Int. J. Precis. Eng. Manuf.-Smart Tech. > Volume 2(2); 2024 > Article
Park and Tran: Hybrid Model Based Autonomous System for Controlling Double Side Grinding Process


Manufacturing uncertainty is rising due to diverse customer demands and new technologies. These challenges are propelling us into the Industry 4.U (Industrial Revolution 4.0 Upgrade) and Industry 5.0 era, focusing on human-centric manufacturing and value creation, necessitating digital transformation. This transformation relies on the core technologies of the 4th Industrial Revolution (Data Analytics, Connectivity, Autonomy, and Collaboration) to perform value-creating processes autonomously and reliably. To adapt, Industry 5.0-oriented Smart Manufacturing architecture was designed. In this system, process planners collaborate with generative AI to generate optimal process plans based on new specifications, pursuing human-centered manufacturing. The manufacturing system is reconfigured with modular technology, and each process is autonomously controlled as a 4.U/5.0-oriented production system. This paper focuses on developing an autonomous process control system for Industry 4.U/5.0-oriented Smart Manufacturing. The study aims to stabilize the quality of the brake disc grinding process in an automotive parts company. The developed control system has been tested on a car’s brake disc manufacturing line, with experimental results showing the reliability of the DT and AI in controlling the grinding process.

List of Symbols


Digital Twin


Artificial Intelligence


Cyber Physical System


Internet of Things


Cognitive Agent


Information Communication Technology


Artificial Neural Network


Human-machine Interface

1 Introduction

Intelligent manufacturing and automation are applied for manufacturing companies to achieve efficiency, productivity, quality and competitiveness. Industry 4.0 is considered to be technology-driven, where an intelligent networking of machines and processes for the industry based on cyber physical systems (CPS) is established [1]. With Industry 4.0, technologies are used to increase operational efficiency. In the Industrial Revolution 4.0 Upgrade (or 4.U), technologies are used to increase value for customers and users [2]. However, at many points in production, these methods, technologies and processes show some limits to make transparency along all product life cycles, to control and optimize complex process behaviors in value creation stages and to orient to the actual customer requirements and needs. To transcend these limits, Human centric Solutions, Autonomy and Digitization as core elements of Industry 5.0 are now empowering manufacturing. With value-driven in Industry 5.0, core human needs and interests are placed at the center of the production process.
Humans and machines are working together to improve the efficiency of industrial production [3,4], in which the hybrid intelligence (one form of collaboration between machines and humans) is focused. In order for machine intelligence to improve decision-making processes, humans and machines must collaborate [5]. So, the advanced technologies such as artificial intelligence, digital twin, human-machine interaction technologies, bio-inspired technologies and smart materials are used in machines to equip them with intelligent functions [1].
For realizing autonomy and digitization characteristics of the production systems in Industry 5.0, data analysis and real-time data connection keep an important role. So, data driven manufacturing is required. In consideration of these characteristics of Industry 5.0, the framework of implementing Industry 5.0 is designed to cope with current manufacturing environment. Based on this framework, this paper firstly focuses on the development of an autonomous process control system as a part of this framework. An autonomous control system is developed to monitor process behaviors and identify current state for reasoning and making decisions in case of disturbances.

