Diagnosing a failure without understanding the machine is like looking for a leak in the dark. Chatbots and PDF search engines read the documentation, but they have no idea how the components fit together. Mimorian is an industrial intelligence platform that models equipment from its diagrams, structures failure diagnosis and captures the know-how of maintenance teams through a multi-agent AI architecture. This modelling is the precondition for a structured diagnosis. Here is how it works.
What is a functional digital twin in maintenance?
A functional digital twin is a map of the relationships between the components of a piece of equipment: which drive powers which motor, which sensor monitors which valve, which safety device cuts off which line. It is a model of the links: who depends on whom, who acts on what, whereas a CMMS is limited to an asset inventory.
This difference changes everything for diagnosis. An inventory tells you that the machine contains a drive, a motor and a turbine. The functional twin tells you that the drive controls the turbine through the motor. When a fault appears, it is this chain of dependencies that lets you trace back to the cause.
How does AI extract a relational graph from an electrical diagram?
The starting point is what the plant already owns: its electrical diagrams, often in PDF or DWG format, sometimes as a scanned image. The AI ingests them, recognises the standardised symbols (the IEC 60617 standard describes these symbols), identifies the connections between components, and builds a graph: each component is a node, each electrical or logical connection is an edge.
Let us take a concrete example. A variable speed drive controls a turbine, and the motion passes through an intermediate motor. One day, the turbine shows a fault. The visible cause is the turbine. But by following the drive, motor, turbine chain, you can see that the real culprit may be the drive upstream or the motor in the middle. Without a model of the links, a technician under pressure replaces the turbine and the failure comes back. With the graph, the AI traces the chain and suggests examining the upstream component first. The visible cause is not always the root cause, and that is precisely what the graph reveals.
Why a graph beats a keyword index
The clearest comparison is a GPS against the index of a book. An index tells you on which page a word appears. A GPS knows how the roads connect, so it works out a route.
A search engine running on the documentation works like an index: it retrieves the pages where "turbine" is mentioned. Useful, but it does not know that the turbine depends on the motor, which depends on the drive. The relational graph, on the other hand, knows these connections. For diagnosis, the consequence is direct: you trace back a chain of cause and isolate the component responsible, instead of churning through text in the hope of hitting the right page.
What gains on the ground: guided diagnosis, traceability, knowledge capture
Three concrete benefits for teams.
Guided diagnosis: the AI puts forward hypotheses ranked by criticality and by dependency, and the technician confirms or rules them out. They stay in control, but they no longer start from scratch.
Traceability: every intervention enriches the graph with a component, failure mode, corrective action link. The knowledge does not stay in one person's head, it is deposited in the model.
Knowledge capture: field know-how becomes searchable. The departure of an expert no longer erases their reasoning, because that reasoning has been structured failure after failure. The industrial sites that genuinely benefit from AI are those that anchor it in their existing data and processes [McKinsey, 2023].
Conclusion
The functional digital twin is the precondition for a structured diagnosis. Without a model of the machine, a maintenance AI remains a PDF chatbot. With one, it becomes able to trace back a root cause chain.
Mimorian combines three layers of value: structural intelligence (the functional digital twin and hybrid search), the orchestration of specialised agents for the field and the expert, and a virtuous circle where every intervention enriches the plant's collective memory.
To go further, read our guide to the functional digital twin for maintenance. For the full framework of what trustworthy AI in industrial maintenance involves, see our complete guide to trustworthy AI for industrial maintenance.