Editor’s pickA New Look At The Industrial Internet Of Things As An Enabler For The Asian Water Industry
Perceptive Engineering,David Lovett,Water,Asia,IOT,IIOT,Diagnostic,Cloud-based,Maintenance
By David Lovett, Managing Director of Perceptive Engineering
IOT (& IIOT) Technologies have begun changing the world. The digital transformation of manufacturing, personal healthcare and smart cities is well underway; most multinational companies have a strategy to incorporate these highly connected tools to re-invigorate or completely transform their business operation.
Like the other sectors, the water and wastewater sector has set out strategies and produced demonstrable examples of IIOT, particularly for improved Asset Maintenance. In parts of Europe and the US they have also steadily turned to predictive maintenance to address the current and future challenges of improving asset performance and optimising sites and facilities. Given our insight into the Asian water industry, we believe these IIOT solutions may well have greater impact in the Asian water sector by enabling cost-effective monitoring and control, specifically for modular water treatment units, for clean, waste and biogas processes.
The advent of lower-cost sensors and widely available connectivity through wireless IOT devices, plus central cloud-based platforms, creates an opportunity for this new wave of predictive monitoring, control and maintenance systems that support the widely distributed wastewater treatment assets in Asia. The impact will be to ensure quality, security and cost-effectiveness of this infrastructure across the Asian region.
Figure 1: Dashboard showing process diagnostics
Most water companies worldwide are looking into methods of maximising performance of existing assets, which is becoming equally important as strategic investment in new capital projects. In some countries the regulators are making sure that particular attention is given to asset effectiveness and encourage operators to consider the pros and cons between replacing existing equipment and developing new maintenance strategies. This situation is particularly relevant for small, widely-distributed WWT sites which are inaccessible and therefore costly to access, as well as being critical for the local community to ensure a clean water supply is maintained.
Scheduled maintenance techniques have been used for some time to manage asset performance, particularly availability. Progressing from scheduled time-based maintenance usually means developing a preventive maintenance strategy, which relies on a combination of equipment usage history and OEM statistical benchmarks to determine a theoretical Mean Time Before Failure. The calculated result enables a company to programme maintenance, to prepare its staff and external contractors, places orders for replacement parts and schedule turnarounds and area shutdowns. Figures suggest that this has helped organisations reduce maintenance costs by a third. However, our experience has shown that preventive maintenance schemes alone do not capture the whole picture and can result in undertaking unnecessary (too frequent) maintenance or, more damagingly, missing a full asset failure when an early stage fault is not detected and addressed. How do we overcome this?
Perceptive’s View: We believe that for some years, water and wastewater companies have installed relatively sophisticated monitoring and base layer automation. These systems are used to enable the treatment works to reach and maintain efficient operating performance. However, historically many analytical instruments have fallen into disuse, often due to high ongoing maintenance costs or lack of readily available trained personnel. For this reason, it is important that preventative maintenance should include all assets, including the critical analytical devices monitoring the overall system’s quality and efficiency. This isn’t easy. Achieving connectivity to many analytical instruments from distributed assets has been an ongoing challenge for the water sector, which has tended to only use the instruments for regulator compliance reasons rather than using the process information to maximise efficiency or availability of the treatment facilities. So, has the situation really changed significantly to make a practical difference? We believe it has and the next section explains why.
Figure 2: Siemens IIOT connector. Photo credits to Seimens
PREDICTIVE MAINTENANCE METHODOLOGY
Predictive is different from Preventative maintenance. It relies on monitoring and understanding the actual operational condition of an asset and its historical use. Comparison of current operation in near real time against “normal” patterns is used to detect upcoming failures. Detection of such an event is the fi rst stage of predictive maintenance; however, a much more challenging task is fault diagnosis, which aims to specify the fault, its signifi cance, then help identify the corrective action required. Getting this right leads to improvement in scheduling, avoidance of downtime and, consequently, increased Overall Equipment Eff ectiveness (OEE) of the asset.
So, what has changed to create such a dramatic uptake in Predictive maintenance in the US and Europe? Underpinning the rapid growth in predictive systems is the advancement of sensor technology, particularly reduced cost, robustness and ease of connectivity wirelessly. These developments have been part of the wave of technologies under the umbrella of the Industrial Internet of Things (IIOT). When this data is connected to readily accessible cloud-based systems, there is a lower cost mechanism of performing statistical analytics to scan through massive sets of data, then generating insights that can be acted upon swiftly. Furthermore, this data analysis has been augmented by a combination of rapid and self-learning modelling methods, identifying faults reliably and with enough sensitivity to directly update real time maintenance systems.