2 Architecture of Industry 5.0 Oriented Smart Manufacturing

Smart manufacturing is a set of manufacturing activities that use networked data and information and communications technology (ICT) to manage manufacturing activities [6]. Toward core values of Industry 5.0 such as human-centricity, autonomy, adaptability, data driven manufacturing, the architecture of an Industry 5.0 oriented smart manufacturing is designed in Fig. 1. The Architecture include six layers (physical, communication, information, modelling, digital simulation, and AI-driven layers). These layers can be classified into four domains including observable manufacturing domain (physical layer), communication domain (communication and information layers), digital twin domain (modelling, digital simulation layers), and user domain (AI-driven layer). If classified by function, these layers are grouped into 3 levels such as ACT, SEE, and THINK levels. In which, at the level 1 “ACT”, the physical machines and equipment of the manufacturing system are controlled by agents such as handling agent, turning agent, grinding agent, logistic agent, inspection agent, and balance agent. With these agents, each unit of the manufacturing system is an autonomous element. Through agents, the physical system is connected to level 2 “SEE” which contains digital models of the physical system. This connection allows receiving and monitoring the status of the manufacturing system. Digital data of level 2 is stored in level 3 “THINK”, combined with algorithms based on artificial intelligence that allow the system to make control decisions to ensure the manufacturing system operates optimally.
For the AI system in the cloud to derive the new product concept and modify an existing product according to market demands or customer requirements, knowledge, information, and data will be collected from the operation and simulation of the system and process, as well as the use phase of the product by using 5G and 6G networks. To manufacture the new or modified product in effect and efficient way, the optimal process plan is generated by the collaboration with planner and generative AI which can suggest alternative solutions. Among them, the optimal process plan is selected by planner through cooperating with generative AI in human centric way.
By using data analytic and digitalization technologies, digital twin system was developed to check the reliability of the generated system and process plan through modeling, simulating and evaluating their behaviors before carrying out real manufacturing. The developed DT allows planning manufacturing processes faster and more efficiently as well as allowing assessment of quality with real-time process analytics.
These whole planning processes are realized in Metaverse environment. In this environment, through the cooperation of AI system in cloud and the system planner, the manufacturing system is reconfigured with modular and DT technique for adapting to the newly optimized process plan.
Through the connectivity, manufacturing process behaviors are monitored and diagnosed in real time way in the Metaverse environment. To take measures against the detected disturbances, the components of manufacturing system have been implemented as a cognitive agent (CA) with the ability to perceive, reason, and make decisions based on collected knowledge, information, and data. This implementation allows the components of the manufacturing system to cooperate with each other to solve problems that arise through negotiation. Within the scope of this study, we only consider building an autonomous control system for the grinding process.

3 Hybrid Model Based Autonomous Grinding Process Control System

3.1 Problem Description

Currently with the brake disc production line, the steps include robot arranging of the semi-finished product, balancing system, grinding process, machine vision, and logistics of finished product.
Fig. 2 shows the quality failures in grinding process and the proposed method to improve the grinded part quality. For the current grinding quality assurance process, with the optimal parameters of the grinding process initially determined, the operator proceeds to set the grinding process parameters on the grinding machine. Products after grinding are inspected and evaluated for quality. In case of quality problems, experts will evaluate and redefine the grinding process parameters. In the proposed research, the grinding machine is connected to the digital twin model and sensor network to receive data of the grinding process in real time.
The control system is built with control algorithms based on artificial intelligence techniques that allow the use of the above data to monitor the current status of the grinding machine as well as the grinding process. From the monitoring results, the control system makes decisions to the grinding machine such as adjusting grinding process parameters in case current parameters affect grinding quality. In this study, an autonomous control system for the grinding process of automobile brake discs was built and tested with DT and AI applications.
With the current grinding process, brake disc parts after grinding have some quality problems such as unsatisfactory surface roughness and burn marks. Fig. 2 shows the products of the brake disc, the grinding process and some errors in the quality of the grinding product. An autonomous control system for the grinding process is essential for the manufacturing line to ensure the quality of grinding products and increase the productivity of the manufacturing line.
To solve the above problem, a hybrid model for grinding process control system has been proposed. Nikolaev proposed a hybrid model for gas-turbine power plant which includes data-driven model and physics-based model. This hybrid model is effectively used to for maintenance of the gas-turbine flame tubes [7].
In our research, the hybrid model is a combination of digital twin model and AI model with artificial neuron network (ANN). ANN architecture has one input layer, one output layer, and three hidden layers. With three hidden layers, the neurons for layers are 8-4-8, respectively.

