MBDlyb

Python library for model-based diagnostics

MBDlyb is an open-source library for design-based diagnostic reasoning in high-tech systems. Developed by TNO-ESI in collaboration with ASML and Canon Production Printing, it formalizes functional dependencies to support fault diagnosis - both during system design and operation:

  • to gain insight in failure observability and make early design changes to reduce the diagnostic efforts during operation.
  • to assist a service engineer in identifying the root cause of a system failure.

What Makes MBDlyb Unique?

Unified Knowledge Base

The use of a single knowledge base for design for diagnostics and operational diagnostics.

Intelligent Test Guidance

Step-by-step assist a service engineer by showing the most likely faulty components and the most promising diagnostic tests by balancing expected information gain from conducting the diagnostic test with the cost of conducting it.

Automated Diagnostic Trees

Generate a-priori diagnostic procedures in the form of decision trees for fault diagnosis.

Graphical models

Structured transformation of function-based knowledge to probabilistic graphical models, such as Bayesian networks or Markov networks.

Graph database

Persistent storage of the knowledge in a graph database.

Capella Model Import

Functionality to import a Capella system architecture model.

Integrated Web Workspace

A web interface that integrates the model editor, analysis tool and diagnostic assistant.

Why it matters

Use MBDlyb during system design to improve failure observability and reduce future diagnostic costs. During operation, it helps service engineers identify root causes faster and more reliably.

Open collaboration

MBDlyb is actively being developed and has been open-sourced under the EPL-2.0 license to foster innovation across industry and academia.

Further information