New Delhi: Stopping unplanned stoppages earlier than they’ll happen is the purpose of the sensible evaluation system being utilized in meeting at BMW Group Plant Regensburg. Predictive upkeep is proactive and preventive – and that is exactly what the sensible monitoring system affords.
Information-driven analyses of conveying gear permits potential faults to be recognized early and prevented – thereby sustaining optimum car manufacturing circulate. The substitute intelligence (AI)-supported system avoids a mean of round 500 minutes of disruption per yr in car meeting on the Regensburg plant alone.
Information evaluation: For the meeting at BMW Group Plant Regensburg, automobiles are typically hooked up to cellular load carriers or skid techniques, which cross by means of the manufacturing halls in a series. Any technical fault within the state-of-the-art conveyor techniques can deliver meeting strains to a standstill – requiring extra upkeep effort and thus leading to larger prices. To stop this from taking place, the innovation group at BMW Group Plant Regensburg has developed a system that may determine potential technical defects early – and thus keep away from any misplaced manufacturing. The conveyor parts affected may be faraway from the meeting line and repaired, away from manufacturing. The benefit is that the monitoring system doesn’t require any extra sensors or {hardware}, however evaluates current knowledge from put in parts and conveyor aspect management. An alarm sounds if anomalies are discovered.For instance, the load carriers used to move automobiles by means of meeting ship numerous knowledge to the service management system. This knowledge is then transmitted by way of the service and plant management system to the BMW Group’s personal predictive upkeep cloud platform. That is the place the evaluation begins: The algorithm continually searches for irregularities, comparable to fluctuations in energy consumption, abnormalities in conveyor actions or barcodes that aren’t sufficiently legible, which might set off a malfunction. If anomalies are discovered, the upkeep management centre receives a warning message, which it assigns to the upkeep technician on responsibility. “The surveillance displays at our management centre run 24/7,” explains challenge supervisor Oliver Mrasek. “This allows us to reply rapidly to any type of fault report and take the affected car out of the cycle.”Implementation: Predictive upkeep isn’t a standalone answer, stresses Mrasek. The system was standardised in collaboration with the BMW Group’s central shopfloor administration and different plant websites to facilitate swift and easy rollout to different BMW Group plant areas world wide. This method can also be cost-effective. “We don’t want any extra sensors, so the one prices are for storage and computing energy.”
Machine-learning fashions developed in-house have been additionally carried out within the system, which makes use of so-called heatmaps with numerous color codes for various abnormalities to visualise the mannequin’s findings. “This permits us to map completely different fault patterns in numerous parts and reply to them in a focused method,” explains Mrasek.
Based mostly on these sensible findings, the algorithms are constantly improved and refined. The group is at the moment within the strategy of connecting extra installations, optimising the system and integrating advisable actions into fault messages. The fault message might, for instance, point out related issues which have occurred in a system. This simplifies troubleshooting for upkeep technicians – for instance, if an impeller on the conveyor trolley is flawed. “Optimum predictive upkeep not solely saves us cash, it additionally means we are able to ship the deliberate amount of automobiles on time – which saves an enormous quantity of stress in manufacturing,” explains Deniz Ince, the group’s knowledge scientist.
Predictability – and two patents: Mrasek and his colleagues have been engaged on data-driven monitoring of conveyor expertise for the previous six years. In the present day, round 80 p.c of the principle meeting strains are already monitored on this manner. “We will’t detect or stop each single fault upfront, after all – however we’re at the moment avoiding a minimum of 500 minutes of downtime per yr in car meeting alone,” he explains. It’s simple to calculate how a lot this provides as much as. At BMW Group Plant Regensburg, a car rolls off the meeting line roughly each minute – each 57 seconds, to be actual – and the system is already being utilized in conveyor techniques on the plant websites in Dingolfing, Leipzig and Berlin.
The purpose is to additional exploit the probabilities of synthetic intelligence, with the system studying to estimate how a lot time stays between detection of the fault and the potential stoppage. This is able to assist technicians resolve how quickly they should carry out upkeep and permit them to prioritise, if wanted. Mrasek additionally sees additional potential in different areas of the plant: “We’re at the moment testing whether or not we are able to additionally use the system for the gear used to fill our automobiles with brake fluid and coolant, for instance.”
Though there are already quite a few choices for predictive gear upkeep, Regensburg’s built-in studying system is, to date, the primary of its variety. Compatibility with predictive upkeep is subsequently already being written into tenders for brand new conveyor expertise. Tools producers are additionally praising the system, since they profit from its evaluations as properly. The BMW Group has already registered two patents for its in-house growth.
