It is becoming obvious that those who want to make better decisions and save maintenance costs in the future have to make use of data from their asset. But consolidating and interpreting such data is a challenging task and so far, companies have been using it for condition diagnosis and not for condition prognosis, writes Moritz von Plate

 

There is almost no asset that does not have maintenance costs. A great part of the running costs are due to unplanned malfunctions that, in the past, no company has been able to predict.

The asset management field is broad and not at all easy to oversee. It is worthwhile to look at the different options because maintenance is an area where companies can save considerable costs.

Through intelligent maintenance, companies hope to optimize the operational lifespan – that is, the remaining useful life – as well as quality, security and environmental standards. It optimizes the availability of the asset, thus avoiding high downtime costs.

Also, maintenance has great savings potential in terms of personnel, inventory, time management and error rates. In the past, the most well-known maintenance strategy was probably reactive maintenance: the engineer troubleshoots as soon as errors arise.

This first generation of maintenance does not deal with issues systematically or based on data. For components that are not business-critical and whose malfunction only causes minimal costs for the company, this method might still be suitable.

However, for components that are critical for the entire production and whose malfunction would incur significant costs, companies should rely on other measures. For critical assets, whose failure would have dangerous consequences, this kind of maintenance is not suitable at all. Increasing interlinkage of components and assets and increasing costs through downtime forces companies to rethink current solutions.

Industrial plants particularly have replaced their first maintenance generation with a second one, introducing preventative maintenance. They determine the mean time between malfunctions and use these as the basis for a regular maintenance cycle. This method, where data analysis only plays a subordinate role, is relatively easy to organize and achieves better results in keeping the asset available than the reactive method.

However, companies often conduct expensive maintenance too early, even though the asset does not yet require it. Also, preventative measures do not systematically prevent malfunctions and failures.

With the increasing digitalization of assets and processes, the third maintenance generation is starting to roll, giving greater importance to data collection. Most companies are trying to introduce condition-based maintenance. It rests upon so-called condition monitoring, in which data on a machine is collected. Ideally, condition monitoring analyses the data in real time to give warning of malfunctions or damages. The term ‘predictive’ is used when the tools give early signals about possible malfunctions. However, what is missing is a specific timescale about when a failure could arise.

With the aid of mathematical models, companies have the opportunity to use this data for a condition diagnosis and prognosis. Together, diagnosis and prognosis form the basis for condition-based maintenance, making it possible to conduct maintenance measures at a time when they are technically necessary and economically sensible. Companies implement condition-based maintenance with the use of data from sensors as well as remote maintenance modules or analytic systems, giving the opportunity for both online and offline surveillance.

Data is crucial

It is becoming obvious that those who want to make better decisions and save maintenance costs in the future have to make use of data from their asset. Through the new technologies of Industry 4.0 and big data it has become essential for every company to collect data about any kind of asset.

By analyzing this data, the system operators try to get answers to different questions such as why the asset is in a critical condition or which components can be expected to cause failures. Through this, system operators want to decrease downtime of their asset and stretch out the maintenance cycle. With the aid of condition and process data such as temperature, vibration as well as lubricant analysis, currently established approaches to data analysis provide significant information about critical components of an asset.

Consolidating and interpreting such data is usually a challenging task, since different condition monitoring/diagnostic technologies often look at various aspects and address different malfunctions. Also, they only describe the current condition of the components without making a clear prognosis. Time-related questions remain unanswered. Even methods such as predictive analytics only set off a warning which says that a failure is to occur at some point in the future. Whether this is to be expected in an hour or in two weeks or only in three months remains open.

This means that, so far, companies have been using the data at best for condition diagnosis, not sufficiently for condition prognosis, even though – as explained above – it is the second critical information on the way towards condition-based maintenance.

Prognostics as the next step

Early detection of risk indicators and time calculations of future malfunctions are made possible by the Cassantec Prognostic Report. It expands the planning horizon of companies by using specially developed algorithms. The report gives the user the opportunity to see beforehand when the condition of a machine reaches a critical state. It supplements other analysis tools that are part of the daily running of companies today. This decisive analytical component as a basis for condition-based maintenance draws on the same data that the company has already collected with the methods explained above.

Making remaining useful life transparent and, ideally, actively controlling it was the goal of utilizing Cassantec’s Prognostic Solution in a hydropower station of the Swiss utility BKW Energie.

In addition to the active management of remaining useful life, maintenance scheduling and preparation were also to be optimized in order to lower the operation and maintenance costs. The Swiss utility also hoped to change the operational strategy of the plant in such a way that the plant’s operation during its remaining useful life and the purchase planning for spare parts could be optimized.

Bannwil hydropwer plant utilized an automated prognostic solution

Currently in Switzerland, power generation is focused on hydroelectric and nuclear sources. BKW is one of Europe’s largest energy utilities with an annual power production of about 146.7 GWh. Some of BKW’s hydroelectric plants will be reaching the end of their originally planned life in the coming years.

To determine the remaining useful life of the power plants and their key components such as bulb turbines, generators and transformers, to control and to extend their life and to determine the optimal renewal time for these components, the hydroelectric plant Bannwil implemented Cassantec’s automated Prognostic Solution.

The remaining useful life of the entire system can be influenced by the operating strategy since the respective remaining useful life is highly dependent on the chosen workload of the individual components.

Load-dependent operating scenarios can be calculated to obtain different perspectives on remaining useful life. Moreover, the prognostic solution supports and complements condition monitoring and diagnostics during regular operation of the system and thus the optimization of maintenance planning, scope and costs.

Project execution

First of all, Cassantec collaborated with BKW to choose the most important components of the Bannwil plant. Then Cassantec generated a prognostic report for the respective components based on current and historical condition and process data. These included, in particular, data on temperature, lubricants, vibrations and rotation speed, which are recorded and stored during plant operation.

A computerized stochastic model analyzes, aggregates and correlates this data with malfunction modes or life limiting factors. The resulting remaining useful life distributions can be aggregated at plant level and updated – for example, daily, weekly, or monthly – in the desired time intervals. Cassantec presents the prognostic reports for the plant and all individual components in a secure online portal. BKW provides the current data needed for the update via a SharePoint Server.

Based on existing data and the complementary know-how of BKW and Cassantec, prognostic reports are developed for all mentioned components. It turned out that the Bannwil plant’s remaining useful life was largely determined or limited by the generator in the third power plant unit.

The reason for this is particularly a progressing defect that can be seen in the increased levels of the vibration parameters for this generator. Based on these parameters it can be seen how the generator’s remaining useful life is dependent on the respective load scenario.

With an operating strategy that limits the power of the generator to roughly 85 per cent of capacity and, if necessary, increases the power of the other two generators, constant bearing vibration gradients and correspondingly extended remaining useful life are to be expected. With the appropriate adjustment of the Bannwil plant’s operating strategy, BKW can extend the remaining useful life of the entire system.

Moritz von Plate is chief executive officer of Cassantec, an independent provider of prognostic solutions for industrial asset management founded in Zurich and operating from Berlin, Germany and Cleveland in the US. www.cassantec.com