,audios,videos,texts,tables,documents,and databases.Energy consumption records,equipment checking and maintenance records,and other oper-ation records are also collected.

Part D is the data management center of the PIDAS.In this part,collected raw data are proces-sed,validated,denoised,reorganized,integrated with metadata based on metallurgical and mech-anical knowledge,and restored to the database of the PIDAS.Raw and integrated data are available for users to access and utilize and are ready to send to company data centers or other data platforms.

Part P prepares a series of platform adapters for connecting the PIDAS with a company data center and all the data centers of upstream and downstream processes or customers beyond the company.Raw and integrated data from plate mills can be sent to the place where they are needed and vice versa.The PIDAS is also ready to receive data from other data platforms.

Part A supplies the analysis toolkit and application program interface for users.In this part,some conventional descriptive and knowledgeable applications are developed for aiding quality and process designs,monitoring production processes and main equipment status,observing production results,correlating processes and properties,outputting production reports,and analyzing the reasons for defects and breakdowns using tables,figures,visualization analysis,and statistic tools.An additional toolkit is integrated with data extracted,transformed,and loaded from the PIDAS database to the local database.The most popular machine learning methods and statistical analysis methods are constructed and continually improved for users to perform data analysis and develop specialized applications.

Part S is constructed as a sustaining station of the PIDAS.In this part,new requirements on new data,functions,or operation methods from users and new modifications,changes on the PIDAS,and version updates from developers are listed and stored as structure data,forming a sustainable evolution center and restart base.

2.3 Changes brought about by the PIDAS

The basic effects brought about by the PIDAS can be summarized as improvements in production efficiency and effectiveness,product quality,and accumulation of easily inherited and applied digitalized knowledge.Here we define the production efficiency and production effectiveness on the pace and cost of material,energy,or human resource to implement a production process or perform a task for the produc-tion management,analysis,or quality improvement.Consequently,time reduction in filling daily pro-duction report tables,backtracking reasons of defects,trial production of new steel grades or specifications,and delivering a batch of products,are taken as an improvement in the production efficiency.Moreover,increases in the yield ratio,reduction in energy consumption,and increases in the per capita steel output are improvements in production effectiveness.

A daily production report was composed of manually collected data from more than five data resources and then reorganized with the help of a data sheet.It is a half-day work of one person,and typing or statistic errors could occur repeatedly.The use of the PIDAS requires just a click of a mouse,costing just a few seconds and without human error.

Mostly,sourcing the defect reasons was spent on collecting and assembling data,not on analyzing them.Generally,days were needed to collect and treat the data from different systems and in different frequencies.When the PIDAS is used,the time is shortened to hours at most.Moreover,with the help of the toolkit integrated with the PIDAS,the time spent on the analysis is shortened.

The PIDAS could provide a preliminary optimized design for new steel grades or orders with extra re-quirements without precedents by correlating the chemical composition,process parameters,and pro-perties,sometimes including the microstructure,based on accumulated historical production data.The time and materials needed for the trial pro-duction,conventionally twice or thrice,could be saved to a large extent.Subsequently,a quicker mar-ket and customer response could be achieved.

Optimized designs could also be implemented on the material design and rolling schedule planning.As a result,a high yield ratio could be realized on the plate rolling,and less energy consumption could be achieved.

Based on the analysis of historical data,the pro-cesses for high-quality products could be taken even further,and those causing defects or resulting in low-quality products could be avoided.Moreover,the integrated and dynamic process control can be pro-spected for achieving the overall QCDDS control targets.

3 Effective digital transformation method of the process manufacturing

In a sense,the PIDAS is a preliminary digital transformation practice of the process manufac-turing.Some breakthroughs have been made to over-come the difficulties of the digital transformation of the process manufacturing,as mentioned before.The main products,processes,equipment,and automation characters are digitized and connected based on already existing data in the current four-level automation computer systems and on the extra measurements added according to business/operation requirements.Some favorable application results are emerging,and much more promising changes are coming,including the reconstruction of the automation system to meet the instinctive demand for the overall production control of the QCDDS.

