March 2020 was the start of a historically exceptional period.
This time can be broken down into three phases. First, a sudden stop on the economy and therefore a halt on a large part of manufacturing (except for essential needs like energy or agrifood). This was followed by a phase of minimum production resumption. It is estimated that the French Industry fell to less than 50% of so-called “normal production” in April and May 2020. Lastly, we’ve seen a revival and a return to stable production levels. However, even these are estimated to be 10% lower, on average, than the historical normal volume.
Like most parts of life these days, the short-term future remains very uncertain. Normalcy is still disrupted by curfew restrictions—or even containment measures—in France, Spain, Italy, and many other countries. The United States experienced three significant spikes in COVID-19 cases as states struggle to set consistent protocols.
These days, the question I ask myself is: how could have, should have, or would have new digital technologies (particularly Artificial Intelligence) helped us since the start of the COVID-19 crisis?
How AI Could Aid Manufacturers during the COVID-19 Crisis
To answer this question, let’s first define these technologies. It’s important to understand how we define Artificial Intelligence (AI) and how it is applicable to industry. According to Luc Julia, AI does not exist. No computer system to date is capable of cognitive intelligence identical to that of human beings and it is unlikely that we will achieve this level of cognitive intelligence in computers for a very long time.
Instead, we use AI as an umbrella term for all computer applications that use Machine Learning systems to replace or augment human tasks. Julia prefers the notion of Augmented Intelligence. Since different programs are designed to replace different kinds of human tasks, there are actually several families of AI with each one using specific technologies to function properly.
Types of AI for Manufacturing
There are seven AI patterns identified by Cognilytica. Below are three relevant examples used in manufacturing:
- Autonomous Systems are aimed at reducing human effort, capable of adapting by learning and evolving situations.
- In Industry, Autonomous Systems include: Predictive Analytics, which help teams in their decision-making, to plan for maintenance and/or production configuration.
- Visual Recognition consists of product models for sorting and classifying anomalies (this application of AI is often the one presented by the media, with facial recognition in particular, hence sometimes confusion when presenting other types of applications for AI).
- interacting with humans, as do voice assistants who can help with training or execution
- recommendation systems, access to objectives, which find the best solution to a problem
- the detection of models and anomalies, making it possible to identify good and bad practices, using say textual or voice data for example
- Hyper-personalization treats each individual as a unique being, which makes it possible, for example, to produce a particular product per customer.
- For example, a sports shoe manufactured individually on demand or specific customer design choices built at time of order
Users of AI and Machine Learning in Manufacturing
To become an effective technology for manufacturers, AI and machine learning needs data and users. The data will be made available to AI thanks to the new essential infrastructure: the Industrial IoT platform, which allows the collection, centralization, protection, preparation, and access of data to all users, through connectors to other systems or more simply applications with embedded AI.
There are even new careers specializing in AI that will gradually become acquired skills for production teams.
The new Data Scientist’s mission is to create tools adapted to particular use cases by designing computer programs using AI algorithms.
Not to be confused with a Data Analyst who is in charge of using existing applications to benefit from AI. Or the Data Architect whose mission is to prepare data for AI from raw collections on machines or information systems.
AI has above all become more and more accessible, thanks to the arrival of applications with embedded AI, which allow everyone to benefit from it without having to develop and code, clean and prepare data (automation of information in the data lake of the IIoT platform), or become an advanced statistician. Engineers will become natural analysts (citizen data scientists) augmented by assistant applications, just as in our lives we are augmented by the applications of our smartphones (generic applications easy to use, like a smartphone application like Maps or Waze, but which use a lot of technologies like GPS, IoT, big data, and AI).
The Value of AI
The impact of AI, and its gradual deployment in factories, will help increase productivity by solving complex problems, assisting teams, or automating tasks. But another fundamental characteristic is the virtuality of this technology and therefore its instant and permanent accessibility on IoT platforms that operate as collaborative networks.
And the acute COVID-19 crisis has revealed the need to adopt these solutions. We polled industrials on the biggest organizational challenges during COVID-19, and the majority responded with frozen budgets. As Industrials are expected to do more with less, access to automated reporting and actionable insights becomes ever more critical. The ROI IIoT platforms powered with AI can often quickly yield ROI that outweighs the cost.
When blue-collar workers returned to factories, white-collar workers most often stayed away. And if the production teams were already equipped with AI systems, the factories were able to benefit from them independently (process control, maintenance or predictive quality, logistics optimization), then the technical support or project staff were able to continue to propose actions of continuous improvement and resolution of problems thanks to the use of their available data.
Of course, the daily production monitoring has been made easier thanks to the automation of reports and virtual meetings based on connected dashboards. Not only have factories that had already embarked on a digital transformation through the integration of applications with embedded AI been able to continue to operate and progress, but they have also discovered new practices and changed the way they operate collectively.
At Braincube, we have been able to measure this phenomenon by two objective criteria: an increase of more than 80% of the use of our tools since May 2020, and requests for faster deployment from customers during the integration process.
In conclusion, this COVID-19 crisis acts as a great revealer for our society. We discover, among other things, less polluted cities, the economic importance of network infrastructures (internet and mobile), the fragility of certain industries and the great resilience of others, the real state of our health system, the irrationality of political decision-making (which should use data other than polls!), and finally the need to transform in order to strengthen.
Virtual technologies are accessible with much lower investment levels than hardware technologies. And since it is in times of crisis that we transform ourselves the most, pushed by the pressure of the situation, it becomes essential to build a new vision and a roadmap that will allow the deployment of new uses with high added value. as quickly as possible. AI should not be seen as the solution (it is a new technology available among others) but rather COVID-19 as the opportunity to ask the right questions.
The Importance of Legitimizing AI
Production teams use AI-enabled analytical tools in an IIoT platform to gain access to the data that can answer their questions or offer them prescriptions at the right time. But how legitimate are these AI solutions? How can industrials ensure the suggested parameter modifications that AI proposes are the “best”? CEO of Braincube, Laurent Laporte, discusses the importance of legitimizing AI in Industry.
AI vs. Machine Learning vs. Data Science
The popularity of Artificial Intelligence (AI), Machine Learning, and Data Science have grown in manufacturing from the pursuit of the connected factory and Industry 4.0. But with this expedient growth has also come confusion. Understand the key differences, how they intersect, and use cases for each.
4 Ways to use Machine Learning in Manufacturing
Continuing to find new ways to improve operations requires increased creativity, capacity, and access to critical data. Industrials use Machine Learning to identify opportunities to improve OEE at any phase of the manufacturing process. Learn how to use Machine Learning to solve some of the biggest challenges faced by manufacturers.