At the time, in 1997, I was still a student and I was doing my research paper with a small company from Verviers, Automation & Robotics, specialized in quality control for the ophthalmic industry. This SME had developed a system that, thanks to a camera, could identify defects on lenses. With the team, my job was to automate detection and classification of these defects (scratches, bad polishing, etc..). From there came the idea of using Artificial Intelligence algorithms to optimize performance in industry.
I was using tools that were developed at the Montefiore Institute at the Université de Liège, ULg Belgium, by Professor Louis Wehenkel, and mainly applied for decision support in electric power transmission networks. I joined this laboratory as a research engineer and we started to collaborate with various industries in the region.
Then, the idea of creating PEPITe came naturally to promote the tools in the industrial sector where a lot of data was already available, but very little used. As a young engineer in mechatronics, I saw AI as a disruptive way to make factories more efficient by exploiting this data.
Our first software tool, Pepito® laid the foundation for what DATAmaestro® is today. However, it was not yet easy to use by non-experts and we used it mainly for research programs and for our services to the industry. We wanted to make it easy to use and to integrate it into an industrial environment. Very few people were working in AI in the early 2000s, but as young engineers, with the co-founder of Pepite Dominique Archambeau, we had the dream of revolutionizing the world of industry. Step by step, we convinced plants to get involved in the subject with us. This process took much longer than we had imagined, but our dream was taking shape...
The Artificial Intelligence we use is based on what is called "machine learning". This field is based on capturing and processing raw data to build rules and models to help make decisions in real time. Ultimately, this accelerates and factualizes decisions that are often based on intuition or approximations.
Artificial Intelligence algorithms have certainly improved, but they are based on the same foundations. In DATAmaestro®, we have combined a range of known techniques that are particularly powerful and robust for manufacturing applications. AI techniques are certainly not new, even if we talk about them a lot more today.
However, what has evolved a lot over the last 20 years is computing power and storage capacities! This evolution has made it easier to implement data and computation intensive algorithms.
Domain knowledge remains fundamental in the use of AI. It is important to understand the issue to be solved and the business context. Then it is essential to understand the data you are working with in order to correctly formulate the problem, select the right algorithm for that problem and validate the results. An Artificial Intelligence that solves everything does not exist. Each case will require a very specific AI. That's why we focused on industry.
We have been using and refining our AI tools for over 20 years to create reliable and robust methods. The success of an AI project does not only depend on the reliability of the algorithms but also on the data quality and availability and the way the algorithms are used. If the data is bad, the AI will be too!
This is why, in order to implement an AI project, we systematically carry out a Digital Data Diagnostic to evaluate stored data, which gives us a first idea of the applicability and feasibility of AI.
PEPITe sees Artificial Intelligence as a means and not as an end! Yes, we are a technology company active in AI. But our goal is to maximize impact on plant performance.
A key issue for companies is sustainability and their environmental impact. A sustainable industry is profitable but is also very careful about its environmental impact. This includes using raw materials sparingly, minimizing waste and emissions.
These objectives, inscribed in PEPITe's DNA, naturally lead us to implement AI tools and know-how to help industries maximize their performance, to minimize use of natural resources and bring a sustainable and measurable impact on the planet.
Find more about our commitment and read our article "WORLD CLIMATE DAY: WHAT CAN WE DO AS AN INDUSTRY?"
The best way to control and optimize performance is to measure it. It's the same as when you're running. If you measure your heart rate, your speed, etc. you can better manage your run and improve faster.
For an industrial plant, we start by helping to continuously measure key performance indicators such as productivity, energy consumption, raw material, yield, carbon footprint, product quality, etc.
Data analysis, and in particular AI, will then allow a plant to understand what influences their performance and help to quickly identify what needs to be done to improve.
AI is an enabler for performance improvements and processes efficiency. However, there are limits to this technology. Data quality is essential and AI implementation requires solid domain knowledge. For example, while AI is perfectly capable of detecting correlations between events, it is unable to determine the cause and effect relationship without the contribution of human knowledge.
Example: Do cars stop because the light turns red or does the light turn red because cars stop?
In industry, AI tools help operators and engineers to be productive by letting workers focus on higher value-added activities. AI can help increase capacity and speed of problem solving for engineers and helps guide operators. It helps decision making by providing key information aggregated in real time. This can relate to energy performance, production line anomalies, real-time predictive product quality or more. These tools continuously ingest data and provide output information. They can detect operating anomalies at an early stage, to avoid minor process upsets through to potentially catastrophic events.
Some examples:
There is indeed nothing magical about it, but there is a lot of data and calculations... Moreover, the title of the movie Imitation Game, which tells the story of the mathematician Alan Turing, is very well chosen. AI imitates as much as possible what it has seen in the past and correlates events on the basis of advanced statistical calculations. The power of these calculations and the data processing give the impression of magic.
We are still a long way away from this famous general Artificial Intelligence that rivals the thinking and reasoning abilities of the human brain. Today, without the human intelligence that provides the context and the method, AI would be sitting on a shelf...
