Context
In the road and rail transport sector, safety is a major concern. Leaf springs are critical components of suspension systems, used in particular on trucks and trains. A broken spring can lead to a loss of vehicle stability, or even serious accidents that endanger passengers, drivers, and the goods being transported.
The main challenge lies in the ability to detect the first signs of failure before the spring breaks. Preventing such breaks is therefore a critical issue. Detecting the first signs of weakness before a failure occurs significantly improves safety while optimizing maintenance costs.
Needs
The customer wanted to monitor the condition of the leaf springs in real time and be able to anticipate their failure by analyzing the vibrations recorded just before the braking phases.
To do this, two accelerometers were installed on either side of the blade. However, these sensors were not synchronized, which complicated data exploitation. The customer therefore needed a reliable health indicator capable of reflecting the condition of the spring and enabling quick and relevant decision-making.
Solutions
DiagFit has enabled the implementation of a fault detection solution based on the analysis of raw vibration data, without requiring a physical model of the system.
The data from the accelerometers was first pre-processed to correct problems related to their lack of synchronization. Next, a diagnostic model was built based on characteristics automatically extracted from the signals. This model made it possible to categorize the condition of the springs into three levels:
- Green: spring functioning normally
- Orange: abnormal behavior requiring inspection
- Red: critical failure; immediate vehicle shutdown recommended
This approach enables continuous, online monitoring without human intervention, which can be directly used by maintenance teams.
Results
The solution developed by DiagFit delivered remarkable results, demonstrating the relevance of its approach. During testing, the model accurately detected all faults: the two defective springs were identified without error, achieving a true positive rate of 100%. Similarly, the five springs in good condition were all recognized as such, guaranteeing 100% true negatives. These results confirm the reliability of the system, even in a complex environment characterized in particular by sensor desynchronization.
Beyond these figures, the model also stood out for its ability to detect anomalies upstream, even before visible symptoms appear. This early detection paves the way for proactive maintenance that is more efficient, safer, and better targeted.
For the customer, the benefits are numerous: enhanced safety through the anticipation of breakdowns, better planning of interventions, a significant reduction in unplanned downtime, and overall optimization of maintenance operations. Finally, the robustness of the model, which is capable of operating with unsynchronized sensors, demonstrates the solidity and adaptability of the DiagFit approach, which is perfectly suited to real industrial constraints.
This use case clearly illustrates DiagFit’s ability to transform raw vibration data into actionable decision-making information, helping to enhance the reliability and safety of transportation systems.