3.2 Development of DT System for Describing Grinding Process

The concept of digital twin is the development of modeling and simulation technology. Traditional simulation methods still limitations in evaluating the performance of the system to be designed. DT technology allows overcoming the above limitations by integrating modeling, analysis, IoT and AI integration [8,911].
The digital twin concept consists of three parts: (i) the physical entity (object or process) and its physical environment, (ii) the digital representation of the entity, and (iii) the communication channel between the physical and virtual representations [12]. Jairo and Yang proposed a novel framework for the design and implementation of DT applications to the development of autonomous control engineering [13]. DT is used to visualize the physical model, track changes, and understand and optimize the system throughout the product lifecycle analysis. Additionally, data collected from the digital twin is used to refine physical models, optimize manufacturing processes, manage supply chains, and manage product quality. The application of DT in smart manufacturing allows reducing time to market by designing and evaluating production processes in a virtual environment before production [14].
In this study, with a DT-based grinding system, a hybrid digital twin model is proposed. Hybrid digital twin is a combination of two models, physical-based digital twin and data-based digital twin. The laws of physics and empirical knowledge are applied to the physical digital twin, whereas the data-driven digital twin uses machine learning algorithms to find hidden relationships between inputs and outputs.

3.2.1 Systematic Procedure for Developing the DT System

Fig. 3 shows the steps for developing the DT system to a grinding machine for grinding brake disc products. These steps include Generating 3D model with mechanism of components; Describing physical/mechanical behavior of grinding machine; Generating process controller model of virtual system; Development of grinding force prediction model; and Validation of functionality.
In mechanical modeling step, geometry analysis and mechanical coupling analysis are performed to obtain virtual models of real systems. In physical property modeling step, static/dynamic characteristic analysis, force transmission mechanism, and process mechanism are carried out for simulating the system’s operations. And with the final step is control modeling, in this step information/signal flow analysis, control function analysis, and mapping control function to mechanism are performed.
The results of this step demonstrate how the DT system differs from traditional simulation methods. Integration of AI and DT has more efficiency with production planning, production control, and quality control applications [15]. For DT system deployment, IoT, cloud computing, artificial intelligence, advanced visualization tools and other related technologies must be integrated together to provide a virtual environment while also allowing physical system monitoring [16]. The input parameters are workpiece speed, vibration error, wheel entry speed, and coolant flow. The output parameter is grinding wheel speed. With sensor networks and IoT, data about the grinding process is analyzed and processed. This data is compared with monitoring data from the DT model. In case the deviation between the digital model and the physical model exceeds the allowable limit, the AI model will generate a new set of optimal grinding parameters to update the grinding machine through the control system.

3.2.2 Implementation of DT for Grinding Machine and Process

For generating DT model, analysis of grinding machine for describing model, physical model for predicting mechanical loss, operating model of virtual controller, and evaluation of DT system were carried out. Analysis of Grinding Machine in Terms of Static and Dynamic Behavior

In the step of analysis of mechanical behavior of grinding machine, sequence of the actions on the grinding machine is analyzed as shown in Fig. 4. The actions include product input, cylinder lift, product chucking, carrier device move up, carrier device forward, carrier device move down, un-chucking, carrier device return, work drive rotation, work spindle rotation, wheel spindle rotation, grinding wheel forward, and grinding wheel return. Calculating Mechanical/Electrical Loss of Grinding Machine with Physical Model along Force Flow

In the analysis of physical behavior and characteristics along force flow, the contents evaluated and analyzed include: power flow in spindle in consideration of loss, modelling spindle power losses, mechanical losses in spindle, electrical loss in spindle, and modelling belt power losses.
To accurately assess power losses for the machine’s drive system, modeling belt power losses and power flow in spindle are established as shown in Fig. 5. Losses of the power converter, motor’s electrical losses and unproductive mechanical power are calculated.
For calculating the electrical losses, the total losses due to electrical phenomena are considered such as copper losses in the stator and rotor, iron losses, and stray losses as shown in Fig. 6. In which, cooper losses are calculated from the electric resistances of stator and rotor materials. Iron losses are the result of hysteresis and eddy currents induced in the stator and rotor by the rotating magnetic field. An average stray power loss is calculated at 1.2% of useful power. From the mechanical and electrical power losses, the power total for grinding machine is calculated. Realizing a Controller of DT for Grinding Machine

With the step of modeling behaviors for implementing virtual controller, the contents considered include:
  • Analyzing motion control and logical control through ladder diagram;

  • Data setting for the communication of the controller with the virtual machine;

  • Description of the control characteristics.

