The challenge

It is essential for medium-sized part suppliers to identify production errors early on. If defective parts are subjected to the random sample-based final inspection or delivered to the OEM or end customer, this not only leads to high warranty costs, but can also undermine confidence in the supplier and lead to the withdrawal of the order. The most important characteristic to measure and improve in this context is reliability: how consistently a product or system performs under the required functions without failure under stated conditions for a specified period of time.

It is subject to a random process and cannot be measured directly – a challenge for small-sized manufacturers like MaTec as well as mass producers like KLS. Both have to identify, save, and analyse all relevant data during production. Saving these thousands to hundreds of thousands of parts with simultaneous recording, storage, and evaluation of relevant parameters over a longer period of time requires extensive storage space. Automated evaluations with stochastic methods, artificial intelligence, and machine learning processes require high computing power. Hence, large companies have a significant competitive advantage here: As a rule, these challenges are taken on by the internal data centres or by IT service providers working exclusively for the company with their own hardware or hardware that is rented on a long-term basis. An HPC-based solution would also give small to medium manufacturers the chance to professionally forecast and influence reliability.

The solution

The solution, called HiPerRel, implements applications for observation, calculation, and reliability improvement for metal and rubber parts manufacturers. The SaaS platform “Plexalytics” is used as a basis and operated externally by a technology provider. The more data are collected from the machines, the more accurate the results will be. However, the required computing power also increases. SMEs usually do not have their own data centres. To close this gap, an external High-Performance Computing centre is used to statistically check and apply the relevant measures that increase the reliability of the products and reduce the probability of failure. To obtain results the use of HPC allowed delivery of reliable info overnight rather than after several days, enabling production parameters to be adjusted promptly if needed.

Business impact, Social impact, Environmental impact

The connection of the Plexalytics Search Based Application platform to the HPC architecture provides the first business model with which not only large corporations but also small to medium-sized series manufacturers can be served.

The result of the experiment helps small to medium metal and rubber parts manufacturers recognize errors and their influence in a very early phase of the production process, statistically check and apply the relevant mitigation measures to increase the reliability of the products, and reduce the probability of failures. KLS and MaTec will experience substantial future gains as a result of these advancements, leading to improved reliability of their parts and, consequently, increased customer satisfaction.

The capacity for early calculation of reliability reduces the number of failed parts. This not only improves the quality assurances of the manufacturer SMEs but also achieves substantial savings in material and energy resources which bring contributions to carbon reduction and improved use of natural resources.

Benefits

During the experiment, a wide range of potential benefits were identified. The following points give a concrete overview:

• Using HPC and Plexalytics within service-based software enables overnight calculation of parts reliability models, instead of taking 8 days. This swift analysis allows for quick interpretation of failure causes and adjustment of production parameters within a day.
• KLS and MaTec report an approximately 15% reduction in the long-term part failure rate due to the application of HPC to Plexalytics.

• Due to the application models for reliability predictions in highly-automatised serial production as well as in small-batch manufacturing, the technology provider will expand its customer base by approximately 25% within the next three years.

Organisations involved:
End User: KLS Ljubno d.o.o. 
Technology expert: MaTec Gummiwerk GmbH
HPC provider: Pumacy GmbH