Providing high-availability and high-performance systems is a top priority for any Manufacturing company. Reducing planned and unplanned downtime is an important part of this, since each minute of downtime is a loss in productivity and revenue.
One of our customers provides material handling systems (MHS). These systems are running 24/7 and every minute of downtime has a large impact on the supply chain.
Their question: can predictive maintenance techniques reduce our downtime?
Together, we brainstormed if there were parts of the MHS with visible traces of wear – since in that case we could use visual analysis techniques. Resulting in the focus on a particular part of the MHS: the shipping sorter.
This shipping sorter enables –through sorting– that products are distributed to the right location.
It consists of roughly four hundred rail-based induction-powered carriers that transport and deliver parcels to the right conveyor locations in the warehouse. These carriers provide on-the-spot sorting and delivery by integrated belts.
When you step into the warehouse the first time, it looks like a large roller coaster for parcels. The high-speed carriers are running on rails and use four running and guiding-wheels to keep the carrier on track.
These wheels and carriers are prone to experience wear and tear over time.
This wear can lead to unexpected downtime. In addition, also planned downtime takes longer than wished for, since it requires a maintenance support engineer to be onsite. On site, the engineer actually needs to inspect the wheels visually and measure the size of the wheels using a caliper.
To automate this process and predict the wear and tear, we (co)developed the “Carrier Health” system. The goal of this AI-based predictive maintenance solution is to automatically detect wear in the components, and thus making manual inspection obsolete and reducing future downtime.
The first step consists of a laser sensor that measures the height of the carrier in relation to the track. When the wheels wear off, the carrier will come closer to the track and lose height. This can cause the magnets on the carrier to come too close to the induction coil, which can lead to an emergency stop of the system to prevent collision damage. The emergency scenario is a practical example of unplanned downtime.
The second step is placing four advanced cameras with their own backlighting surrounding the track. These cameras continually take images of the running- and guiding wheels of the carriers when passing by. The images are ingested in a data platform with an AI engine that applies corrections. By visually determining the wheel position on the track, the AI engine calculates the wheel dimensions with an accuracy of 0.15mm.
A great improvement on accuracy and coverage compared to previous manual caliper measurements which only covered about 10% of the carrier wheels, and left room for human error.
The third step consists of a vibration sensor placed on the track. This sensor detects vibrations caused by worn out bushes and/or bearings, and sends the data to the platform to be analyzed. Here, the vibration data is linked to a specific carrier by analyzing the vibration curve in the data in relation to the carrier’s position in time.
All of this data from the laser sensor, the cameras, and the vibration sensor is collected, processed, and combined into heatmaps on sorter, carrier, and component level. With that, the system determines the wear and tear on wheels.
It consequently suggests the site reliability engineers to intervene in time by quickly pinpointing components requiring maintenance. In addition, the maintenance planner can fully use and prioritize the upcoming maintenance windows or plan emergency repairs to prevent unplanned downtime.
So yes, the predictive maintenance techniques used in the “Carrier Health” system are actually reducing downtime.