Context
In the steel industry, where equipment robustness and precision are essential, hydraulic actuators play a key role in the smooth running of production lines. On a sheet steel welding line, these hydraulic actuators are subject to high stresses and must operate reliably to guarantee continuous, high-quality production.
However, these actuators can be prone to a variety of maintenance problems. One of the major risks is oil leakage, which can have far-reaching consequences. An undetected leak can lead to a reduction in actuator performance, causing anomalies such as :
- Fire hazard due to the presence of oil in a high-temperature environment.
- Tool slippage, compromising the precision and quality of welds made on sheet steel.
- Poor weld quality, which can affect the strength of assemblies and necessitate costly rework.
- Production line stoppage, impacting overall output and leading to financial losses.
To anticipate these failures and optimize equipment maintenance, cyclical data from the actuators was collected. This data, structured in 11 columns, enables real-time analysis of the condition of the hydraulic cylinders and identification of early warning signs of failure. Thanks to this monitoring, predictive maintenance can be implemented, reducing unplanned downtime and extending equipment life.
In a sector as demanding as the steel industry, where any malfunction can lead to major disruption, the intelligent use of this data is a strategic asset for ensuring smooth, efficient production.
Need
Detecting and anticipating failures using actuator time series is a strategic requirement for many industrial companies, particularly in demanding sectors such as the steel industry. By exploiting cyclical actuator data, companies seek to better understand equipment behavior in real time, identify the early stages of failure, and intervene at the right moment. This ability to predict failures not only reduces maintenance costs, but also improves safety, production quality and overall plant performance.
Solution
DiagFit generated a predictive model from the actuators’ time series, providing a concrete response to the company’s predictive maintenance challenges.
DiagFit makes it possible to predict breakdowns before they have an impact on the line. It uses advanced algorithms to detect anomalies, generate alerts and provide maintenance teams with the information they need to intervene at the right time.
By transforming raw data into usable, predictive information, DiagFit becomes a genuine decision-making tool, reducing unplanned stoppages, extending equipment life and securing industrial processes.
Results
DiagFit detected 100% of cycles labelled as unhealthy, demonstrating the reliability and accuracy of its predictive model. In addition to this performance, the software also identified an additional cycle not initially labelled as unhealthy, but with similar characteristics to the faulty cycles.
This proactive detection suggests an enhanced ability to anticipate, to spot faint signs of degradation even before a failure is formally noted. This type of detection is particularly valuable in critical industrial environments such as a welding line, where every early warning can prevent production interruption or quality degradation.
These results confirm that DiagFit doesn’t just reproduce existing labels, but goes beyond them to provide predictive added value, essential for an effective preventive maintenance strategy.
