With apologies to Jerry Lee Lewis this is neither about him nor Rock and Roll, it is about vibration in motors.
Abnormal shaking or vibration in motors is a key indicator of a probable or pending failure. In fact, there is an ISO standard for maximum allowed vibration for given machines. ISO 10816 defines the vibration levels which will tend to cause early failure or “wear out” (Figure 1). There are many ways of detecting vibration with external sensors and boxes that monitor these motors. A more cost-effective method is evolving, Artificial Intelligence (AI) for predictive / preventive maintenance.
Figure 2 shows a typical motor inverter design for driving a Brushless DC Electric Motor (BLDC) or induction motor used in industrial and consumer products. You'll notice the design is a very minimalist design and likely cannot bear the cost of additional sensors. What can we do?
If we step back and look at the information available to us in the motor algorithm, we have speed and current. Can we utilize this to detect issues? It is not enough to just look at instantaneous speed and current as the algorithm may do. Some motor installations may have given levels of “unbalance” which may be acceptable and show up in this speed and current data. We need to evaluate and detect “abnormal” speed and current data. Enter AI!!!
Renesas' Reality AI solutions combine advanced signal processing with artificial intelligence on inexpensive edge nodes, for example this vibration sensing motor control solution. Reality AI solution can be added to any motor control solution. It will take the speed and current information and evaluate the data through AI functions. Since Reality AI is targeted at “edge nodes”, it requires very little resources in the MCU, just a little additional memory to execute the AI functions. No additional hardware is required.
Reality AI will detect “abnormal” operation based on the already available data. If your motor control has a vibration that is normal to its operation, this can be trained out with the main focus of finding “abnormal” operation.
If we take a quick peak at Figure 1 again, we see additional sensors as “optional”, a MEMS accelerometer sensor and audio sensor. Since Reality AI uses very limited memory resources for its AI implementation, it is easy to add additional sensors for additional accuracy, for example the MEMs Accelerometer may be mounted to monitor vibration / shaking on a specific axis, for example along the motor axis for very specific failures such as bearing slop along the shaft.
One thing to note, we still need to get the data off the system (to HOST MCU for example) and Reality AI provides an easy-to-use MQTT-based API for integration with dashboards and workflow systems.
So, let’s detect when a “whole lotta shakin” is going on when we don’t want it and prevent failure in critical systems operation or reduce cost by reducing down time in production equipment.
For more information on Reality AI visit https://www.renesas.com/products/microcontrollers-microprocessors/reality-ai
For more information on Motor Control Winning Combinations and other solutions that will reduce your time to market visit https://www.renesas.com/win