Manufacturing, in most cases, is a high-volume, best-margin game. Most large manufacturing companies across the globe are sitting on petabytes of data. They have the potential to produce more data as the data storage cost keeps reducing. Such companies are definitely Data-rich; however, they are Insight-deficient. Even today, we encounter news of large manufacturers performing product recalls or churning out subquality products.
Production heads are under constant pressure to adhere to delivery timelines and quality control benchmarks. Unplanned downtimes and poor quality products are the two most devastating outcomes for any manufacturing company.
2020 had been a less forthcoming year for the manufacturing sector in general. Despite this, there a few companies whose performance has been unphased by global cues. These companies have learned the trick of the trade — Smarter Manufacturing. Smarter, as relative as it sounds, is focused on continuously improving the bottom line. Computer vision-based quality checks, connected equipment, predictive maintenance are some of the many use cases that help companies produce superior products with the best possible production cost. Such companies are truly leading the digital manufacturing revolution — Industry 4.0.
Under the hood, a superior Industry 4.0 solution leverages the power of cloud computing to produce almost real-time insights that are beyond human capability. Understanding the scope of cloud computing in tandem with Artificial Intelligence (AI) becomes essential to foresee how Industry 4.0 shapes up. Let us take some examples of Amazon Web Services (AWS) that have the potential to transform any shop floor, no matter its size and complexity.
Amazon Monitron — AWS
Amazon Monitron service is like an over-the-counter solution for Predictive maintenance. The primary components of this service are tiny wireless sensors and a gateway box. Both of them are readily available on Amazon.com. Once you get it, use some epoxy to stick it to the industrial equipment of your choice, pair it with the Amazon Monitron mobile app, and within minutes you have deployed a Machine Learning (ML)-based predictive maintenance solution.
Internally, the hardware captures four main parameters: Temperature and Vibrations on the x-axis, y-axis, and z-axis. Using these as an input, Amazon Monitron will build an ML model and determine a baseline condition for normal functioning. Over time, once it detects any anomalies from the data, the app will alert its user.
Such an out-of-the-box solution can indeed be transformational for small businesses that would otherwise have limited access to resources for running a Data Science team or deploying sophisticated sensors. With no ML expertise and the pay-per-use nature of the cloud service, a scalable Digital shop floor would be a reality. Such sensors can be deployed easily on motors, gearbox, pumps, compressors, and bearings to monitor equipment health continuously. Even though the anomaly detection process would not be so transparent, the instant value that a shop-floor generates on deploying such solutions benefits the manufacturer and the industry as a whole.
Amazon Lookout for Equipment — AWS
Now, let us assume a shop floor with a sensor network in place which captures the data of many pieces of equipment. Even though this company is capturing the sensor data, they do not have the expertise to harness the captured data's full potential. They want someone to interpret the data for them and use the insights to manage the factory floor smartly. Enter — Amazon Lookout for Equipment service.
This service creates a curated ML model from sensor data per industrial machines and identifies early warning signs that can cause equipment failure. For this service, the prediction accuracy has the potential to be high because the ML model can take data inputs from up to 300 sensors per industrial machine. If we considered a Compressor for illustration, some of the equipment's data parameters would be RPMs, inlet/outlet flow rates, temperature, pressure, etc.
Amazon Lookout for Equipment service can pinpoint individual sensor or sensors that show abnormal equipment behavior among many equipments sending data via hundreds of sensors to AWS for analysis. These insights are sent as alerts to the technician or can trigger remedial action, such as logging a ticket.
Cloud-based predictive maintenance services such as this will help manufacturing companies adopt cutting-edge technology faster without requiring heavy investments in talent and technology.
Amazon Lookout for Vision — AWS
This computer vision-based service is used to spot visual defects on a fast-moving production line accurately. This service aims to help manufacturers produce high-quality products and accelerate the quality control activity that is currently done manually.
The process to deploy Amazon Lookout for Vision is reasonably quick and straightforward. The Amazon Lookout for Vision console needs a minimum of 30 images to establish a good baseline state. The service would automatically build a curated deep learning-based model within minutes of deployment. On detecting defects on a fast-moving production line, alert notifications are sent to plant engineers to take corrective action while recording the anomaly's image details in the console. The best part of this service is that the lighting, camera angle, and pose variations will not affect the anomaly detection. Like the other services mentioned above, this service does not require any expertise to train and deploy the ML model. The applications of this service can be endless not only for manufacturing but also for other related industries.
AWS has various other services that can facilitate a smarter factory. AWS Panorama can help improve inventory tracking and site safety by enabling installed cameras or edge devices with computer vision. They have computer vision-based solutions curated to industry. This service also has a Software Development Kit (SDK) that allows cloud specialists to deploy custom-built computer vision solutions to the existing edge devices. Apart from this, AWS has a suite of IoT related services that help data collection, organization, and analysis scalable and dependable. For curated AI/ML solutions, Amazon Sagemaker and its related services help carry out end-to-end Data Science projects.
Other cloud service providers like Google Cloud Platform (GCP) and Microsoft Azure have an equally capable ecosystem of ML as a service. Google Cloud AutoML of GCP and Azure ML of Microsoft Azure are worth exploring.
From smartphones to smart TVs to smart appliances, I believe ‘smart factories’ is a matter of time. Continuous transformation to Industry 4.0 is the most silent revolution that is happening around us. Empowered by the flexibility, reliability, and advancements in cloud services, Industry 4.0 will be a reality sooner than one can imagine. The beneficiaries will be not be limited to large corporations, but the small and medium-sized manufacturers will also transform their shop floors. This quote from Punit Renjen of Deloitte sums it up “Today’s disrupted are yesterday's disruptors. And as Industry 4.0 gains traction and speed, how the world works and lives is being redefined, reengineered, and reinvented. The line between digital and physical is blurring”
I hope this article gave some useful information about advancements related to AI in the manufacturing industry.
Thanks for reading!!