Ski Lift Monitoring System: Structural monitoring of ski-lifts

Anomalies in ski-lifts caused by faults such as misalignment in the sheave train or wear in the sheave wheels can affect the ski-lifts’ nominal operation and human safety. Therefore, the monitoring of the ski-lifts, and the detection of faults at early stage is crucial.

The monitoring of ski-lifts is complex due to large mechanical structures with many critical components. Currently, the monitoring is conducted via Non-Destructive Testing (NDT) methods with the most used one being the manual visual inspection. These methods are error-prone, costly, and time-consuming and they may lead to safety risks and operational inefficiencies due to failure in detecting faults at an early stage. Moreover, no alternative methods have been investigated.

Lawo-Lab AS has developed a self-sufficient automated and remote monitoring system for ski-lifts, the Ski-Lift Monitoring System (SLMS) which is based on vibration data acquired via sensors. The effectiveness of the SLMS has been investigated during a past research project with test data acquired by sensors in a single mast and only for a few hours.

In the current project, the new improved version of the SLMS system was developed and tested, with the new version being based on an automated machine learning method for an effective and remote monitoring of any anomalies in the ski-lifts.

During the project, a wireless sensor network was initially set-up with one sensor per lift and effective and reliable mechanisms for the acquirement, transmission, storage of data for a full season from all the masts.

Then the acquired signals will be processed by checking the lifts’ behavior via spectrums and spectrograms.

Finally, the machine learning method was applied for monitoring each ski-lift’s structural state and characterizing the state.

The method was equipped with multiple statistical AutoRegressive models based on data from multiple sensors.

The models describe the ski-lifts’ behavior under varying operating conditions and features from the models were used for detecting anomalies.