AI: female engineers set the tone

Why should an engineer learn about artificial intelligence methods? Laura Neuendorf is doing her PhD at TU Dortmund in Prof Kockmann's group “Laboratory of Equipment Design”. She explains what the potential of AI might be in her field. 

DECHEMA: The KEEN research project involves more than 20 industrial and scientific institutions that are jointly pursuing the goal of introducing artificial intelligence (AI) technologies and methods to the process industry. TU Dortmund – Laboratory of Equipment Design is one of the three incubator labs within the project. What does that mean in concrete terms? 

• __Prof Kockmann’s group at TU Dortmund develops modular laboratory and pilot plants for continuous chemical and pharmaceutical production. Our laboratories include small columns for extraction and distillation, devices for continuous crystallization and microstructured reactors. Within KEEN, we do not only provide data sets for the development of AI applications, but also develop AI solutions ourselves.
  My PhD thesis, for example, deals with AI-assisted image analysis applied on extraction columns. To ensure the transfer of the developed methods into practice after the end of the funding, AI-based business models are formed for the three incubator labs. There is a great potential for the implementation of AI in the process industry especially for process modelling, engineering and the optimization of plants. 

DECHEMA: Where did your enthusiasm for artificial intelligence methods come from? 

• __Like probably most people, I first came in contact with the term "artificial intelligence" through the media. Headlines like "AlphaGo Zero: AI becomes unbeatable in Go without humans" or "AI is chess world champion" awaken interest. In my bachelor's study, I attended the lecture "Introduction to Programming". It was curiosity that finally drove me to apply AI methods in the context of my master thesis. The approach was kind of obvious, since my topic dealt with the analysis of X-ray images from micro-computed x-ray tomography. As a bioengineer, I first had to familiarize myself with the programming and with AI methods. From my point of view, programming is an important skill that will gain more and more importance. I was the first in our research group to present a thesis in the field of AI. 

DECHEMA: What is the challenge in developing an AI application? 

• __Surprisingly, the learning process was very fast. This is also confirmed by the students I supervise as part of my PhD. Within a very short time, the basic framework of the program is set up. I myself learned a lot by trial and error. Programming hints and tips can also be found on the net. The overall goal of my work is to be able to optimally operate an extraction column by the use of AI-based recommendations. When operating a separation column, e.g. the stirrer rotation speed, temperature or the flow rates have to be adjusted. In order to do that, there is the possibility to identify the operating condition by an AI image recognition method. However, to derive a recommendation, additional process parameters, such as pressure loss and temperature, must be considered. If the droplet size could be reliably determined by an image recognition process, we would probably have a second option that could even work without additional data. At the moment, I still regularly check with my own eyes whether the AI application delivers correct values. In my master’s thesis, I have already proven how well AI image recognition methods work. They are now capable of beating the human eye in image recognition, which was presented at the annual ImageNet Large Scale Visual Recognition Challenge (ILSVRC), the software competition for object recognition. I really enjoy programming and the implementation of AI methods. Of course, there are several challenges to overcome. Data pre-processing also takes a lot of time, which should not be underestimated. Here, I would like to refer to the ACHEMA Innovation Challenge within the KEEN-Hackathon by ABB.  

DECHEMA: For the ACHEMA Innovation Challenge you have formulated a challenge that is related to your doctoral thesis. What makes your challenge so interesting? 

• __The challenge I formulated is very illustrative. The distinction between dogs and cats is a prime example of machine learning. Artificial neural networks are ideally suited for analysing images according to certain features. The process starts with an appropriate training phase. The challenge can be a good starting point to learn about AI methods. Particularly due to the Corona pandemic, topics of digitalization have awakened public interest. Perhaps now is the right time to get familiar with AI. I myself have already participated in a hackathon organized by the Sievers Group. In interdisciplinary teams, we programmed a game called Conway's Game of Life. Everyone brings her or his own strengths to the process. I learned a lot about interpersonal communication during the process. I would also like to encourage female talents to participate in the challenge. In terms of AI, I follow Allie K. Miller, Global Head of Machine Learning Business Development, Start-ups and Venture Capital at Amazon Web Services; but there are countless other role models in the field of technology. Meanwhile, there are also hackathons just for women. With my challenge, I would like to take away the fear of coding for young scientists and developers of all disciplines. Participation is worth it, I promise. 

https://www.achema.de/en/the-achema/innovation-challenge

The interview was conducted by Dr. Simone Rogg.