Predictive Maintenance IoT Case Study: ML & Edge Computing Prototype

  • IndustryHigh-Tech
  • Solution IoT


SaM Solutions and Toradex marketing + R&D departments were tasked with discovering how the latest edge computing and machine learning technologies could be used in an IoT predictive maintenance context. This predictive maintenance IoT case study outlines the various technological challenges, and how SaM Solutions and Toradex were able to overcome them.

In an industrial setting, the overall efficiency and production output of equipment relies heavily on the performance of machine motors. If SaM Solutions and Toradex could successfully develop an IoT-based predictive maintenance system, maintenance engineers would have the ability to not only monitor production assets 24/7, but be alerted of any potential machine malfunctions or outages before they even occur.

The goal was to have the latest Toradex board connect to the Amazon Cloud, where it would then apply machine learning algorithms to the data collected from the embedded sensor through edge computing capabilities. Through achieving this, the resulting predictive maintenance system would be able to evaluate motor behavior, and then accurately predict whether malfunctions or outages would occur.

Toradex specializes in embedded computing technology, offering ARM®-based System on Modules (SOMs) and Customized SBCs. Complemented with direct online sales and long-term product availability, Toradex offers direct premium support and ex-stock availability with local warehouses.


The edge computing prototype was created using Amazon Greengrass technologies implemented on a Toradex Aster Carrier Board with a Colibri Module, which then connected to the Amazon Cloud platform to apply machine learning algorithms to the data collected from a vibration sensor.

SaM Solutions successfully built a prototype in which the sensor attaches to the DC brushless motor and passes specific motor parameters to the Toradex board. Amazon Greengrass is running on the board and is taught to recognize the states of the motor on the fly (e.g. stop/run/malfunction).

The machine learning model is deployed from AWS to the device using Amazon Lambda to adjust the motor behavior recognition algorithms, after which it can run on the board without needing connection to the Internet and AWS (i.e. as an edge computing device).


Aster Carrier Board, Colibri iMX6ULL 512MB Wi-Fi, DC brushless motor, vibration sensor MPU-6050
Services & Libraries
Amazon Greengrass, Amazon Lambda, TensorFlow, Keras, Amazon SageMaker


This successful implementation of an IoT predictive maintenance device has the ability to save industrial and manufacturing organizations potential lost revenue resulting from mechanical failures and machine outages. An overview of this predictive maintenance IoT case study can be viewed in video form here.

SaM Solutions instituted a dual-technology approach to creating a solution that allows a DC brushless motor to be analyzed as a standalone solution, while also being able to connect complex machine learning technology to a cloud-based database.

More information on our partnership with Toradex is available on this page.

Please fill out contact information or call us directly at (857) 777‑6073