We notice on a day to day basis the importance of the continued availability of plant shop floor system. Continued availability of monitoring systems, data historians or process, sampling systems is of paramount importance and early identification of issues is pre-requisite for continued operation. Gone are the days when we used to measure the success of the company solely on the basis the production capacity. Now more and more organizations are looking for process capability, which is mapping the entire value chain in a real time view based on process digital twins. Such systems can help mapping the current and even historical maps for production environment giving the ability for organizations to optimize the processes. Initially IoT based digital twins are advocated for equipment performance assessment or even identifying anomalies in equipment functioning. Recently more and more emphasis is placed on extending such capability to complete value stream mapping and identification of process improvement. In my previous blogs I have given an overview of digital transformation covering various aspects of process optimization, user experience and legacy system modernization. In this blog I would like to cover the concepts of shop floor manufacturing systems and how Process digital twin help in transformation
Companies traditionally operate on applications and equipment which operate at an individual basis like tracking the plant historian data or assessment of laboratory information management systems or even operational performance dashboards which give an overview of data, issues or real time integration of laboratory data for detecting anomalies. Such systems are employed to production data acquisition, equipment or laboratory monitoring data and task or log management. Manufacturing shop floor offers unique challenges like for example a minute down time can result in huge impact to cost and quality. What is required now is developing capability of visualizing a holistic picture of end to end process scenarios including supply value streams. Though for many organizations it is an incremental cost to existing scenario, it should not be cancelled out but rather companies should adopt a careful approach of transitioning to future state by digitalization of process components and remove the impediments and obstructions on a real time basis..
The major technologies which act as drivers for the implementation of digital twin include the availability of processing power of computers, Wireless networks like upcoming 5G, IoT, Big Data and availability of advanced analytics. The major technological enables include Cloud technologies, Data Science, VR, AR, AI, IoT etc. Looking at the technological enables we can create a hypothetical 4 layered digital twin model like lowest level physical layer of sensors, devices, gateway systems at the shop floor, followed by availability of communication layer of necessary protocols, service providers. These two lower level layers are supported by Information layer which host the database for the storage of the data. Which can further be supported by additional layer of Visualization, processing, analysis algorithm including machine learning. For an end user these layers are obstructed by incorporation of business process layers. So, for an high level it could be data connectivity, data processing and business process presentation components.
Based on the architecture we discussed above we can identify necessary building blocks for the digital twins like Artificial Intelligence or Machine Learning capabilities to provide necessary intelligence and decision making capabilities, Virtualization models for visualizing the 3D asset visualization to form the replica of physical world, Sensor systems connection to physical world like IoT devices, AR/ VR. The architecture can also be extended with introduction of necessary functions or services which can be consumed like data validation or input data format or operability conditions.
Once a basic architecture is understood, lets try to see how we can go about implementing a digital twin for manufacturing Process. The basic approach suggested include identification of opportunities for implementation within the organization, identifying the right process for implementing digital twin, conducting a system pilot to identify criterial for success, updating the pilot with real time implementation and finally monitoring the process in real time basis.
There are various software systems which can be used for implementation of the model simulations like Experimental modelling like boundary physics models like meteorological modelling, physics based models which rely on advanced fluid dynamic models like Finite Difference Method, Finite Element Method, Finite Volume Method and Discrete Method using software systems like Ansys, Mathworks,, COMSOL, SIMSCALE, ANYLOGIC. Other software system-based implementation is via data driven modelling like Linear Regression, Support Vector Machines, ANN using systems like Tensorflow. The basic methodology of building a system include
- Setting up of process set up like reactor system
- Development of digital twin using simulation software like SimScape, Ansys, Mathworks etc
- Parameter tuning in the simulation software with data from the reactor systems
- Incorporation of cloud connectivity to real time reactor system
- Deploying the model either on premise or in cloud
- User Interface development for scenario management including Data Science methods.
Some of the adoption strategies for the Digital Twin in organizations include
- Digital Twin and analysis of the process data should not be confined to traditional databases, Companies needs to invest in databases specialized for data analysis
- When a piece of equipment is purchased, vendors should make an conscious effort to provide Digital Twin Prototype of the equipment, process or reactors as part of the purchase. It is known fact that the even the same equipment can be put to multiple uses.
- Digital twins need not be deployed only on customer centres, rather can be implemented / deployed in cloud systems, and should start doing pilots for the same
- The Manufacturing system not only have a Process data analysis but also infrastructure (Computer systems) management system for possible failure of the system components
- The Implementation of the Digital Twin should not be a big bang approach rather it is to be implemented for process areas which are critical for operation
- Industry wide efforts should be made to streamline the data quality, data portability issues for digital twins by standardization of templates, data quality metrices and other guidance
- All organization should start assessing the digital adoption questionnaire by validating various components like Process Management, Technology utilization and People & Organizational capabilities.
To conclude, there are various barriers for successful implementation of digital twins for manufacturing sector for managing the processes. Some of the issues which needed to be resolved include Technical capabilities of the people, Lack of standards and protocols, Cybersecurity issues, Cost to the organization and cost-effective technology. However, companies can start investing time to identify their key process areas and start conducting pilots with various physics-based models or data analytics to develop a road map for implementing the same in near future. For a company to adopt process level digital twin’s efficient introduction of process changes, adoptability to changing scenarios and revamping the technology landscape is also required. Such scenario will result in better customer service and increased product quality, cost optimization and streamlined information flow.