  • Configuration of the virtual PLC.

Fig. 7 shows the details of the above steps for mapping the operational process signal to digital grinding machine components. Based on the analysis of acquisition data, signals from hardware, software, and sensors of the real grinding machine, a virtual grinding machine with behaviors similar to the real machine has been built. With analyzing motion control and logical control through ladder diagram, the input and output parameters for controlling grinding processes are determined. Data types of these parameters are established for communicating the controller with the virtual machine. From the results of these two steps, description of control characteristics is carried out. And finally, configuration virtual PLC is established. The virtual PLC enables to simulate the grinding process with the input and output parameters similar with the real grinding process.

3.2.3 Implementing a DT System for Grinding Machine

For implementing the digital twin system for grinding machine, two stages are carried out that include the integration of the modelled elements of the grinding machine with their operational process; and the control mechanism of the grinding process is established [17]. In the first stage, with each elements of the grinding machine, geometric attributes, physics attributes, and behavior attributes are added. This information enables a dynamic interaction between machine’s elements and operational processes. In the step of embedding the physical characteristics into grinding machine, moment inertia, friction coefficient, clamp force of the workpiece table, the index table, and spindle are calculated as shown in Fig. 8. With a grinding system consisting of many machine elements, modeling, integrating physical characteristics and operational process of the machine elements take up a lot of time when building the DT system. For implementing this stage, a relational rule model is used. With this method, the integration of the digital model and operational process is achieved using the steps including [17]:
  • Carrying out a logical mapping of the machine elements’ attributes to the operational processes. This is done by identifying the value-added (geometric, physics, behavior) machine elements;

  • Developing the operational process and the grinding machine’s digital model;

  • Developing a relational rule model that logically implements the interaction between the operational process and the machine’s digital twin attributes. This establishes the logical connection between the digital elements with their operational processes, respectively. With this dynamic interaction, the machine-process digital twin information model is established which is used for describing the machine’ dynamic behaviors.

In the second stage, the control mechanism is established in which the grinding force with engineering model for digital twin is calculated as presenting in section 3.2.4.

3.2.4 Calculating Grinding Force with Engineering Model for DT

For realizing the functionality for predicting the grinding force of the DT system, grinding force model and control model for describing cutting mechanism are established. The grinding force value affects to the surface quality of the part. So, a data-driven digital twin model for predicting grinding force was carried out [18]. The grinding force model for describing cutting mechanism and engineering model for getting the grinding force value are shown in Fig. 9.
In the grinding force model, the relationship between the grinding force with the grinding wheel velocity, depth of cut, the grinding stone shape and the workpiece shape is determined. From this model, architecture of the control model for getting the cutting force is established.
To predict the grinding force, the effective cutting depth of grinding and the cutting speed must be determined. The values of cutting speed when grinding in the X and Y directions are determined by the following formula:
In which D1: grinding wheel diameter; ω1: rotation speed of grinding wheel; ω2: rotation speed of brake disc; and L: center distance between grinding wheel and brake disc.
To control the grinding process, the control diagram is also described in Fig. 9. To predict the grinding force F, the effective cutting depth of grinding must be determined in consideration the relationship of this value with the programmed cutting depth, deflection and thermal expansion.
Ha, Ks, Hs, and λ are transfer functions to get the depth of cut ae, grinding force, grinding stone shape, and deflection (x) respectively. The formula for calculating the effective cutting depth of grinding is as follows:
In which ae: effective cutting depth of grinding; ap: programmed cutting depth; as: grinding wheel wear; x: deflection; and xexp: thermal expansions.

