ABB Ability™ Predictive Maintenance is a service that harnesses the power of cloud computing, machine learning and ABB’s expertise to propose highly targeted maintenance actions for critical drive applications – before any problems occur. This gives better availability, longer process uptime, and peace of mind, says Anthony Rawson, UK service manager, drives and controls, ABB.
What is ABB AbilityTM Predictive Maintenance?
ABB AbilityTM Predictive Maintenance is a condition-based service to improve the reliability of critical drives in demanding environments. The service harnesses the power of cloud computing, machine learning and ABB’s expertise, together with a remote monitoring system. Through predictive analytics and active monitoring, the future health of the drive is understood and issues are actioned before any issues arise.
As a term, predictive maintenance has long been associated with electric motors, so why is it so new
to variable speed drives?
Predictive maintenance is a term that has been around for decades. In the world of motor-driven applications, the term is associated with mechanical equipment such as electric motors. Predictively maintaining a motor extends to the understanding of the condition of highly visible components like bearings and the impact of vibration.
It is analogous to the tyre wear on a car. By looking at the condition of the tyre you can safely and confidently predict when a replacement is due. Predictive maintenance on a mechanical motor is one dimensional, you practically see what maintenance is needed, without having to worry about other dimensions that you may not be able to physically see.
Yet, as surprising as it sounds, predictive maintenance has never accurately been applied to solid state electronic devices like variable speed drives (VSDs). Even though VSDs have been packed with digital sensors for the past 20 years, understanding how to use the data available and what to forecast was proving a challenge.
We know that customers are monitoring their motors but do not really have sufficient intelligence on the condition of the drive.
Why is predictive maintenance such a challenge for variable speed drives?
Static electronic converters, like VSDs, are packed with discrete components like capacitors and IGBTs. And it is virtually impossible to accurately predict when such devices are likely to fail as, unlike with the bearings in an electric motor, you have no way of visually inspecting the components. With a VSD there are multiple dimensions – including the impact of temperature, current and voltage – that require a lot of
calculations and understanding especially as many of the applications on which they are installed can be different and quite demanding
So, what has changed?
W e have combined new technologies such as cloud computing, advanced algorithms and machine learn-ing with ABB smart engineering expertise to enable us to create the first predictive maintenance service for VSDs used in critical applications. In short, predictive maintenance is a direct outcome of the availa-bility and application of big data.
The advanced algorithms, for instance, are developed based on decades of real life experience in multi-ple applications across many industries. Together with analytics, this enables us to estimate drive com-ponent lifetimes and propose maintenance to reduce the risk of failure.
Our data resources stretch back to the earliest days of drives, and across multiple power-involved indus-tries, to yield unique, actionable insights that deliver digital advantage for our customers.
Today’s drives are highly reliable, so why is predictive maintenance needed?
Most drives are famously reliable, lasting more than 20 years with the correct maintenance. You can have hundreds of drives on a plant and have the occasional breakdown. Yet, when a drive does fail the cause is often not clear. By tracking the performance data of each drive, we can build a picture and identify trends that pinpoint any weak drives in the installed base.
What are the limitations of traditional VSD predictive maintenance practices?
It is relatively straightforward to predict or forecast the failure of a drive if considering only a handful of parameters and knowing the drive’s history. Earlier our calculations were all based on probability theory, resulting in service actions every nine years. This was coupled with physical site visits to gather fault information and manually establish what will happen to the drive.
Traditional monitoring gives the user detailed information about one particular part of the plant. It is like looking at the equipment through a keyhole – all that can be seen is one small detail, not the piece of equipment as a whole.
Using data alone, it can be easy to overlook failure modes that depend on the eyes and ears of the operators. Likewise, it can be easy to miss how observations from the field relate to the condition monitoring data.
While most maintenance data can be collected automatically, this does not apply to everything. The digital maintenance concept differentiates between conditions that can be picked up through data and others that still need hands-on execution on the equipment.
