DATA-DRIVEN QUALITY ASSURANCE OF SENSORS IN IOT/IIOT THROUGH SELF-DIAGNOSIS AND SELF-HEALING (SDSH)

Authors:
Dr CUI Shan, Scientist III, Acoustics & Vibration Laboratory
Dr MOU Jianqiang, Scientist I, Acoustics & Vibration Laboratory

Internet of Things (IoT) has been generating a lot of interest in digitalisation. Particularly for industrial applications, the Industrial Internet of Things (IIoT) has seen adoption in an increasing speed. Vast amount and variety of sensors have been and are being deployed to generate insights and realise automated control to improve process and product/service quality, to improve efficiency and eventually to improve profitability, sustainability and customer satisfaction. To achieve the desired outcomes, the quality, i.e. accuracy and reliability of sensing data in the IoT or IIoT is crucial.

At National Metrology Centre (NMC), we have developed a data-driven approach for sensing data quality assurance in sensor networks for IoT and IIoT, namely Self-Diagnosis and Self-Healing (SDSH). Self-Diagnosis refers to autonomous and in-line monitoring and diagnostic of sensor health using a metrological ruler, i.e. the measurement uncertainty of the sensors. Self-Healing refers to the subsequent auto-compensation of error readings identified based on metrological principles.

SDSH minimises lab-based calibration, which is laborious, interruptive to operations, costly and not feasible for IoT/IIoT applications due to the amount of sensors involved. It also enhances sustainability of IoT/IIoT as it minimises the resources needed for the maintenance of the sensing data quality and ensuring long-term reliability of the sensing data.

Here we would like to highlight three applications/projects which NMC is working on.

Application 1: INTELLIGENT BUILDINGS

Green buildings require smart design and intelligence in building condition and control for better sustainability. The intelligence comes from the many sensors installed in the building such as for indoor air quality (IAQ), temperature & humidity, energy meters, and so on. For the building controls to be effective, reliable and minimise energy consumption, the sensing data must be reliable and accurate to the acceptable level.

To help buildings achieve the sustainability goal, NMC’s team is deploying IAQ sensor networks with SDSH function to drive building’s fresh air ventilation control to achieve effective demand-controlled ventilation. Through continuous monitoring and correcting the sensing data by SDSH automatically, energy efficiency in fresh air ventilation is achieved without sacrificing indoor air quality and the long-term energy saving performance is sustained as a result of minimised sensing error.

App 1 - Intelligent Buildings

Green buildings require smart design and intelligence in building condition and control for better sustainability. The intelligence comes from the many sensors installed in the building such as for indoor air quality (IAQ), temperature & humidity, energy meters, and so on. For the building controls to be effective, reliable and minimise energy consumption, the sensing data must be reliable and accurate to the acceptable level.

To help buildings achieve the sustainability goal, NMC’s team is deploying IAQ sensor networks with SDSH function to drive building’s fresh air ventilation control to achieve effective demand-controlled ventilation. Through continuous monitoring and correcting the sensing data by SDSH automatically, energy efficiency in fresh air ventilation is achieved without sacrificing indoor air quality and the long-term energy saving performance is sustained as a result of minimised sensing error.


APPLICATION 2: SMART FACTORIES

App 2 - Smart Factories

Industry 4.0 opens up opportunities to manufacturing companies to implement digitalisation, maximise automation, improve productivity and quality, reduce human error, increase competitiveness and eventually save cost and enable long-term growth. One of the core technologies in Industry 4.0 is IIoT and the sensors in IIoT. To achieve the full potentials of IIoT, its sensing data accuracy and long-term reliability must be ensured with minimal resources for sustainability.

Temperature and vibration sensors are among the most commonly used for product quality monitoring/prediction, machine condition monitoring/predictive maintenance and more. NMC’s team is further developing SDSH for temperature and vibration sensors in a manufacturing shopfloor environment. Machine learning method is being applied to learn sensor behaviour and to separate it from unwanted disturbances and noises. The target is to provide reliable sensing data with minimal disturbance to manufacturing process and minimal resources including manpower, time and service cost.


APPLICATION 3: STRUCUTRAL HEALTH MONITORING

App 3 - Offshore Structures

Offshore structures exposed in extreme ocean waves and winds from time to time are subject to structural ageing and degradation. An efficient way for ensuring the offshore structural integrity and minimising the risk of incident is utilising the structural health monitoring systems, in which a large amount of strain, vibration and acoustic sensors are deployed. Since these sensors operate in harsh offshore and marine environments, a critical issue is the data accuracy and fidelity of the sensor measurement in continuous operation over long term. As these sensors are generally embedded in the offshore structure, the traditional approach by shutting down the offshore structure operation or dismantling the structure for calibration of the sensors isn’t favourable. 

The team at NMC are developing data-driven technologies and solutions to assure the accuracy and fidelity of the sensor measurement data in-situ, without unnecessary interruption to the operation of the offshore structure. The offshore structural operational process is typically non-stationary, random and time varying, which makes the diagnosis of sensor health and identification of the sensor fault from structural heath degradation more challenging. Hence state-of-the-art data analysis algorithms for data correlation and signatures extraction corresponding to the sensor health status is being developed. With training data traceable to physical measurement standards, senor health diagnosis is being achieved by machine learning and artificial neural network. For possible deterioration of the sensor health leading to sensor drift, self-healing, i.e. autonomous compensation of the drift will be implemented based on metrological principle.


More Applications:

SDSH can be adapted and applied to more areas such as autonomous vehicle sensors and water quality sensors.

Contact Dr Cui Shan at cui_shan@nmc.a-star.edu.sg or Dr Mou Jianqiang mou_jianqiang@nmc.a-star.edu.sg for discussion and collaboration on your interested application areas.