Five aspects of the experiences for the cons-truction and application of the PIDAS could be sum-marized here as a reference for the digital transfor-mation of the process manufacturing.For conve-nience,the five letters of the PIDAS could also be used to describe the five aspects/steps:planning a long-term vision,initiating action by utilizing the legacy system,developing with a joint team,asso-ciating to actual manufacturing scenarios,and spiraling upward following new demands.P,I,D,A,and S might be one of the effective ways to per-form digital transformation for a process man-ufacturing enterprise,as shown in Fig.3.

Fig.3 One of the effective digital transformation routes for the process manufacturing

3.1 Planning a long-term vision

Digital transformation is a long-time process,especially for the process manufacturing.People can-not just expect to plan a program that can be imple-mented in a short time and then acquire a great leap forward in business earnings or productivity.

To plan a long-term vision,the company is suggested to culturally foster a climate for digital transformation.Everyone in the company,including top leaders and line operators,should be aware of the values,necessities,essence,difficulties,and basic ways of digital transformation.

The values and necessities of digital transfor-mation are nowadays believed to be a consensus.How-ever,its essence is seldom thoroughly explored.To perform digital transformation is to follow a process that gradually makes the production and customer experience better based on fact-perspective progress driven by all-around data and the collaboration of all people involved,aided by the interactive inspir-ation between industrial and special professional knowledge and ICT.During the process,the company changes its operation mode by growing high-tech features of intelligent decision making and smart control.Then,individuals in the company change the work method by cooperating in cyber-space and instructing,investigating,analyzing,designing,and optimizing data and artificial intelligence (AI) tools.

In most cases,the greatest difficulty for a process manufacturing company lies in the lack of digital knowledge among employees who have accu-mulated abundant and valuable specialized manufac-turing knowledge.They may feel frustrated by the helplessness of being involved or the fear of being replaced and are not aware that their specialized manufacturing knowledge is the most precious treasure to the digital transformation of the process manufacturing.Hence,it is inevitable to train such employees to improve their willingness to embrace digital transformation and their competence for contributing their knowledge and skills for playing a new role in the digital era,with successful precedent examples of the team and project in the way of case teaching and goal-and-action-oriented practical learning inside the company.In Baosteel,the PIDAS development and application team has been selected to summarize the knowledge structure and development and application experience.The first one of the series courses was successfully com-pleted,indicated by the fact that many trainees in areas such as product manufacturing,process and material design,automation,customer service,and new steel and process development presented their solutions to their routine problems with the know-ledge,method,and inspiration from the training.

The previous successful examples came neither from nowhere all of a sudden nor from outside ICT companies.Instead,they came from the wide practices encouraged by the company and digital culture and carried out by willing and spontaneous teams.In Baosteel,the PIDAS is one such practice.The most recent practices in Baosteel for digital transformation are the Industrial Brain Program.A series of projects are set from the AI of iron and steel making to the digitalized research and develop-ment.

A long-term vision planned to cultivate digital culture,encourage wide practices,and train em-ployees with inside practical examples could lay a solid foundation for practical actions and promising fruitful achievements.

3.2 Initiating action by utilizing the legacy system

Sometimes,digital transformation is alienated as initiating an integrated data system to collect all data thought to be of great importance for the company and upload them to the cloud for people to use when needed.In fact,building an integrated data system is only technical support for digital transformation but absolutely not the digital transformation itself.The essence here is to make data highly feasible to reach the people who need them through the web and analysis tools,such as machine learning,visual analysis,and dynamic control.Some people from the ICT side,especially those who have great achievements in informati-zation,firmly believe that they can program any business process that people can clearly describe.

As mentioned above,digital transformation grad-ually brings changes to the manufacturing process.The progress continually and quickly evolves,and it is impossible to clearly describe the whole process throughout the transformation.The common dilemmas are that needed data are unavailable,and available data are not the ones required.A definite description could only be made for a short time.As a result,it is hardly possible to design and build a new integrated data system of high feasibility for usage.

In most of the current manufacturing process enter-prises,all kinds of primarily organized,isolated,or transient data already exist in different legacy systems,such as the well-equipped four-level com-puter control systems and different information systems for the administration of quality,equip-ment,finance,purchase,and human resources.In such a legacy system,not only the data but also the knowledge accumulated for enterprise operating or process controlling is of high value.When they are collected and integrated,as what we have done in constructing the PIDAS,big progress has been made.

3.3 Developing with a joint team

After the digital culture is cultivated,most individuals in the company would not expect an outside ICT company to build a technical support system for them to perform a digital transformation.According to Gartner’s investigation results,more companies prefer in-house software development.