3.2.5 Evaluation of the Developed DT System

The results of simulating the grinding force of the DT system and measuring the force of the real grinding machine are shown in Fig. 10. With the real grinding machine, from grinding process parameters, the grinding force is calculated and displayed on the HMI device and control room.
To make it more convenient to deploy the system and process data, the data obtained from the real system and the DT system is the motor current. From this value, the grinding force is calculated.
With the digital twin system, from the virtual grinding machine and grinding process model, the grinding force is simulated and displayed. In the case of the deviation between the grinding force value simulated by the DT system and the grinding force value from the grinding machine, the control system based on AI will generate a new optimal grinding value [19,20]. This renewing process parameters will be sent to the machine controller for realizing grinding process.
To evaluate the developed DT system, motor current values from the real grinding process and from the DT system were collected. Fig. 11 shows the deviation between the actual motor current value (displayed by blue color) and the simulated motor current value (displayed by red color). The formula for determining the precision value is as follows:
in which n: grinding process time (s); Ysim: simulated current value; Yreal : real current value. With n equals 25 (s) and the upper current value, the precision value: ηUpper = 95.01%; and for the lower current value: ηLower = 95.37%. The deviation between the actual motor current value and the simulated motor current value within 5%. This value demonstrates the high accuracy of the DT system.

3.3 Design of Hybrid Model for Controlling Grinding Process

The architecture of the developed system which enables to adjust the grinding process parameters in consideration of the grinded part’s quality is shown in Fig. 12. In the monitoring system, the hybrid model is used. The sensors-based monitoring module enables to get the data from the vibration signal. The digital model enables to have the reference values which are stored in database (DB) of the monitoring system.

3.4 Implementation of An Autonomous Grinding Process Control System

3.4.1 Design of Framework for Monitoring Grinding Process

For carrying out the monitoring module of autonomous grinding system, data flow model is described in Fig. 13. The current status of the real grinding process is collected by using the object linking and embedding (OPC) server due to connecting the programmable logic controllers (PLC) to this OPC server. These data are sent to the monitoring system and DT system. These systems will use these data to make a decision in considering the quality of the grinded part.

3.4.2 Design of Autonomous Process Control System Architecture

Fig. 14 shows the combination of using digital twins, cognitive agents and artificial intelligence to build an autonomous control system for the grinding process. The control system is a cognitive agent that receives data from the physical system through vibration sensors and data processors, grinding process parameters from the database.
The digital model connected to the physical system allows the grinding force to be diagnosed. Based on the above data, the system monitors the current status of the grinding process. In case an error occurs or the grinding process value deviates from the desired value, the system will generate new grinding process parameters based on artificial intelligence algorithms and update the ongoing grinding process to ensure the quality of grinding products.

3.4.3 AI Assisted Generation of Process Control Algorithm

To deploy the DT system to the shop floor, the block diagram and software tools are used as described in Fig. 15. Control data from the PLC is feedback to the digital twin model of the grinding machine using virtual PLC and Siemens NX MCD. Data and digital models are displayed in the control room and the human-machine interface (HMI) devices for system status monitoring. For realizing autonomous behaviors, artificial neural network (ANN) is used for generating process control algorithm. Optimizing AI Model for Deriving out Threshold Value

In order to reduce quality failures such as surface roughness and burn mark etc., monitoring vibration and its features during operation are carried out. For process control system and failure recognition, the interrelationship between process parameters and the vibration signal feature for failures must be established. For that, experiments with variable process parameters are needed. DoE experiments are designed to collect failure data and establish relationship with process parameter data and vibration. The process parameters include grinding wheel speed, workpiece speed, wheel entry speed, and coolant flow. The vibration signals are collected from the grinding process by the vibration sensor as shown in Fig. 16.
These experimental data were used for training the artificial neural network (ANN) with 80% training data and 20% test data. The initial ANN model is shown in Fig. 17 with four process parameters at the input layer, three hidden layers with the number of neurons 10-8-4 respectively. The vibration value is the output parameter.
For optimizing the ANN model, different ANN models with different number of layers and nodes were tested as described in Fig. 18. With the smallest value of the Mean Absolute Value (MAE) (0.001) and the highest value of the accuracy (98.1%), the optimal ANN model with three hidden layers and the number of neurons 8-4-8 respectively is selected. Generating Control Algorithm for Adjusting Control Parameter Selecting Control Parameter