Perceptive Engineering, a Singapore-based company engaged in advanced process control, develops model-based predictive solutions for various sectors including water and wastewater. Perceptive’s software platform now allows rapid digital modelling for the key process units on wastewater treatment plants, which has been incorporated into an AI monitoring system to identify abnormal process conditions and analytical instrument problems. The system uses a combination of statistical detection tools with a conditional machine learning engine to pinpoint faults and suggest corrective actions. The system is derived from many years of practical experience in monitoring, controlling and optimising these process units, and building tools that can improve effi ciency even when assets behave abnormally.
Perceptive Engineering has recently developed an open cloud-based application, WaterNeuron™, available on AWSCloud Computing, Microsoft Azure and Siemens’ MindSphere open platforms. The WaterNeuron™ App aims to reduce the costs and expertise associated with modelling, by utilising powerful self-learning techniques which capture process behaviour over a period of time and transform the data into a model representative of that plant. Alternatively, a “turbo-charged version” is also available, which uses existing process unit models to give an approximate representation of the plant, to speed up creation of the fault detection database. The prime advantage of the cloud-based platform is that the end user only pays for the data being analysed and the insights that the analysis brings. Additionally, through the use of the Siemens’ MindSphere platform, the system adheres to the highest data security levels and facilitates connectivity to other open Apps.
The App makes use of the extensive data infrastructure around the water and wastewater plant, which is augmented by machine learning capability and amplifi ed by existing process knowledge.
Adopting this cloud-based approach enables organisations to start seeing the benefi ts as quickly and as widely as possible across their asset base, for a cost that is customised to the treatment plant size or the volume of data analysed. This pay for use/result service reduces investment in both cost and time. For example, a typical UK water company has more than 500 sites. Installing statistical monitors across all their sites would require a massive capital expenditure. A cloud-based approach to asset condition monitoring and predictive maintenance, with pay-per-use pricing, allow cost flexibility and effi ciency. A similar scenario can develop in Asia with modular distributed WWT units located at point of need and maintained remotely using IOT techniques.
Figure 3: WaterNeuron Predictive Maintenance Dashboard
A Predictive Maintenance Assistant is a key feature of Perceptive’s WaterNeuron™ APP, which can be adopted to create or augment existing scheduled maintenance systems. The concept underpinning the App is its configurability and self-learning capability, using signals that are readily available on-site, either in real-time or from daily operator entries.
In brief, the system works by capturing critical information in an Edge Server running Perceptive’s software. WaterNeuron™ utilises advanced, robust statistical tools to create a ‘rolling’ window of reliable data, using a digital model that is automatically created within the software. In this way, it remains adaptive, keeping the model up to date as the process changes. This self-learning capability can compensate for changing influent characteristics (diurnal flow patterns, storm and first-flush events, population increase), as well as shifts in operating mode (e.g. from carbonaceous to nitrifying treatment), all of which can substantially alter the relationships between process variables.
Whilst the model adapts to prevailing conditions, it also has the ability to ignore any data it determines as untrustworthy, such as instrument issues pinpointed by the “Data Quality Monitor”. The system also reduces data noise levels that would otherwise result in erroneous faults. This powerful combination of data pre-processing, advanced classification techniques and machine-learning algorithms generates a robust process monitor providing reliable and optimal predictive maintenance advice.
The App can reference a simple model of the process to enhance its ability to quickly detect, identify and report known fault conditions. The user can adjust the data window for the self-learning algorithm, to ensure detection remains as sensitive, flexible and responsive as possible. The App can be configured to send emails and alerts to mobile devices and transfer information into a readily accessible relational database. Utilising an App that functions as a Predictive Maintenance Assistant will lead to consistent, accurate monitoring of water quality, resulting in improved environmental compliance.
In summary, Asian water companies that incorporate next-generation digital platforms and predictive maintenance techniques into their operational plans, will be best placed to capitalize on the major opportunities for flexible modular distributed wastewater treatment units in the region. These units off er a new modular approach providing safe, environmentally-robust treatment for a widely distributed population, without incurring the enormous capital overhead of a sewer network.
* Article can be found in Water & Wastewater Asia Jul/Aug 2019 issue.