The LiU-ICE Benchmark -- An Industrial Fault Diagnosis Case Study

TL;DR

The paper introduces LiU-ICE, an industrial fault diagnosis benchmark built around a realistic four‑cylinder internal combustion engine air path, addressing the challenge of scarce faulty data and model inaccuracies in fault detection. It provides a state‑of‑the‑art mean‑value engine model with $94$ equations (including $14$ differential constraints), $90$ unknowns, $10$ knowns, and $4$ fault variables, formulated as a differential‑algebraic equation (DAE) and implemented in the Fault Diagnosis Toolbox. The data compendium comprises $25$ WLTP‑based driving cycles at $20$ Hz, including $2$ fault‑free and $23$ faulty scenarios, with faults injected around $120$ seconds in each dataset and covering sensor faults ($f_{ypic}$, $f_{ypim}$, $f_{yWaf}$) and a leakage fault ($f_{iml}$). Structural analysis using the Dulmage–Mendelsohn framework demonstrates that all faults are structurally detectable and isolable, with a redundancy degree of $4$ enabling multiple over‑determined residual generators. The benchmark, already used in Safe Process 2024, is publicly accessible via gitlab and supports rigorous evaluation of fault diagnosis methods under industrially realistic conditions, bridging theory and practice for residual generation and diagnosability research.

Abstract

This paper presents the LiU-ICE fault diagnosis benchmark. The purpose of the benchmark is to support fault diagnosis research by providing data and a model of an industrially relevant system. Data has been collected from an internal combustion engine test bench operated in both nominal and faulty modes. A state-of-the-art model of the air path through an internal combustion engine with unknown parameters is provided. This benchmark has previously been used in a competition at the 12th IFAC Symposium on Fault Detection, Supervision and Safety for Technical Processes (Safe Process) 2024, Ferrara, Italy.

Figures (7)

• Figure 1: Picture of the engine test bench.
• Figure 2: Schematic of the air path through the engine (used with permission from eriksson2002control).
• Figure 3: An example of the model implemented in Matlab.
• Figure 4: A structural representation of engine model.
• Figure 5: Dulmage-Mendelsohn decomposition of LiU-ICE model.