The Predictive Maintenance Platform (PMP) is a condition monitoring application for the operating conditions of industrial equipment.
All manufacturing equipment with moving parts are subject to degradation which require servicing or component replacement, but traditional maintenance approaches based on set schedules ignore actual equipment condition. In Condition-based monitoring, maintenance is instead scheduled according to the estimated condition of the machine from inspection or from sensor data.
Predictive Maintenance builds on condition monitoring by feeding sensor data into dynamic predictive models for failure modes in an attempt to foresee maintenance requirements as far into the future as can be deemed practical. This can translate into more efficient maintenance planning, less machine down time and longer operating life through investment in system intelligence and other ERP data like equipment life cycle.
The process of evolving from Condition Monitoring to Predictive Maintenance begins with establishing information criteria and building appropriate systems to sense and deliver the data, followed by more intricate phases involving the optimization of thresholds through experience and historical data, and finally the implementation of predictive models able to provide accurate forecasts of the future condition of manufacturing equipment.
Time and frequency analyses of vibration data is especially useful for the identification of anomalies. Different analytical techniques can be used, which can include deep learning and AI technologies.
With respect to Figure 1, this solution is designed to get users to step three, in order to gain familiarity with the environment and equipment in which vibration or environmental analysis may be performed.
The architecture we propose is based on an STEVAL-IDP004V1 master board and up to four STEVAL-BFA001V1B smart sensor nodes, which export the following condition monitoring data over a serial protocol:
- environmental pressure, humidity, and temperature data
- time and frequency domain vibration data from the embedded accelerometer, processed by STM32F4 microcontroller
The data is collected and further processed in an Edge gateway consisting of an STM32MP157C-DK2 kit running X-LINUX-PREDMNT software, which includes the AWS IoT Greengrass service.
The DSH-PREDMNT dashboard completes the journey with a web-based tool to manage device provisioning, configuration, data injection and analysis, and simple thresholds for anomaly detection from a centralized Cloud service.
The AWS IoT Greengrass Edge Computing service allows local computation of Lambda functions on Edge gateway nodes with the same logic available on the Cloud to ensure continuity even when connection to the Cloud is unavailable; shadow devices on the Cloud are automatically synchronized with the Edge nodes as soon as connection is reestablished.
- Vibration monitoring data in the form of vibration speed (RMS), peak acceleration, and FFTs performed by STM32 core on data acquired from ST industrial accelerometer.
- Temperature, humidity and pressure data from ST environmental sensors.
- Condition monitoring example demonstrating Edge node processing in communication with a Cloud application via a secure gateway.
- End-to-end communication framework allowing Condition Monitoring platform to develop into a Predictive Maintenance solution.
- Further processing potential on Edge node with AWS IoT Greengrass and Lambda functions.
- Cloud Dashboard to register and provision the devices, configure a gateway for Edge processing, assign a gateway to a group of devices, analyze real time and historical data, and set thresholds to trigger alerts for particular equipment conditions.
- Free usage terms for a limited number of sensors and gateways, and for a limited time, as part of the DSH-PREDMNT Cloud application user license agreement.
- Based on STM32Cube and STM32OpenSTLinux expansion packages.
- Serverless deployment of the Dashboard application in user account through Cloud Formation tool.
RoHS Compliance Grade