Machine Learning / Artificial Intelligence

We are using them everyday in our smartphones, navigation systems, web browsers and maybe even the fridge. If you ask different people from different backgrounds what the uses are, they will tell you different things. My impression is that Machine Learning or Artificial Intelligence can do almost everything that a human brain can. Just like a human brain, it needs training. For humans it can take 16-20 years of education to learn and master a topic, for ML/AI it can take a few months and a lot of data. I believe that in the future, there will be a big progress in this field and it will be easier to apply these technologies in real life as an end-user.

I'm not an expert. Actually my first deliberate effort in using machine learning came just a few months ago while working on a quality improvement project. With the help of a company that makes a data analysis platform, we were able to identify different failure modes and then break down a big problem that seemed complicated into smaller problems that can be solved more easily. We used a clustering algorithm to discover patterns in our measurement data. ML can find similarities and rules where humans are overwhelmed with information. Then we created a classification model to give near real-time information to the operator.

There are many algorithms and models that can be used for different purposes. ML libraries like Tensorflow, SciKit-Learn and others can be used in different programming languages and have a huge number of available models. You just need to identify which model is best suited for your problem. All these are available to use in your preferred programming language and there are many tutorials. Just search online for your problem + machine learning and probably you will find someone who had the same problem and used ML to solve it.

With implementation of digitalization, Industry 4.0 and Internet of Things we now have access to a huge collection of process data(coming from sensors, measurement devices or human input) that can be analysed using ML to find the best process parameters so that the part or product would have the desired quality.

Quality inspection can be fully transferred to AI. Measurement data and images from cameras can be used to determine if the parts are good or bad, then based on this decision to sort the parts and create the reports. All this product data can be fed back to the machine controller to adjust the process parameters so that defective parts would not be produced any more.

We could also use AI to improve logistics performance: optimize the transport routes & frequency, improve forecast quality, discover patterns in customer and supplier behaviour to better go through crisis periods. Ultimately, we could have AI suggest management decisions and predict more accurately the outcomes.

Check this out: https://teachablemachine.withgoogle.com/ you can easily create your own image or sound classification model that can be transferred to your smartphone/tablet for real-life use. Of course, if you use just a few images, the model will not accurately classify all the real-life diversity.

I'm still a beginner and I'm interested to learn and apply ML/AI together with lean principles as I believe that they complement each other.

One of my next personal projects will be to load the data from my apartment's temperature sensor and create a new software for the heating controller that I implemented 2 years ago so it would adjust the temperature based on my past preferences and daily schedule.

Do you have some cool project where I could contribute? Send me a message!