Autocare diagnostics official website

About the project

In this project, the primary objective has been to design an AI-based remote diagnosis system for heavy vehicles. The project is part of the CDIO-Project course TSRT10 at Linköping university and the group has worked with the project from a planning phase, through development and into implementation. The final deliverable constitutes a proof-of-concept AI-driven Remote Diagnosis Solution designed to illustrate the detection and isolation of various faults within the Selective Catalytic Reduction (SCR) system utilized in heavy-duty trucks. This demonstration is facilitated through an interactive user interface (UI).

System

Exhaust system

With an already developed model of the SCR system, this project has extended the the model with an exhaust system. The main focus for the exhaust system is to detect sensors fault and urea clogging.

Fault detection

The fault detection main assignment is to simulate residuals for both the SCR-system and the exhaust system. The residuals are then evaluated in a decision matrix to get a preliminary residual based diagnos. In case of an alarm, the data is packaged and sent to the fault classifier.

Remote connectivity and Fault classification

The fault classification's main goal is to provide a secondary diagnosis based on the initial fault detection, ensuring thorough fault isolation and assigning a certainty score. A machine learning model serves as the fault classifier, enhancing diagnostic precision and efficiency. Data from fault detection is transmitted to the classifier via the remote connectivity subsystem.

The system overview depicted in the image illustrates a promising proof of concept for remote fault detection and diagnosis within the SCR-system. This solution seamlessly combines the analysis of sensor data, model predictions, and cloud-based fault classification. By leveraging this integration, the system not only enables the remote identification of faults in the SCR-system but also lays the groundwork for potential scalability and enhancement.