Experiments with Taguchi L27 orthogonal array analysis were carried out to find the effects of process parameters on surface roughness and vibration characteristics during the process. With four process parameters including grinding wheel speed, workpiece speed, wheel entry speed, and coolant flow, the grinding wheel speed is the most influencing factor in terms of surface roughness and vibration as described in Fig. 19. To reduce the experiment and measurement cost, we used vibration signal as monitoring signal.
The grinding force affects the quality of the grinded part. Therefore, determining the grinding force value during the grinding process to ensure it is within the allowable limit is very important. The DT system was built using vibration signals to diagnose the grinding force. Through experiments as shown in Fig. 19, the relationship between vibration values and grinding process parameters is determined. Rule Based Autonomous Control Algorithm

For the decision making to keep the part quality during the grinding process, a rule based autonomous control algorithm was established. The failure mode and effects analysis method for developing the grinding control system were applied [21]. With the greatest impact on the quality of grinding part, the grinding wheel speed is selected as the main process parameter for controlling. In order to ensure consistency, the part quality during the grinding process, from calculating the deviation between the vibration signal value of the real grinding process with the threshold value, the grinding wheel speed is adjusted. In Fig. 20, in case of the deviation is bigger than 0.5, the grinding wheel speed reduces 2.5% (from 850 to 828 RPM) in order to keep the surface quality of the grinded part in the allowable limits.

4 Implementation of An Autonomous Grinding Control System

To develop the autonomous process control system with DT, and AI applications, the steps include grinding process behavior analysis and quality issues; influencing factor identification; design of monitoring system architecture; design of the autonomous control system architecture; data acquisition module; data preprocessing for control system; generating AI model for training; developing control algorithm; and implementation and testing. An autonomous grinding control process system was implemented as shown in Fig. 21. The implemented modules include getting processing data from vibration sensor; data collection and storage; feature extraction; vibration monitoring; error detection; autonomous control for generating new process parameters.
In monitoring and diagnosis system, time domain signal features are used to monitor process behaviors in term of vibration and force. From the data send by the monitoring and diagnosis system, the autonomous control module will make a decision in case of detecting error such as a bad surface of the grinded part (by comparison of deviation between the real vibration signal value with the threshold value). A new process parameter in considering the part’s quality is generated and sent to the controllers of the real machine for carrying out the grinding process. The feature extraction module, and rule-based control module were programmed.

5 Application of the Developed Autonomous Control System to Practice

To deploy the developed autonomous control system into practice, the developed system was tested for the manufacturing line for the brake disc grinding process, modules of the system have been built including: monitoring interface; smart process monitoring and control system HMI; DT module; and AI based control system as shown in Fig. 22. In the normal state, the grinding process parameters through the machine controller are transmitted to the machine to execute the actuators to perform the grinding process. In parallel with this process, the parameters of the grinding process are transmitted to the virtual machine in the DT system via the virtual controller.
Through the monitoring module, data from the grinding process on the real machine (acquired from sensors) and data from the DT system are compared. Comparison results through diagrams are displayed on the HMI and control room. In case there is a difference between real data and diagnostic data from the DT system (specifically in this study is grinding force) that affects the quality of grinding products, the AI-based system will generate a set of new grinding parameters. These parameters are then updated to both the real machine and the DT system.
The system performance test results have been successfully implemented for a new manufacturing line as shown in Fig. 23. The requirements from the manufacturing system include 4 types of brake disc, 3,600 brake discs per day. With the traditional manufacturing line, the number of qualified products is 88%; 12% of products have failures such as burn marks, surface roughness does not meet requirements. With the manufacturing line applying DT and AI, the number of qualified products reaches 95%. This allows saving materials and energy, improving productivity and quality.

6 Conclusions

To adapt to current harsh manufacturing environments, a new manufacturing concept such as Industry 4.U/5.0 is emerged. It applies advanced technologies to manufacturing to create more value for people. Based on this concept, the architecture of the Industry 5.0 oriented Smart Manufacturing was designed. In this architecture, DT, generative AI, Cognitive Agent, Data driven Method, Connectivity through ICT are core technologies that equip the manufacturing system with Human centric Solutions, Autonomy and Digitization. The proposed manufacturing system enables to adapt to the changes from customer requirements as well as improve the efficiency, productivity, and quality. To deploy this concept to practice, an autonomous grinding process control system with the target to keep the quality stabilization of the brake disc grinding process in an automotive parts company has been developed, built and tested for performance in a real manufacturing line. This is the initial result of the project. The results show that applying the autonomous control system to practice allows to save material, manpower and improve the part’s quality. The contributions of this research include:
  • Proposing the architecture of Industry 5.0 oriented smart manufacturing in consideration of the concept of Industry 4.U/5.0 with the application of generative AI, autonomous process control, digital twin, cognitive agent, 5G/6G networking;

