As well as revolutionizing how we deal with text and images, artificial intelligence (AI) also holds the potential to optimize engineering processes. Dr. Pouria Bigvand, Director of Product Management and Research & Development at AUCOTEC, explains how AI can assist with engineering of production plants. This is an area that he researched in his PhD thesis many years ago – and his focus on it has remained undimmed.
Since the hype surrounding ChatGPT at the end of 2022, AI Tools no longer just used by a small group of engineers but by the broad masses. Consequently, the opportunities offered by AI become even more visible – also for companies. This, of course, could be achieved by the application of generative AI, specifically the development of large language models (LLMs) on the basis of the almost unlimited amount of training data on the open internet. However, experts agree that generative AI is just a smaller subset of AI and in some specific fields the availability of training data is not as broad as published texts on the internet. In these fields, supervised machine learning (ML) is a strong approach parallel to LLMs.
Quantity and quality?
On this basis, says Bigvand, there are some use cases in engineering of production plants that can significantly benefit from usage of AI. To name a few:
- auto generation of data models and diagrams by LLMs, i.e. automatic generation of a data model of a component such as a heat exchanger without any predefined rules
- auto correction of data models and diagrams, i.e. suggestions for corrections to the data model based on similar tokens found by LLMs
- HAZOP, i.e. identification of constellations that can be considered hazardous based on supervised ML models
- migration of legacy documents using computer vision and ML models.
Turning PDFs into digital twins
© AUCOTEC AGDr. Pouria Bigvand, Director of Product Management and Research & Development at AUCOTEC
AI can also learn to “understand” diagrams. This means it can be trained to classify illustrated components in PDF or PNG files. AUCOTEC is tapping into this area of application to provide a unique form of support for projects seeking to migrate as-built plant documentation to the data-driven Engineering Base software. The aim here is to migrate all diagram types from process, I&C, electrical and hydraulics in non-machine-readable forms such as PNG or PDF and generate a data model in parallel to PDF files with hot spots for navigation purposes. “Having said that,” continues Bigvand, “experienced experts are still needed to carry out reviews and improve the AI model. Several rounds of corrections and fine tunings are required per data set.”
The training is worth all the effort, as it will then be possible to create object models very efficiently from the plant documentation, which is decades old. Doing so will greatly facilitate maintenance and revamp work. “Although much of this is already possible for machine-readable formats such as DWG in Engineering Base, AI will bring new life to documents that have long been considered dead,” says Bigvand.
More than just searching
At present, when engineering tool providers talk about their tool’s AI features, they are usually just referring to an advanced search function that is able to process large amounts of text and provide reasonable answers, lists or components. However, we at AUCOTEC believe AI is capable of much more in the world of plant and machine engineering when “structured” data is available and on a large scale. Plus, as Bigvand says: “Being able to capitalize on the capabilities of AI will be crucial if you are going to stay competitive.”
Engineering Base would appear to fit the bill perfectly here: open for integration, data centric across different disciplines, and exceptionally consistent and transparent too.
