From autonomous cars to speech recognition, artificial intelligence (AI) is making inroads into our lives and businesses. But one lesser-known area where AI is gaining traction is in manufacturing, where real-time data and machine learning are helping drive down costs and increase outputs. According to the McKinsey Global Institute, manufacturing is the sector that generates the most data yet has the least digital penetration.
Still in the early stages of adoption, AI offers a potential competitive advantage to manufacturing businesses that are open to new ways of operating. If that describes you, read on to learn how you can use AI to drive exciting improvements in your company’s outputs, costs, and even culture.
Two areas of AI impact – and the one that stands out
AI is making its mark on manufacturing in two major ways. The first is computer vision, which involves using automatic object detection to control complex systems – most famously applied in Tesla’s self-driving cars. In a manufacturing setting, shop owners can take advantage of computer vision by installing cameras on the factory floor to track inventory and automatically identify product defects.
But the area where AI holds the most promise to improve manufacturing is in advanced machine monitoring.
Machine monitoring involves collecting data about a shop’s manufacturing processes to help make better decisions about those processes. Manufacturers can use the information they glean (like how often machines are lying unused) to increase output and identify cost-saving efficiencies.
AI takes this data collection a step further by using machine learning to visualize, diagnose, and predict machine problems in real time, further increasing equipment productivity and efficiency.
AI-based machine monitoring is developing quickly
AI-based machine monitoring is in its infancy, meaning it has lots of room to grow. However, the practice still has challenges to overcome before being adopted by the industry as a whole.
First, in my experience as chief data scientist at MachineMetrics, manufacturing is five to seven years behind other industries in terms of technology adoption. When you're producing physical parts, your guiding metric is typically a tangible end product. As long as that product looks good and sells well, there may be less impetus to optimize and modernize.
Second, many manufacturers are hesitant to let their data leave the factory, insisting on an on-premise machine monitoring implementation where all data is stored on internal servers. This creates the problem of siloed data, which is not ideal for adequately training an AI-based machine learning model. Imagine if Netflix tried to build a recommendation engine using data from only one household!
However, companies like MachineMetrics have moved to cloud-only solutions and started to amass large, diverse datasets across many disparate processes, leading to the first glimpses of more general-purpose or application-agnostic algorithms for many types of machines. As an objective indicator of this trend, we’ve seen a recent uptick in patents awarded and publications in major journals regarding strong AI algorithms in manufacturing.
Lastly, there is often resistance among manufacturing business owners to adopt AI platforms out of fear that they will eliminate jobs. But this fear is misplaced.
The purpose of using AI for machine monitoring isn't to replace humans, but to supplement human expertise with computer-guided capabilities to make the factory run more smoothly. Software can take over things like tracking the number of parts created or how often a machine is up or down – rote, menial tasks that workers used to do by hand. This frees up workers to focus on higher-value work, like running initiatives to improve efficiency using the new data. The additional revenue driven through AI-based machine monitoring can also be invested in training programs to help workers develop skills that will benefit the business.
At MachineMetrics, we find that shop culture also improves with the increased data visibility gained through AI and advanced machine monitoring. When people can more easily reference data and machine-learning driven, intelligent recommendations to support their decisions, this transparency helps foster a healthier, more rigorous, and more productive work environment.
Getting started with AI-based machine monitoring
If you’re a manufacturing business owner looking at AI-based machine monitoring, there are several key things to consider.
First, you have to walk before you can run. If you have no digitization or automated visibility into your processes, you need to start with descriptive analytics: a system that tells you when your machines are actually running and how many parts you’re producing so you can optimize their use. With descriptive analytics, for instance, you can identify a machine that’s been sitting idle for two hours at the beginning of every first shift. Optimizing an inefficiency like that can save you thousands of dollars right off the bat.
Next, whether you’re setting up a descriptive analytics system or a more advanced machine monitoring system, you’ll encounter various technological requirements that a partner company can help you identify.
But outside of the technology, you also have to make sure your company is a good cultural fit for machine monitoring and advanced analytics. This means being thoughtful about what types of efficiencies you’re looking to drive instead of simply looking to adopt the latest technology. Machine monitoring with AI is not a “magic box” that will solve all your problems without human input and close collaboration.
This means getting buy-in from multiple key stakeholders in the company and partnering closely with your technology vendor. Implementing a machine monitoring system, even one without AI, is a complicated venture into pioneering territory for most shops. And it’s one that requires organizational maturity and an eagerness to improve.
AI-based technologies like predictive machine monitoring and computer vision are still taking root in the manufacturing world, and their impact holds tremendous promise. As time goes on, processing speeds will improve, technology will become cheaper, and algorithms will get better. And more shop owners will learn that AI is not a replacement for their staff, but a powerful lever to identify efficiencies, optimize processes, and elevate their workforce.