If the PIDAS could be used as an analogy,the driving force to make in-house software necessitates embedding the special professional knowledge in the manufacturing process and making it employee-friendly and continually improvable with the enterprise evolution.A joint team with members from the ICT and manufacturing would make it much easier to fulfill such necessities.

The PIDAS development team is composed of members who have education/career backgrounds in computer science,industrial automation,material science,and material processing engineering and work backgrounds in process modeling,control system development,application of plate mills and hot strip mills,manufacturing process administration and information system development,flat steel product development and production,and commissioning and operation of plate mills for carbon steel and specialty alloys.The developers are also a part of users and have been working together for over ten years.Hence,the problems,scenarios,jargon,speci-alized and technical terms,and requirements are easily understood,defined,and digitalized.The key data and their sources,qualities,and effects are well recognized and treated during the construction of the PIDAS.Hence,the developing effectiveness and efficiency are well ensured.

3.4 Associating to the actual manufacturing scenarios and spiraling upward following new demands

Associating with actual manufacturing scenarios makes the values of systems sensible and achie-vable.Positive feedback and inspiration are con-sequently obtained.These values are very important for developers to persevere in the long-term deve-lopment and to get all-around support and opportunities to optimize the system.

The PIDAS has been winning in its evolving opportunities and accumulation of experiences,technical breakthroughs,skills,and knowledge step by step ever since the good results from using a pro-totype automation model for tuning post-roll cooling control.The prototype was followed by the first PIDAS,which integrated all data from the rolling line and promoted great successes in the high-grade pipeline steel product development and production and in the development and application of the TMCP.Then,the requirements for covering the whole plate mill were received and implemented,and the PIDAS is required to cover the different plate mills of Baosteel or Baowu Group.The PIDAS is simultaneously spiraling upward to version 2.0.

4 Several noteworthy issues

Many practitioners summarized some myths about digital transformation.Nonetheless,opinions often differ from one another.According to our exp-erience during the construction and application of the PIDAS,the following issues are considered.

4.1 Plan-do-check-action cycle rather than the top design is more followable

Digital transformation is an evolving business process.The target is prospective but not clearly descriptive.Concrete continual actions might grad-ually unveil the future.

4.2 Intelligentizing data rather than digitizing knowledge is more effective

The route of data-information-knowledge-wis-dom has been widely investigated and practically built.On the contrary,directly digitizing expert know-ledge is still frontier research.What we have at pre-sent are data that record or contain expert knowledge in the format of documents,manuscripts,reports,models,audios,and videos.Digitizing such data and intelligentizing them could be practical and effective.

4.3 Digital transformation rather than enhan-cement of informationalization is the core of all actions

Great leaps in productivity have been made in the steel industry when performing informationalization during the Third Industrial Revolution,and the pro-cess is still lasting in quite a large scope.The main feature of informationalization could be recognized as firmly controlling the enterprise with a fixed oper-ation flow or mode.From this point,enhancing infor-mationalization is completely the opposite of digital transformation.Digital transformation is a continual evolution process of an enterprise evolving to a better mode with the help of data and ICT.

4.4 Algorithms or AI serves as an aid rather than a substitute for human beings

Sometimes,algorithms or AI is mistaken as the universal measures that can even replace human beings in any situation.Algorithms or AI can help people solve problems,of which some have never been solved before.However,it is impossible to define a situation and apply algorithms or AI without the help of human beings.

4.5 Solving problems met in daily work rather than those imagined theoretically

When the benefits of digital transformation are emerging,much more will be expected in some companies.To further take advantage of digital transformation,several problems are theoretically listed to be solved as urgent and important items.However,some theoretical issues are proved to be neither urgent nor important,even not practical at all.Accordingly,some methods are suggested to list the problems met in daily work and to solve them with priority in terms of practicality,feasibility,impor-tance,and urgency.

5 Conclusions

As one of the practicing projects encouraged by the company,the PIDAS has been successfully developed as technical support for Baosteel’s digital trans-formation.Its developing route could be sum-marized as follows:planning a long-term vision,initiating action by utilizing the legacy system,developing with the joint team,associating to actual manufacturing scenarios,and spiraling upward following new demands.The route is supposed to be one of the effective examples of digital transfor-mation for the process manufacturing.