Equipment diagnostics for hard-down failures

TNO-ESI develops a model-based digital assistant for diagnosing hard-down failures, i.e., failures that are consequence of hardware faults: Operational Diagnostics. The assistant computes the most likely hardware components to be broken, given the current observations. These observations include automated observations, such as timeseries and event log data, and the results of tests performed by the service technician. Moreover, the assistant advises on which test to conduct next, by balancing the value a test could bring and the cost associated to conducting the test. As such, the methodology assists service technicians in their search for the root cause of an issue, reducing the need for extensive expertise.

The methodology uses a probabilistic model that is derived from the system design, usually formalized as a Model-Based Systems Engineering (MBSE) model, containing the following concepts:

• System function breakdown
• System function inter-dependencies
• Functional deployment on hardware

The MBSE model is augmented with observability and reliability information, such as events that can be logged, tests that can be executed and hardware fault rates.

Moreover, the model can be used during system design to assess the future observability of hardware faults with the given sensor and test configuration: Design for Diagnostics. This assessment can be used to add, move or remove sensors and tests to reach the desired diagnosability of the system.

This methodology has been implemented in MBDlyb, our Python library for model-based diagnostics

• MBDlyb
• Carefree MBSE report
• SD2Act report