With predictive maintenance, many different streams of condition data are combined to create a rich context for the maintenance team.
How does digital predictive maintenance for VSDs work?
The service enables users to maximise operational efficiency through highly targeted maintenance actions for critical drive applications. The latest machine learning algorithms and analytics made it possible to estimate drive component lifetimes based on wear-out and stress monitoring.
When you are trying to consider multiple parameters then you need a sophisticated tool such as machine learning. Today, our engineers can gather data on how individual components are ageing and the relationships between thermal effects and other parameters. Using drive data stored in the cloud, the manual and labour intensive data gathering of traditional maintenance assessments is removed.
For instance, you may have an operational point with a defined set of curves based around the current, voltage, temperature or environmental circumstances. If any of these parameters varies, then a new operational point is created. Through machine learning we can accumulate the knowledge which will show us if anything is wrong and not only tell us what to do to correct it but assist in the fix.
With predictive maintenance, the end user now has real time visualisation on the future status of the drive, enabling health to be more accurately, and therefore cost efficiently, predicted.
This approach allows to propose precise maintenance time and actions to reduce the risk of failure. Now even the slightest anomalies in the process can be found and reported. More reliable availability of drives can translate into improved performance and, therefore, return on investment.
Where do you start to extract the data?
Because of the complexity of a drive, ABB needs to gather data gradually from specific sections. This starts with the power stage, as this is the most valuable and difficult to exchange if it fails.
For instance, the DC capacitor is one of the components within the power stage and temperature can impact on the applied voltage. According to the maintenance schedule, the DC capacitor is changed every nine years, but sometimes such replacement can result in a short that can affect the IGBT power semiconductors.
As such, ABB is currently gathering the condition data of the capacitor. However, IGBT technology is continually developing, with devices getting physically smaller, yet more powerful. As such, it is a critical component, the parameters of which need to be monitored.
Is this an instant game changer for VSD users?
It is important to realize that it is still only a first step, even though it is a significant step forward. While drives have been digital for over 20 years, it is only with the advent of cloud technology and data lakes that we are now able to collect different types of data and analyze them on a large scale.
It is still early days and we are on a rapid learning path. However, we can now channel that data and using algorithms and machine learning techniques we are able to predict deviations. But machine learning, by its very nature, is a gradual process. Just like it would take a human years to become expert, it will take a period of time for machine learning to gather detailed knowledge.
For predictive maintenance to be truly effective, knowledge needs to be accrued over time from many drives operating in different applications. This way, by combining the different data sets within the same platform we may start to discover patterns and findings that we do not even know about yet, such as components working in a different way. The more data we have, the better findings we will have. So future versions of our predictive maintenance service will be even more powerful.
So, at this stage predictive maintenance is a concept-based approach as data needs to be collected and collated over several years, as little can be predicted with, say, one year’s results. ABB is establishing long-term relationships with its customers. During this relationship ABB is proactively monitoring the drives and alerting customers of any challenges that it discovers.
How long will it be before predictive maintenance is useful?
It is already! We have a number of trials in place worldwide and the outcomes for our clients are very positive. We have been able to anticipate and prevent failures of highly critical applications. As a result of our findings, we have advised our clients on the way to optimize drive usage and maintenance to reduce risks on their operations continuity. This has saved our clients hundreds of thousands of Euros in repairs and lost production.
Predictive maintenance, and the broader range of digital services ABB is developing, will bring us closer to our customers: digital services makes ABB technical expertise more readily available to clients to capi-talize on.
However, it is also true that the more data we analyze from the field, the more knowledge we and our clients accumulate and the more accurate the service will become. This is a process that will take a few years, but we are confident it will bring increasingly greater benefits to our clients.
When the iPhone came out it wasn’t really a game changer. It took time for cloud computing to emerge before the iPhone truly came into its own. Likewise, our predictive maintenance service will evolve. The power of cloud computing, the speed at which artificial intelligence (AI) is emerging and the strengths of machine learning means for the first time we can start to calculate what is happening within a drive ear-lier.
