How big data and machine learning are revolutionising manufacturing
Wed 8 Mar 2017 | Piotr Niedzwiedz
Piotr Niedzwiedz, CTO and co-founder at deepsense.io, offers an insight into how big data and machine learning technologies are transforming the manufacturing process…
As manufacturing businesses are increasingly on the lookout for opportunities to boost automation, it is clear that big data and machine learning are set to revolutionize the future of production.
Companies are already collecting huge amounts of data from various resources, including industrial machinery, and monitoring precise details of the production process in the hope of improving quality at every stage.
When companies have gathered enough of the right data, predictive technologies can allow businesses to manage the servicing of machinery based on sensor data and advanced analytics, rather than on a fixed schedule.
For a piece of machinery or an autonomous vehicle, for example, these technologies can help predict when and how they are likely to break down. Businesses can then service the equipment before it starts to be a problem and generate losses.
Computer vision and deep learning are proving to be a key asset in quality control environments
Thanks to Industrial IoT (IIoT) and predictive analytics, data available today can be used to foresee future breakdowns, and it’s increasingly possible to see what patterns precede problems. We can use machine learning and predictive analytics tools to mine data for these patterns and detect failures sooner.
In this way, maintenance is a crucial industrial process that can be optimized via predictive capabilities, offering huge time and financial savings. The number of companies implementing predictive maintenance solutions is rising, and with the support of machine learning and IIoT, fast predictive maintenance is becoming more and more effective.
Computer vision and quality control
Similarly, visual inspection systems which rely on the latest advancements in computer vision and deep learning are also proving to be a key asset in quality control environments.
While one wouldn’t typically associate such a project with machine learning, a packaging machinery producer has recently introduced a new fruit sorting machine that uses an advanced machine vision system to sort fruit into different pack grades. The software uses defect classification to detect different types of fruit skin defects. It can also be optimized to work more efficiently in specific applications depending on the fruit and defect type.
A major challenge for manufacturers lies in the proper exploration and usage of big data
Cases like this show how simple processes can be leveraged to bring measurable results in production. What’s more, these solutions can be, and very often are, brought to market so that smaller or even larger businesses can take advantage of the technology in similar manufacturing scenarios.
Of course, we have to remember that universal products based on machine learning generally achieve lower accuracy rates than a customized approach. It is therefore advised that companies oriented towards innovation should consider a customized solution, which can achieve accuracy rates of around 90%.
Outsourcing data support
A major challenge for manufacturers lies in the proper exploration and usage of big data to assure that it really is optimizing resource and process management. Organizations are increasingly looking for third-party advice to learn how they can use their data more effectively, and to discover new methods for data collection, in order to achieve their business goals faster and to outrival the competition.
Predictive analytics and machine learning are also quite advanced data science techniques which require specific skills and experience. For companies deciding to incorporate predictive analytics into their production processes, the priority needs to be finding talented data scientists. Companies starting their journey with data science should consider outsourcing talent and extending their teams with data science experts for specific projects.
Data engineers are also needed to design and implement systems for capturing and storing the manufacturing data, making it available for analytics, and ultimately for processing new data streams in real-time.
An expert partner can help support these data science requirements, helping businesses unlock potential at every stage of their data science maturity.