Good practices for monitoring a network of IoT devices

Growing a high performing IoT network

To shape the (sustainable) future of water, Shayp offers water monitoring and real-time water leak detection services based on the deployment of wireless data loggers for fluid meter monitoring. Such a solution goes beyond the initial development of its hardware. Indeed, Shayp is actually managing a growing network of IoT (Internet of Things) devices which entails multiple challenges. This article offers insights on Shayp’s experience in the adoption and management of IoT technologies in order to offer (water consumption) monitoring services, and shares good practices following the contribution of Shayp in a research project of IoT monitoring.

Challenges encountered by Shayp in IoT network management

Given the increasing number of its IoT water sensors being deployed, monitoring this growing network is a topic of concern for Shayp in its pursuit of ensuring high quality services to its customers. Indeed, an IoT network entails multiple challenges. 

Some devices might be malfunctioning for various reasons such as poor network signal for instance in case of, for instance, an external intervention (bad weather conditions, antenna relocation by the operator, device accidentally hit), or simply production defect. Such issues are problematic for a company as they are not always easily detected. This means that the data collected and then analysed might present inaccuracies, impacting the quality of services provided to the customer.

On a financial level, the opportunity to get a real-time overview and control over the IoT network status allows Shayp to analyse the profitability of its deployed devices. As a matter of fact, if device replacement is too frequently needed, the resulting higher costs reduce the profit margin.

Constantly willing not only to improve the quality of its offer and its business sustainability, but also to develop and contribute to scientific and technological progress, Shayp regularly partners and works with various universities and research institutes.

Following this ambition of taking part in scientific research and optimising its IoT network management, Shayp seized the opportunity to partner with the research laboratory on Artificial Intelligence of the Université Libre de Bruxelles (IRIDIA) and Degetel Belgium. Shayp was the first candidate for beta testing a monitoring tool as part of a research project focused on a new method of IoT monitoring.

Best practices: using Machine learning for IoT Monitoring

If you know the behaviour of your devices, you will probably already have some simple and efficient rule-based scripts to track potential trouble. But how to know if all the potential scenarios are covered? For several years, advancements in Machine Learning models have brought new possibilities in multiple fields, with great successes for instance in image and speech recognition, medical diagnosis and classification. Can this innovative approach be used for monitoring and improving the quality of service of IoT networks? This is the question that IRIDIA and Degetel Belgium challenged with the help of Shayp by developing a framework, called Philéas AI, to analyse the activity of IoT systems and to identify possible issues via anomaly detection based on a set of machine learning methods.

In its essence, the Philéas service listens to the incoming traffic and learns the properties of the IoT stream. It can then generate a prediction about the future traffic and compare it with the actual traffic observed. If a noticeable difference happens, an anomaly has been discovered and reported in the developed dashboard. It can be very useful for quickly detecting new scenarios for which the customer currently has no solution.

The core engine uses autoencoder neural networks. The advantage of autoencoders is their cheap training phase, and their ability of performing instantaneous predictions. It is considerably useful for real-time and pseudo-real-time applications. Furthermore, the service focuses exclusively on the analysis of metadata, machine-to-machine logs without business logic and user information. Thus, the absence of sensitive data makes the calculation more GDPR-friendly.

Shayp wants to push the science forward

Taking part in this project, Shayp has learnt a lot and continues to dedicate all its efforts towards always having the best customer quality. In that perspective, partnering with top of the game experts and researchers helps Shayp not only to be a top-player in its industry but also to contribute to scientific research and push the state of the art forward with advanced technology.

-------------------------------------------------------------

If you are more interested about the technology, an academic paper is available here: 

Philéas: Anomaly Detection for IoT Monitoring

Other useful links

• Shayp - Water consumption anomaly detection in real-estate - https://www.shayp.com
• Philéas - Preventive Security for IoT monitoring  - https://phileas.ai
• Degetel - Agence digital et conseil en services numériques - https://www.degetel.com