Read this before you move on with industrial data science
Frederik Van Leeckwyck
on , updated
Curious about the possibities that machine learning (ML) can bring to your industrial company? Find out why ongoing instrumentation and automated data collection is the key to building smarter industry systems, and how the technology choices you make today will impact your future data science initiatives.
It’s hard to find a production company that isn’t excited about the huge potential artificial intelligence (AI) can bring to its operations. According to Deloitte’s global survey on AI adoption in manufacturing and production, 93 percent of companies believe AI will be crucial to drive growth and innovation in Industry 4.0. Yet, only 9% of organisations are currently leveraging the power of data science.
Why autonomous production operations are still far off
Many data scientists are convinced that in the future it will be possible to create a fully autonomous or ‘lights-out’ factory, by ingesting a combination of production data from sensors, machines, and people, and then applying it to derive advanced algorithms.
Yet, these intelligent and autonomous systems are still a long way off from today’s plant floor. What is more, many industrial companies still lack the right digital technology to properly implement data science initiatives, such as machine learning.
Instrumentation as the key to getting smarter
Despite the hurdles still in the way of ML adoption, the production sector is getting more instrumented by the minute. While the challenges to finally turn the lights are yet to overcome, today’s automation battle is about making systems smarter and smarter. For that reason, factories keep adding sensors and collecting ever more data.
We all desire to build smarter and smarter systems... It is the grail of coding and moving things forward... Therefore, the future is instrumented... You can’t start learning until you’ve instrumented.
This does not mean that the end goal of all industrial applications is AI. PID controllers, for instance, used in process control applications to automatically regulate process variables such as temperature or pressure, use feedback loops and work neatly without AI. However, to manage more complex applications, you will need predictive models.
So, where to start when aiming to move on with any data science initiative?
Why data collection should be your first move
The first step in any industrial data science project is to collect OT data from machines and equipment using a historian. This enables you to improve processes, find new insights, and continuously improve effiency, quality and yield. Yet, the time-series database at the core of the historian is also the perfect soil to grow data science applications on.
How historians power your data science project
To pave the way for future data science applications, the technology choices you make today will have a large impact on their success. You will need data collection technology that is extremely robust, can integrate any data source, and has no limitations on the amount of data you will want to collect and process. Another thing you will want to avoid is being hampered by a traditional pay-per-user business model.
This is where open-source based historians come in.
Data historians built on open frameworks will heavily support any data science initiative by seamlessly integrating OT with backend IT. As a result, data scientists, programmers and engineers can also develop AI and ML capabilities for backend IT systems. Developers and front-end OT engineers can then leverage those capabilities in the OT environment.
Data to the people
In our experience, employee adoption of open-type historians generally spreads throughout the company like a wildfire. Since there are no limitations on the number of users, anyone in any department is finally free to start querying production data, generate graphs and get new insights.
As such, data becomes a fundamental and widely used company asset. This also means that when a data science project is started later, the benefits of ML will not only accrue to a select club of historian or AI enthusiasts.
Here are four ways in which open-type historians will boost your data science project:
Lower entry. They will drastically lower the entry barrier for future implementation of advanced applications such as AI, ML and DL.
Unlimited data. Unlimited amounts of high-resolution process data are collected, creating a solid foundation for any data science initiative.
Scalability. They offer a flexible, modular system that enables you to start optimising a single asset and scale up further as needed.
Developer-friendly. Programmers have access to the historian data using familiar tools and APIs. For them, open-type historians feel like a familiar world.
Data science in Industry 4.0: first the basics
There is no doubt that machine learning will bring huge benefits to businesses in the process industry, in terms of monitoring and maintenance, quality control, R&D, and decision-making. Lights-out production, however, is still a distant dream. The only way to move forward with data science and build smarter systems is to increase instrumentation and gather more data.
Factry Historian, which uses open components Grafana and InfluxDB, acts as a robust and scalable foundation that can collect an unlimited amount of data from a zillion sensors through a single interface. Advanced solutions can be built on top in order to transform data into useful insights and predictions, which will help to improve processes and quality, and make better business decisions.
Lay the groundwork for industrial data science
Collect any type of process data. Build applications on top.