Why invest in AI for quality control? - A sample calculation.

Episode 6

Author: Hanna Nennewitz

Why invest in AI for quality control? - A sample calculation.

Episode 6

Author: Hanna Nennewitz

In the last episode, Peter Droege, CEO and co-founder of Maddox AI, explained what the biggest problems of visual quality control are in the industry. The insight was that AI-based systems are a good option to address these issues. But what about finances? Is it even worth investing in an AI-based system? These are the questions I turn to in this episode. Once again, Peter Droege can give me the answers to these questions.

Peter explains in our conversation that there are five financial levers that can be turned with the help of AI-based systems, such as the Maddox AI system. The levers are as follows:

saving personnel costs,
minimizing returns / complaints,
increasing traceability,
preventing excessive scrap rates in production and
process improvements.

1. Personnel Costs

As noted in the last episode, manual inspections in visual quality control are especially a major issue in high-wage countries like Germany. At the same time, the shortage of skilled workers makes it hard to find people who even want to do such a strenuous job. “One of our customers,” Peter reports, “had seals inspected manually in shift work. After installing the Maddox AI system, they were able to significantly reduce the number of those who had to work in quality control in order to redeploy labor elsewhere. With a single Maddox AI system, nine employees can now turn their attention to more valuable tasks.” Savings, of course, always depend on the specific use case, in the case of said customer they were between 200-250 thousand euros due to savings in labor costs.

2. Minimization of Complaints & Returns

Returns in connection with the necessary re-sorting actions can quickly become very expensive. Alongside this, incorrectly shipped parts are bad for customer relations. With an AI-based system in visual quality control, the likelihood of receiving returns can be significantly reduced. This is mainly because an AI-based system inspects more accurately and consistently than a human. One Maddox AI customer received two complaints in the year before the Maddox AI system was installed. The complaints and re-sorting work alone generated costs of a quarter of a million euros. Since the installation of Maddox AI and the resulting optimized control of the parts, our customer has received no further complaints, which saves relevant costs.

Peter tells me about another customer who manufactures tractors, some of which are highly customized. The tractors are shipped all over the world. Before the installation of Maddox AI, a wrong configuration in the production in Germany was not noticed from time to time. If the tractor was then shipped to the USA, for example, and the incorrect configuration was only noticed there on site, it was very costly to correct this error. Since the manual inspectors have been supported by the Maddox AI system, the number of customer complaints has been significantly reduced. Each saved error usually directly saves a five-digit amount.

3. Traceability

However, should a return occur despite AI-based quality control, Maddox AI customers also have the advantage that each part can be individually tracked. “For example, if I have a serial number or a QR code, I can track a part one-to-one. If a complaint does occur, I can use an image to prove that the disputed part left the factory without defects,” Peter explains. The documentation therefore offers further protection against high reclamation costs.

4. Preventing Excessive Scrap Production

Digitalization, often accompanied by the installation of AI-based control systems, can also prevent excessive scrap production. Via SMS or mail, one can be warned by the Maddox AI system should the scrap rate exceed a certain threshold. Peter told me about a customer who produces several parts per second. Before the installation of Maddox AI, only random quality checks were carried out every one or two hours. As a result, it was only noticed very late if the production was faulty and, in the worst case, the parts produced in the last one to two hours had to be identified and disposed of at great expense. By installing the Maddox AI system, the shift supervisor is now automatically notified by SMS when the scrap rate becomes too high. Maddox AI acts as an early warning system in this case and helps not to waste resources unnecessarily.

5. Process Improvement

Another adjusting screw that goes hand in hand with digitalization is the possibility of process improvement. What does that mean in concrete terms? Peter explains it this way: “Through Maddox AI, we know where and how often defect types occur on a component. If I now see, for example, that the same defect always occurs in the same place at a certain time of day, then I know that something has gone wrong in the production process.” For example, one Maddox AI customer produces using injection molding. The molds used must be cleaned regularly, otherwise rubber residues or other contaminants build up. If these residues are not noticed early enough, only defective parts are produced. Maddox AI automatically provides structured root cause analyses in the form of defect concentration analyses that detect structural defects at an early stage. With the help of these defect concentration analyses, process improvements can be initiated that directly address the cause of an increased defect rate.

The use of AI-based systems offers you many opportunities to reduce costs in production. If you would like to take advantage of these opportunities as well, feel free to contact us. Our team will be happy to take the time to evaluate the possible use of Maddox AI in your company. Now that I’ve covered the problems in visual quality control and the financial benefits of using AI-based systems, I’ll have a more detailed look at the problems discussed with Peter in episode 5. In the next episode, I’ll try using the Maddox AI system and a rule-based system myself.

Article

Usability - Can I also Operate Maddox AI?

Episode 7

In this episode, Hanna Nennewitz tests the usability of Maddox AI by trying to create a quality control model by herself. In comparison, she tries to programme the same testing task with a rule-based system.

Read the article

All blog posts