  • Developing the practice oriented autonomous process control system with high reliability;

  • Implementing the DT system with high accuracy of more than 95% through detail analyzing the real behavior of the grinding machine along force flow, considering the mechanical and electrical losses;

  • Design of the autonomous process control system architecture using hybrid model and AI;

  • Adjusting process parameter through rule-based algorithm and data processing by monitoring process behaviors in real time;

  • Contributed significantly to reducing the failure rate from 12 to 5% by successfully applying the process control system to an automobile brake disc grinding manufacturing line.

Fig. 1
Architecture of industry 5.0 oriented smart manufacturing
Fig. 2
Quality failures in grinding process and the proposed method to improve the grinded part quality
Fig. 3
Steps for developing the digital twin system
Fig. 4
Analysis of sequence of the actions on the grinding machine
Fig. 5
Analysis for calculating power losses in belt transmission and in spindle during grinding process
Fig. 6
Calculation of the electrical loss and the mechanical loss during grinding process
Fig. 7
Steps for realizing the virtual grinding machine and mapping the operational process signal from the real grinding machine
Fig. 8
Embedding the physical characteristics into the virtual grinding machine
Fig. 9
Grinding force model and control model of the grinding process
Fig. 10
Results of simulating grinding force between DT and real system
Fig. 11
Deviation between the actual current value on the real grinding machine and the simulation values from the DT system
Fig. 12
Architecture of an autonomous grinding control system
Fig. 13
A monitoring system for the grinding process
Fig. 14
Integrating digital twin and cognitive agent for autonomous control system
Fig. 15
Tools for realizing the digital twin system to the shop floor
Fig. 16
Collecting experimental data for realizing the autonomous control system
Fig. 17
Initial ANN model and algorithm for optimizing the ANN model
Fig. 18
Optimizing ANN model
Fig. 19
Analysis of the grinding process parameters for determining the main process parameter for controlling
Fig. 20
Deriving out the rules for controlling grinding process
Fig. 21
Design of autonomous control system for stabilizing grinding process
Fig. 22
Implementation of the developed autonomous control system
Fig. 23
Applying the developed system to practice


1. Xu, X., Lu, Y., Vogel-Heuser, B. & Wang, L. (2021). Industry 4.0 and industry 5.0-Inception, conception and perception. Journal of Manufacturing Systems, 61, 530–535.
2. Piller, F. T.Nitsch, V.Luttgens, D.Mertens, A.Putz, S. & Dyck, M. V. (2022). How digital shadows, new forms of human-machine collaboration, and data-driven business models are driving the future of industry 4.0: A delphi study. Forecasting Next Generation Manufacturing: Digital Shadows, Human-Machine Collaboration, and Data-driven Business Models, Springer: 1–31.
crossref pmid
3. Adel, A., (2022). Future of industry 5.0 in society: human-centric solutions, challenges and prospective research areas. Journal of Cloud Computing, 11(1), 40.
crossref pmid pmc pdf
4. Nahavandi, S., (2019). Industry 5.0-A human-centric solution. Sustainability, 11(16), 4371.
5. Piller, F. T.Nitsch, V. & van der Aalst, W. (2022). Hybrid intelligence in next generation manufacturing: an outlook on new forms of collaboration between human and algorithmic decision-makers in the factory of the future. Forecasting Next Generation Manufacturing: Digital Shadows, Human-Machine Collaboration, and Data-driven Business Models, Springer: 139–158.
crossref pmid
6. Mittal, S., Khan, M. A., Romero, D. & Wuest, T. (2019). Smart manufacturing: Characteristics, technologies and enabling factors. Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, 233(5), 1342–1361.
crossref pdf
7. Nikolaev, S., Belov, S.Gusev, M. & Uzhinsky, I. (2019). Hybrid data-driven and physics-based modelling for prescriptive maintenance of gas-turbine power plant. In: Proceedings of the Product Lifecycle Management in the Digital Twin Era; 16.

