Experiment 8: DIgital Twin for Agile Changes (DITAC)

Motivation of the Experiment

From the perspective of the end-user, the aim is to have an available infrastructure for Virtual Commissioning, with the flexibility to be used also in the design of custom machines. As a matter of fact, currently, at Restart, no one tool is in use for simulating what has been designed, and the options for a complete development environment (capable of integrating the mechanical-pneumatic part and the software part) do not present easy-to-approach architectures.

The reasons beyond the usage of a Digital Twin can be listed as:
Reducing commissioning interval: “testing” and commission is now performed at the customer's production site, also including travel costs for personnel who could carry out the activity in the office.
Reducing system downtime regarding after-sales changes: as in the previous point, with the aggravating circumstance that the costs may involve production stops for the customer and / or activities outside the normal production activity (nights and weekends).

Optimizing the usage of resources, including also hardware and materials, avoiding the use of material erroneously estimated in the design phase. Furthermore, the aim of the project includes the design and testing of procedures and methodologies for faster development of digital twin, facilitated by a digital twin ecosystem. Moreover, the project will represent the opportunity for experimentation of data analysis through machine learning, supported (in case of lack of data) by “artificial” data eventually provided by the digital twin. For what concerns the ML sub-experiments part, following a phase of exploratory data analysis (EDA), a series of models will be trained to make predictions on the future state of the Restart Automation production line, also exploiting for the model training the data coming from the digital twin developed by IDM-Systems.

Purpose of the Experiment

The main goal of using digital twin-based technologies in Restart Srl is to reduce the efforts required for after-sale (on-site) modifications, with a significant improvement in service and a greater interest from new customers. The ultimate goal is to extend the company's range of action, through greater optimization of jobs that require a business trip for on-site modifications.
The main benefits provided by the introduction of a Digital Twin solution will be on the reduction of uncertainty in the case of changes after the realization and putting into production of the machine, improving:

  • the amount of work “on-site” (fewer hours needed)
  • the amount of machine downtime necessary (fewer hours needed)
  • the overall quality of the process (less pressure during changes, since not performed “on-site”)
  • find a non-trivial correlation between production phases regarding quality, providing earlier scrap detection (resource-saving) and better product quality.

From the perspective of software used, it will include both instruments for digital twin and machine learning development:

  • Siemens Tecnomatix Process Simulate
  • Siemens STEP7
  • Siemens TIA Portal
  • Siemens SIMIT
  • Pytorch
  • Fast.ai
  • Pandas, scikit-learn and matplotplib
  • Utilize in the cloud:
  • Possibility to reach ML models onsite and separate ML model training

Technical Impact

Regarding the development of digital twins, the main goal is to develop a stable, robust process and framework completing the current processes and using an existing, available market lead simulation software solutions (Siemens Tecnomatix Process Simulate) to create a fast, reliable and cost-efficient digital twin.
The target is to modularise IDM framework/method to provide a future-proof and easy-to-use solution for similar projects. The reason that these solutions are modularized and standardized makes the extension of the library easy. Therefore, creating another digital twin, even a different type of Industrial Product, is possible and the method for developing and integrating it into the framework will be already established.

Moreover, the consortium will also evaluate the analysis of data provided by the machine through a Machine Learning approach, including the use of Deep Learning (DL) models such as Convolutional Neural Networks (CNNs) or Long-Short Term Memory (LSTM) networks if classical ML models (e.g. Gradient Boosting methods) are not yielding satisfactory results. In this context, the main challenge is to analyse the data stream generated from both the real and the virtual machines.

Expected Economic Impact

From the user point of view, the main economic impact is expected in the improvement of after-sales services; in particular, requests of changes after installation will be fulfilled more easily and at a lower cost, supported by a less uncertainty for modifications and a reduction of downtime for changes after installation. As a matter of fact, in the Restart case, after-sales requests for changes from the final user (e.g. due to necessities of production of a similar model of the same product) are not isolated and they request significant work on the machine, causing interruption of work/production.

The ultimate goal is to extend the company's range of action, through greater optimization of jobs that require a business trip for on-site modifications.

From the point of view of ISV, two new revenue streams are expected:

    • digital twin as-a-service
    • machine delivery or brownfield modification where the digital twin approach could give significant technical advantages

As for the research part by CNR-IMATI, the expected immediate economic impact is to allow savings for Restart Automation S.r.l. in terms of saved parts that are marked as scrap and need to be reworked or are permanently lost. The hope is that, in the long run, this can lead to significant savings for Restart.

Project Partners

 

Restart Automation Srl.

is the experiment
end-user.

IDM-Systems Zrt.

serves as the
infrastructure provider.

CNR-IMATI

serves as the data model
& algorithms provider.