Artificial Intelligence can solve more and more problems. The question is:
How do you maintain productivity when transitioning an industrial process to AI?
Over the years at Cornis, we’ve helped wind industry leaders to transition seamlessly from a mostly human process to an automatized one, here are the steps we follow.
Cornis enables wind industry leaders to optimize their processes. Here are the steps we always follow.
1. You have one job
Start by deciding which job, which tasks you want to automatize and concentrate your transition on it.
If you try to apply AI in every process at the same time, you won’t see productivity improvement soon.
To define an effective scope, we ask these questions:
What is the input data?
Everyone has a vague idea of what are their input data. For example “wind turbine inspection pictures” for wind turbine inspections.
The more you can describe your input data, the smoothest the transition will be.
Input data are every elements your teams need to do the job. For example images, on-site written information, cultural or expert knowledge, etc.
With inspection pictures, do you own a procedure describing how you get these data? Do you use a specific camera designed for the job? Is there more data available? Does your technician fill out a paper describing the mission? Do you get the pictures in folders, ordered or mixed with non relevant content?
If you use your process several times a year, your input data must respect guidelines. Take a bunch of example of these data to figure the similarities on this task input data.
Who is doing the job?
Include in your transition project the experts doing the task you are working on. They are the one who know best how to do the task and what to expect from an automated process.
Transferring people’s experience to algorithms is a major part of AI transition. It’s essential to find every person who contribute to the task you want to automatize.
What is the expected result?
The usual answer to this question for industrial companies is a report or an Excel file. If your process is online, be more specific to transition easier. You don’t want to spend most of the time of your AI project creating a report generation algorithm.
With wind turbine inspections, we focused on determining if a picture has a defect or not. And we left the other part of the process out of scope.
Take time describing the result you expect. It will be easier to train and improve machine learning algorithms.
2. Explain like I’m 5
Headlines presenting machines as smarter than humans are only marketing.
Machines are dumb, but they have huge memory and computing power.
Humans are smart, but they have low memory and computing power.
You need to split the task into simple steps to allow the machine to learn it. You need to deliver your expertise in such a way that anyone can understand it. This is one of the most laborious tasks to integrate Machine Learning in your business.
We did this subdivision for expertise on wind turbines. And we identified that 90% of the job — defect detection — needed low expertise. Automatizing this first task led us to divide by two the expertise time.
3. Design your workflow with no automation
To manage an AI transition, your process needs to be robust enough that it still works in case of no automation.
Before being an automated process, it will be a hybrid one. Machine Learning algorithms will only be efficient on a small part of your data.
In the early stages, AI algorithms quality differs from one input to another. Be ready for the worst-case scenario if you don’t want a productivity drop.
When we developed the Panoblade production chain, our algorithm had around 10% efficiency. It went from 90% of the images stitched for few inspections to 0% for most. We delivered every inspection on time thanks to an architecture designed for failure.
Your AI enabled workflow needs to work well without automation.
Don’t replace your good old manual process, transform it to be AI Ready by ensuring these two elements:
You need a quality assessment procedure.
Describe how you assess the quality of your work : what are the steps? Can you give a quality score?
An automated quality assessment procedure is not mandatory. Well designed User Interface can deliver high productivity on quality assessment.
You need to have a tool to correct a proposition.
You need to make sure you have tools ready to correct an AI proposition.
As soon as you have these two elements, you can use AI in your process:
- You process every new data through your early stage AI algorithm
- The quality assessment procedure helps you select the data to correct
- Using the correcting tool, your users can adjust the AI response
- A feedback loop allows the AI algorithms to improve results
In our experience, you start saving time right from the beginning. Assessing the quality of a mediocre AI result takes usually less time than doing the job on all the data.
This architecture allows you to have constant feedback on the Machine Learning processes.
You will use Quality Assessment results to track AI performance. You will be able to use the detected errors to retrain the algorithms.
Learn more about automation and AI in industrial processes on Cornis Blog.