8. Leng, J., Wang, D., Shen, W., Li, X., Liu, Q. & Chen, X. (2021). Digital twins-based smart manufacturing system design in Industry 4.0: A review. Journal of Manufacturing Systems, 60, 119–137.
9. Peres, R. S., Jia, X., Lee, J., Sun, K., Colombo, A. W. & Barata, J. (2020). Industrial artificial intelligence in industry 4.0-systematic review, challenges and outlook. IEEE Access, 8, 220121–220139.
10. Banitaan, S., Al-refai, G., Almatarneh, S. & Alquran, H. (2023). A review on artificial intelligence in the context of Industry 4.0. International Journal of Advanced Computer Science and Applications, 14(2), 23–30.
11. Sujatha, M., Priya, N., Beno, A., Blesslin Sheeba, T., Manikandan, M., Tresa, I. M. & Thimothy, S. P. (2022). IoT and machine learning-based smart automation system for industry 4.0 using robotics and sensors. Journal of Nanomaterials, 2022(1), 6807585.
crossref pdf
12. Vyskočil, J., Douda, P., Novák, P. & Wally, B. (2023). A digital twin-based distributed manufacturing execution system for industry 4.0 with AI-powered on-the-fly replanning capabilities. Sustainability, 15(7), 6251.
13. Viola, J., & Chen, Y. (2020). Digital twin enabled smart control engineering as an industrial ai: A new framework and case study. In: Proceedings of the International Conference on Industrial Artificial Intelligence; pp 1–6.
14. Soori, M., Arezoo, B. & Dastres, R. (2023). Digital twin for smart manufacturing, A review. Sustainable Manufacturing and Service Economics (pp. 100017..
15. Huang, Z., Shen, Y., Li, J., Fey, M. & Brecher, C. (2021). A survey on AI-driven digital twins in industry 4.0: Smart manufacturing and advanced robotics. Sensors, 21(19), 6340.
crossref pmid pmc
16. Attaran, M., Attaran, S. & Celik, B. G. (2023). The impact of digital twins on the evolution of intelligent manufacturing and Industry 4.0. Advances in Computational Intelligence, 3(3), 11.
crossref pmid pmc pdf
17. Onaji, I., Tiwari, D., Soulatiantork, P., Song, B. & Tiwari, A. (2022). Digital twin in manufacturing: conceptual framework and case studies. International Journal of Computer Integrated Manufacturing, 35(8), 831–858.
18. Qi, B., & Park, H. (2020). Data-driven digital twin model for predicting grinding force. In: Proceedings of the Conference Series: Materials Science and Engineering.
crossref pdf
19. Sharabov, M., & Tsochev, G. (2020). The use of artificial intelligence in Industry 4.0. Problems of Engineering Cybernetics and Robotics, 73, 17–29.
20. De Oliveira Santos, F., & Hahn, I. S. (2023). A systematic literature review and taxonomy proposition of machine learning techniques in smart manufacturing. Multidisciplinary Business Review, 16(2), 66–88.
crossref pdf
21. Febriani, R. A., Park, H.-S. & Lee, C.-M. (2020). A rule-based system for quality control in brake disc production lines. Applied Sciences, 10(18), 6565.


Hong-Seok Park Professor in the School of Mechanical and Automotive Engineering, Ulsan University. His research interests include intelligent manufacturing system, additive manufacturing, and CAD/CAM/ CAE.


Ngoc-Hien Tran Associate Professor in the Faculty of Mechanical Engineering, University of Transport and Communications. His research interests include intelligent manufacturing system and additive manufacturing.
Editorial Office
12F, SKY1004 bldg., 50-1, Jungnim-ro, Jung-gu, Seoul 04508, Republic of Korea
TEL : +82-2-518-2928   E-mail :
Developed in M2PI
Copyright © Korean Society for Precision Engineering.
Close layer
prev next