The core effort in improving machine performance is collecting data. The data provides the foundation for developing advanced digital automation. Artificial intelligence also plays a key role in improving machine performance.
We caught up with Dan Barrera, sales product manager for ctrlX AUTOMATION at Bosch Rexroth to get his view on using automation and AI to improve machine[performance[performance
How can you use AI to improve automation?
Dan Barrera: From my point of view, everything around Industry 4.0, IoT, and digital transformation starts with machine data collection. With data, we should have the foundation to start thinking about cost/time reduction and quality improvement. However, analyzing historical or live production data can take many hours and valuable human resources. This is where artificial intelligence (AI) can help improve automation systems.
AI can analyze and run several tests using the collected machine/production data, with the purpose of identifying the following:
-
From the machine builder point of view (OEM): weak points in a machine or even improvement on the design to increase speed and accuracy on the system plus overall improvement in efficiency and quality.
-
From the end-user point of view: safety risks, bottlenecks, and the ability to predict when components are about to fail in a machine or production line (predictive maintenance).
What processes can be enhanced with AI?
Dan Barrera: Machine builders can input this machine/production data into digital models (digital twins) and AI can be applied to run several simulations to understand how an automated system will behave over time and under different conditions (faster speeds, heavier paid loads, longer running times , higher/lower temperature conditions, etc.). The goal for the AI is to identify weak points that cause breakdowns and provide options to improve machine design and performance, all of which in a reduced amount of time since computer CPUs can process data much faster than humans.
For end-users, the idea is similar, but they will apply the AI to the production data and look to improve safety conditions, detect bottlenecks and for predictive maintenance. The overall goal is to decrease time, cost and increase quality and safety.
Explain the AI aspects of predictive maintenance.
Dan Barrera: If we talk about smart factories, digitalization, or Industry 4.0, then as a minimum we must consider machine and production data collection. With historical (minimum 6 to 12 months) and live production data, AI can determine when a component or a system is not running at 100% efficiency. First, we would need to teach an AI what optimal conditions (green), anomalies (yellow) and defects (red) look like. Then the AI can use the available data to detect when a system or component is transitioning to non-optimal behavior (anything less than 100%). Then, AI can analyze the data to indicate a future component failure and changes in the manufacturing process that exceed the limits of the any system (payload, speed, etc.) which could cause early machine failure.
How can manufacturers leverage current applications to avoid disruptions?
Dan Barrera: By means of digital twins, conditioning monitoring and predictive maintenance, manufacturers can utilize these digital tools (applications) to improve efficiency and minimize disruptions:
-
Digital Twins are a digital representation of a physical object (ie components or a complete manufacturing line). Digital Twins can help emulate the physical behavior of real life electrical and mechanical modules. For example, end-users can utilize digital twins to test (offline) new or existing production programs to understand the behavior before they get loaded into the line. By doing all the work offline, users can avoid system failures (due to an untested code), save energy, save waste of material and avoid shutting down a production line for testing purposes.
-
Conditioning monitoring is the approach that we can use to make sure the system is running like it should. Using different sensors and measurement systems, we can use conditioning monitoring to check the health of a system. If we apply algorithmic formulas or AI, we can predict (predictive maintenance) the exact moment of when a part or a system is going to need maintenance. The goal is to monitor the overall health of the system and avoid an unscheduled production shutdown. The key is to be more proactive than reactive.
What are some of the smart capabilities that can improve automation?
Dan Barrera: With the most recent advances in automation technology (ie controls & drives, AI, sensors, vision, communications, safety, etc.), we’re able to design and engineer more open, flexible, and scalable systems that can help us improve automation solutions or offer complete automation for smart factories of the future.
From my point of view, some of these capabilities that can improve automation are:
-
Industrial Communication (5G). With the need for data collection and connectivity within the shop floor for data exchange, 5G will become an essential component for the “Factory of the Future”. This wireless network is meant for high data speed, low latency, secure and high network capacity.
-
LiDAR Technology. LiDAR sensors are an important component for mapping and localization of applications. For industrial manufacturing, we see a focus for this technology in autonomous mobile robots (AMR). It is important for these autonomous vehicles to safely detect moving objects and personnel around the production shop floor.
-
High-Speed Smart Conveyance: Speed and dynamics are driving modern factories. The aim is to ensure short throughput times by optimizing material flows. Linear motion systems are the key here. These high-speed smart conveyor systems bring flexibility and high speed to manufacturing – with maximum precision and flexibility.
What is the net gain in applying AI to manufacturing automation?
Dan Barrera: This is very specific for each machine builder or end-user. There are many levels at which you can apply AI and that defines the level of investment, effort, and ROI.
For example, machine builders can apply AI to their systems to find weak points and improve their overall system design. The gain here is a higher quality machine and getting them faster to market. Machine builders can also apply AI to the main machine application code to improve the machining process (reduce time, increase quality, etc.). However, this will involve higher investments and longer development time, but the gain is a smarter machine that can produce higher quality products, with less cost and time.
For end-users, the gain is probably a bit transparent. If we talk about predictive maintenance and AI at a basic level, we can calculate the benefit by adding the savings of downtimes versus the investment of these smart approaches. However, there can be a greater overall benefit by applying AI + digital twins to run offline tests with the purpose of improving the manufacturing process.
Each layer of AI implementation has a cost and a complexity that affects the ROI. Business owners need to define their pain points, needs, challenges, and compare those to the different digital solutions available in the market. Basically, it’s about defining a digital transformation plan before understanding the full benefits of AI for the business.