What is driving the interest in predictive maintenance?
The service is generating much interest because of the changing face of industrial maintenance. A combination of staff cuts, shortage of talent and changing technology is forcing industry to re-assess how it protects its plants going forward. Maintenance is not what it was 30 years ago when a company knew everything on site. Many maintenance departments do not have as many employees as previously.
Technology is changing rapidly. Companies really do need to be concerned. We are now offering a new vision. We are offering a way to capture and handle all the data relating to some motor-driven applications. So, having the reassurance that critical plant is being monitored offers peace of mind.
Is predictive maintenance suitable for all VSD applications?
This service is not meant for smaller drives as it is easier to simply replace these. With predictive maintenance we are focusing on large power drives in heavy industries where the impact of a malfunction on a critical application or process can be damaging and hugely costly.
Large powered drives are a major part of the production. They are a hugely valuable asset. Any break-down can have a major impact on production, safety and the environment. Any kind of repair or spares can be difficult to provide, especially as many of these target industries can be in remote locations.
Where is the service being piloted?
ABB is interested in locating drives controlling large motors in demanding processes. So far there are six pilot installations worldwide, with each chosen because of their technical, harsh and demanding appli-cations. The chosen applications include a mine hoist and metals industry processes. We are particularly focused on applications with harsh starting and stopping sequences that produce an ageing effect on the drive.
What are the initial findings from the pilots?
We meet on a regular basis to discuss the performance of drives, its lifecycle management and maintenance planning. This is all based on a high quality report with the initial analysis on usage and wear of equipment and confirmation that the drive is used according to the specifications.
One of the early findings is that the risk of accelerated ageing is eliminated for applications where the motor and drive are ideally matched and not over-dimensioned.
In another case, the ABB engineers were able to identify, remotely, that a particular set of drives were unlikely to survive more than three years due to their current operating profile. So, by recommending changes to the operating principles, ABB was able to extend the life expectancy by nearly 20 years.
What are the concerns going forward?
Inevitably, handling of data, cloud computing, AI and machine learning raise concerns about cyber security and data ownership. To address these issues, ABB has created a data manifesto which clearly states where the ownership of the data resides among other protocols. We have also forged strategic partnerships with leaders in the field of hosting, security and analytics such as IBM and Microsoft. These partnerships, combined with the in-house expertise of ABB Ability, provides our clients with the necessary security assurances.
What long term benefits are expected?
For the first time we are getting better knowledge, avoiding downtime and understanding the status of the drive by looking inside it in real-time. The customer now has a real-time view on the future and can predict the level of certainty around what could happen. There are evidence-based scenarios replacing personal judgements and experience. The impact is immense. We can now resolve a problem in hours rather than the weeks it currently takes. When you have remote locations, just getting an engineer to site can take a long time.
Customer response is tremendously supportive. After all, the service is based on the fact that ABB is here to help its customers by enhancing the reliability of the application and preventing any future mal-functions. They are very intrigued and interested to learn even more about how the process works.
In summary we have:
• Increased reliability: continuous monitoring and analysis of drive data supports maintenance planning to prevent unplanned shutdowns.
• Optimised maintenance: ABB’s predictive expertise raises the profitability of capital investments.
• Local support with global knowledge: contract-based service with active maintenance consultation.
ABB (ABBN: SIX Swiss Ex) is a pioneering technology leader in electrification products, robotics and mo-tion, industrial automation and power grids, serving customers in utilities, industry and transport & infrastructure globally. Continuing a history of innovation spanning more than 130 years, ABB today is writing the future of industrial digitalization with two clear value propositions: bringing electricity from any power plant to any plug and automating industries from natural resources to finished products. As title partner of Formula E, the fully electric international FIA motorsport class, ABB is pushing the boundaries of e-mobility to contribute to a sustainable future. ABB operates in more than 100 countries with about 135,000 employees. www.abb.com