Amongst all the Factory 4.0 chestnuts, artificial intelligence appears to be THE source of gains.
According to BCG, organizations that adopt and integrate AI are better equipped and will double their cash flow by 2030. Yep.
The truth is that currently only 55% of industrial companies are starting to develop AI projects, either in-house or via simple applications.
Yet, its advantages are numerous: quality management (reduction of quality control costs and optimization of maintenance), reporting and analysis of data patterns in real time, etc.
Let's take the example of maintenance time optimization: on average, a factory loses 5% of productivity during machine downtime (and this can easily rise to 20%). This mechanically leads to difficulty in meeting production objectives, delays in customer deliveries, loss of revenue, mobilization of resources related to the resolution of the problem, or costs due to overtime, to name a few.
According to McKinsey, for industry, the greatest added value of AI today comes from its use for predictive maintenance ($500 to $700 billion in the global economy).
Again according to the firm, companies that use AI and machine learning for product defect detection and quality testing can increase factory productivity by up to 50%, generating up to $2 trillion in value in supply chain management and production.
So why do so few manufacturers venture into this area?
User training costs and available tools:
Businesses do not master development and do not have a simple and secure application where they can iterate and exchange with technical users.
Too many different systems to connect to improve prediction. Indeed, if the data from IoT sensors is preponderant to do predictive maintenance (vibrations, temperature changes for example), other data sources can be included to further refine the prediction :
- Location data
- Manual data via human control
- TSS data
- Static data
- History of equipment use
- External data via APIs
- Data from programmable control equipment
Okay, but how do we do it then?
We need to have the capacity to bring together these different data and correlate them on a data platform. This platform must be able to do data science and if possible in low-code so that non-techs can work with techs on these subjects, without friction. Indeed :
- This allows connectivity to many data sources
- There is no limit to the amount and complexity of data
- The platform is inclusive: easy to use, regardless of technical background, and therefore allows for vendor independence and enhanced data analysis opportunities
- It is adapted to the "small steps" method: it does not limit teams to a single use case but rather allows them to start with one case and scale up by integrating machine learning and AI.
So if we recap:
Predictive maintenance is a strategic issue for plants.
Its implementation within the workshop is therefore essential and the integration of an edge solution allowing the feedback of machine data in real time is a prerequisite.
The platform must be able to bring up data from various systems, all without coding to allow adoption in the plant and the rise in competence of employees.
If you want to talk with us about these topics or discover Niagara Edge, feel free to contact us via the live chat on your right or here!