Predictive Maintenance of Industrial Machines using TinyML
In any industry or manufacturing facility, there will be a lot of moving parts that are essential for the movement of man,
materials and machines which are like the pillars of the facility. When an industrial machine halts due to a breakdown,
it causes a huge loss as it will take a lot of man-hours to fix the issue. For example in a factory producing 100000 units using 5 machines,
if one of them breaks down then there's a 20% down in production. Which is a huge loss. If there was predictive maintenance then it would
have detected the failure or future breakdown just by using the sensor data such as vibration, microphone and temperature. How predictive maintenance
works is based on the anomalies in the sensor data it predicts the future breakdowns and suggests the repair or action towards fixing the issue beforehand.
By this one can save a lot of time and effort.
The solution:
Block diagram of the solution.
The solution involves edge computing powered by tensor flow lite to detect the anomalies using the machine learning model running
on the low power PSoC
6S2 + AIROC
This whole unit can be battery powered and only when there is an anomaly detected it can send the data over BLE to the nearest gateway i.e Raspberry pi 4 in our case.
The sensors involved: TLV493D hall effect sensor will be used to measure the change in the magnetic field. This sensor data can be used to measure the angle of the rotor at a
given instant of time. This will help us to detect the anomalies in the rotors, rpm, etc. In order to train the ML model, we will be using the edge impulse studio and the model running on
the low power microcontroller will provide the inference based on the sensor data.
Implementation diagram
About me:
I'm Vishwas. I work as a Senior Research and Development Engineer at Tejas Networks, a leading Network equipment manufacturer here in India. My work revolves around the design and development of complex FPGA based high-speed boards for the network equipment to cater to the needs of 5G. Previously I worked as a Research Assistant at Robert Bosch Centre for Cyber-Physical Systems, one of the largest Cyber-Physical Sytems in the world where I worked on hardware design and development for robotics and drones i.e especially on autonomous drone charging pads, at the Indian Institute of Science, Bengaluru. I'm keen on research and development and I love electronics and building some cool stuff out of it. I write about my adventures in electronics in hackster.io and many other platforms. I have participated in many competitions both online and offline about hardware hacking and won many awards. Till now I have won 6 awards in hackster.io itself. My interests are embedded systems, circuit design and product design. To know more about me please visit my portfolio website http://vishwasnavada.github.io One more added advantage is I have already worked on the design and development of wearable healthcare devices and written a detailed article on the https://hackster.io/vishwasnavada
Also, I have written articles and participated in design challenges in the element14 community and won in some too. Thanks a lot for your time.
Regards,
Vishwas Navada B