We are the Cliff.
The challenge: to anticipate the defects in the electric power lines by creating a predictive maintenance system for faults.
At the SEAT plant in Martorell there are several workshops. The cars have to go through them in different routes or circuits. Each route has a track for the transport of cars which is electrified and is called electrovia.
Cars are each pulled by a trolley or octopus hanging from the electric tracks. Trolleys travel at different speeds and may be stopped for a period of time at certain points. There are also parking lots.
Phase 1: determining the defects in the electric ways
Five defect classes were defined:
– Protective
– Copper
– Position
– Thickness
– Lift
Phase 2: creating a predictive and automated inspection system
The trolley must send, as continuously as possible, data to an equipment called “ground” which will store them and allow their exploitation.
For the inspections (runs through a specific electrovia), a filter is presented that allows to select them by date range, electrovia and type of defect. The result of the filter is a table with a line for each inspection. When the user clicks on one of them, a scale plan of the SEAT plant is displayed, with the corresponding electrovia highlighted in blue (configurable). For each defect a dot is drawn on the plan in a color indicating the class. On the right is listed the list of defects detected in the inspection with data such as section, type, position and QR code.
The model consumes them, draws conclusions about the state of the electric way, according to thresholds, decides whether they are defects or out-of-range measurements, and generates a database for each of the videos. These databases contain annotations for system control, defects, out-of-range measurements and change of parameter thresholds. Our replicator detects new databases and sends the information to the ground equipment.
A scalable impact on the entire network of associated garages
Short-term:
The solution enabled the detection of faults and failures with a reliability of 99.9%, and 50% efficiency in the average verification time. Based on IoT sensorization, a highly reliable process analysis and monitoring system was created in the market.
In the medium term
The preventive maintenance models in inspections and failure predictions, contributed to a reduction in maintenance costs by +80% and a ROI on investment in machinery of +20% over the past situation. This artificial vision model will be scaled up to the rest of the workshops within 3 years.
