Presentation at the Innovation Forum Medical Technology on 14 October 2020 in Tuttlingen on the topic of AI in SMEs

On October 14, the Medical Technology Innovation Forum took place for the 12th time. LA2 was once again represented this year with a booth and a presentation. The presentation was tailored to the forum’s thematic focus—artificial intelligence—and focused on LA2’s latest AI project, which was carried out in cooperation with reference customer DANmed.

After a brief clarification of terms, the presentation addressed the project’s objectives and challenges. The goal was defined as reducing waste in the production processes of the CNC department – ​​the department that generates the largest amount of waste at DANmed. The greatest challenge was the small number of work steps (approximately 600) and the resulting lack of data.

The implementation of the AI ​​model was then briefly explained. This is a binary classification problem, as the model distinguishes between two prediction classes. One class describes the case in which scrap is generated in a work process, and the other describes the case in which no scrap is generated. This model was trained with 75% of the DANmed data until a prediction quality was achieved that allowed simulations to be performed. The remaining 25% of the data was used for the predictions.

These simulations were carried out as follows: First, a parameter was defined to be simulated. In the first simulation, this was found to be the number of breaks. The parameter to be simulated was then varied by a certain value in all data sets to be used for the simulation. The network could then be queried as to how many of these varied operations resulted in scrap. This procedure was repeated for a number of values ​​so that the results could be presented in a two-dimensional diagram. The y-axis represents the number of operations with scrap, and the x-axis represents the various, varied values. The results of the simulation of the number of breaks are shown in Fig. 1.

As can be seen in the figure, according to the simulation carried out, in about 10% (75 instead of 83 – here

(marked in red) fewer work steps result in scrap if only a quarter of the breaks are taken in all work steps. Thus, there is great potential for optimization here.

Next, the employees performing the work process were considered. First, it was simulated that each employee performed all work processes once. This simulation can be illustrated with a bar chart (see Figure 2). Each bar corresponds to an employee.* Information about employees: Names have been censored for privacy reasons.

Figure 2: Simulation of employees*    

The diagram shows that there are large differences in the amount of waste generated depending on the activities of the employees and that there is therefore great potential for reducing waste, e.g. through individual training.

The final simulation involved simulating employee teams, as DANmed’s CNC operations are rarely performed by a single employee. Here, too, significant optimization potential was identified through the diverse composition of the teams.

Finally, an outlook on the future of AI for SMEs was given.

It can be stated that AI is becoming increasingly relevant for SMEs, as the technology continues to improve and become more affordable. However, it is important for SMEs to start collecting data early, as today’s data will be used to train tomorrow’s machine learning models. A system like our MES solution DOQ supports this requirement. The most important prerequisite, however, is the timely identification of problem areas and issues within the company in order to be able to collect data in a targeted manner.

And  – to see how the use of AI can increase the competitiveness of SMEs, here are some examples from the areas of maintenance management and marketing/sales:

• Reduction in repair costs by 47% in the first year
• Reduction of MTTR* times by 60%

(*Average time required to repair something after a failure)

• Increase in technical plant availability from 40% to 80%
• Increase in product sales by 25% (figures from a US SME)
• 20% increase in sales in a business segment (part of a DAX group)

If you’re interested, we can provide you with a video of the lecture. Please contact us.