In short, there is an urgent need for a low-threshold entry into predictive maintenance, an opportunity to explore and experiment without obligations or consequences. That option is now available, in the form of Amazon Monitron, its Amazon’s predictive maintenance solution which is connected with the Amazon SageMaker, a cloud machine-learning platform. Late last year, Amazon launched the Monitron, an end-to-end system that uses machine learning to discover problems and deviations in the operation of assets. It comes with wireless sensors to collect vibration and temperature data, a gateway device for secure data transfer to the Amazon Web Services (AWS) cloud environment, and data analysis functionality.
The “internet” in this IoT-based system is provided by wi-fi connecting the gateway with the AWS cloud, and BLE (Bluetooth Low Energy) facilitating data transfer from the sensors to the gateway. Included in the offer is an app for getting started with predictive maintenance using machine learning (ML), and receiving data reports and alerts to potential failures. No experience required, Amazon claims, and hence no dedicated engineers and data scientists to enter the predictive maintenance age.
Amazon Monitron is an end-to-end system that uses machine learning to detect abnormal conditions in industrial equipment and enable predictive maintenance.
(Image source: aws.amazon.com)
Indeed, the sensors and the gateway are easily installed, the app has a user-friendly interface, with useful data graphs of sensor readings, and it sends a push notification when something unusual has been recorded. So yes, experimenting with the Amazon Monitron will give you a feel for what it takes to start with predictive maintenance, and what it will bring you. However, for serious implementation in industrial practice, the current version of the Amazon Monitron has only limited functionality, and comes with too many downsides.
For one, the app is only available in an Android version; “iOS app coming soon” says the AWS website. The sensors have no mains connection or data communication cable connection. The battery that feeds them has a lifetime of three years at a data sampling rate of once an hour. With the higher sampling frequencies often required in industrial practice, the correspondingly shorter battery lifetime will pose operational challenges.
Here we’d like to compare Amazon Monitron to Sensorfy. And point out some key differentiators that are relevant to OEMs. In contrast to Amazon predictive maintenance solution, we leave the ownership of data to the user, whether or not data are stored and processed in the cloud. The cloud not always is the best place to be.
For example, vibration monitoring usually requires high sampling frequencies, and consequently generates massive data. Cloud processing would require too much communication bandwidth, storage capacity, and processing energy. Here, edge computing is a viable alternative, and nowadays “smart” sensors are available for this.
In the start-up phase, central data processing can be useful for algorithm development, but once the machine learning (ML) model has been finalized, edge computing can be much more efficient and cyber-secure.
Working with the Amazon Monitron offers you a simple introduction to predictive maintenance and the benefits it can bring you. It works with any dataset you want to analyse. You can familiarize yourself with data acquisition using an Internet of Things system, and with data analysis based on Artificial Intelligence.
You can observe how uncommon signals are detected early, so that you can make interventions to prevent incidents. However, soon you will realize that your real-world operation requires a more dedicated system. So yes, do get started with a low-entry predictive maintenance application, and start experimenting. But please keep in mind that a professional consultancy and implementation partner will help you get the most out of such a system.
Predictive maintenance can be applied to existing assets, for example by applying add-on sensors, but even better, it is incorporated right from the start in the development of a new asset, enabling optimal sensor integration.
If you want to get started with predictive maintenance, get free access to our E-book: Predictive Maintenance from Scratch.
Hopefully this article shed light on the importance of asset upkeep in your organization and the different strategies you could go with. We believe the future lies within predictive maintenance and have dedicated ourselves to developing custom made sensors. Our smart sensors are designed and build which collect data precisely and then using data analytics we can detect equipment failures and provide recommended actions to improve productivity. Our sensors match software solutions to help machine manufacturers and their clients run their production process of industrial equipment as efficient as possible. As it is, predictive maintenance is of core value for the highest efficiency and maximum result.
Interested in knowing more about the upkeep of your assets? Get in touch with us.