Detecting Faults Long Before the Damage
A key term of Industry 4.0 is ‘predictive maintenance’ – the forecast of damaging events long before they occur. The software company Indalyz Monitoring & Prognostics (IM&P) GmbH, which is based in the city of Halle (Saale) in the German Federal State of Saxony-Anhalt, has developed an innovative predictive maintenance software solution. This software is based on artificially intelligent algorithms that forecast when an individual component of a machine, complex plant or a machine cluster will reach its critical level or even break down. Future malfunctions are predicted long before the damage event occurs. The operator of a facility such as a wind park can thus organise service, material and personnel efficiently, which in turn reduces the operating costs and downtimes.
By Prof. Dr. Michael Schulz (IM&P) and Tanja Ruedinger (Investment and Marketing Corporation Saxony-Anhalt), Germany
Modern machines have a highly complex technical structure which makes them very costly and maintenance-intensive. It therefore makes economic sense to use each component up to the loss of functionality but at the same time to minimise maintenance costs and reduce downtimes to a minimum. To achieve all this, a pioneering maintenance concept is needed: predictive maintenance.
Monitoring Wind Turbines
Besides being able to monitor the current machine state, predictive maintenance software solutions are able to determine the expected loss of functionality of individual components or the entire plant over a longer period of time.
Wind turbines are a typical area of application for predictive maintenance. These machines are often difficult to reach and are therefore usually monitored preventively with a suitable condition monitoring system (CMS). Once the functionality of components has reached a critical state, the CMS gives a recommendation to shut down the affected wind turbine or change into another operating regime. The economic advantage of a CMS compared to traditional interval maintenance has meant that in recent years most turbines have used a CMS. This also applies to predictive maintenance, which is considered by many experts not only as a future technology but also as the next generation of maintenance concepts.
Predictive Maintenance Software Solution by IM&P
IM&P’s predictive maintenance software product consists of two parts: the multifunctional prognosis core software that handles the actual forecast calculation and the individual, peripheral software, which is tailored to the technical system.
The organisation of sensor-based input data and the calculated prediction results is undertaken by an integrated database. IM&P’s customers can access all relevant information on this database through user-oriented graphical interfaces.
Detecting Faults Long Before the Damage
IM&P’s prognosis software uses data from an existing CMS and is based on a self-learning, artificially intelligent core algorithm. Powerful computer technology processes historical and current sensor data together with engineering know-how, technical and manufacturer-specific parameters and other relevant information such as on-site weather data to create the prognosis. During this process, the prognostic system trains itself. It is this that provides the economic and technological difference between predictive maintenance systems and standard CMS. Prediction did not previously exist at this level of complexity. Prognostic tools developed by other companies are only based on the analysis of statistics and findings gained from many years of experience. Using artificial intelligence for the maintenance management of machines is new.
Revising the Common Maintenance Strategy
A wide range of wind energy plants are already constantly monitored by sensor systems. Sensor data such as structure-borne sound produced by the running wind turbine is fed into a CMS. These sound waves are characteristic of each individual wind turbine and reveal in the technical condition of their single components. If a specified tolerance has been exceeded, the wind turbine must be serviced or, in the worst case, switched off, regardless of whether or not downtime due to unscheduled maintenance is particularly unfavourable (e.g. during wind-rich days in spring or autumn). Thus the common maintenance strategy, which is based on monitoring of the current state of machinery and plant, has disadvantages.
Enabling a Look into the Future
IM&P’s software predicts the future wear performance of technical equipment. At the moment we can predict ahead for between six months and a year whether a component of a wind park system will need to be replaced. IM&P was founded by the physicist Professor Michael Schulz, who for many years researched the subject of artificial intelligence at the University of Ulm and at the Chemnitz University of Technology before founding the company at the Weinberg Campus technology park in the city of Halle (Saale).
Test Results
In a three-year trial test the prognosis software predicted the technical condition of 800 wind turbines, analysing historical and current sensor data of around 2,000 technical parameters and processing around 1.6 million time-series. The forecast horizon was between three and six months, but could have been extended up to one year given the availability of sufficient sensor data. The forecast quality has confirmed our expectations. Given a forecast horizon of six months, around 95% of the critical machine conditions (machine failure) were predicted by our software. Only 0.5% of the individual component malfunctions were not forecast by our software system, mainly due to spontaneous breakages. For the remaining 4.5% the system was over-cautious and recommended maintenance, but no maintenance was needed at that particular time.
Suitable for a Multitude of Applications
IM&P’s prognostic software can also be used in many other fields, if individually adapted. These fields include power stations, railed vehicles, engines and special-purpose vehicles. In addition to this, there are already outline proposals for the use of the prognostic software from Saxony-Anhalt in small and medium-sized hydroelectric power plants and steel factories, and for pumping and piping systems of the natural gas and crude oil industry.
The options in the federal state in the centre of Germany are ideal. A 150-year tradition of mechanical and plant engineering, boosting the area to a centre of high-tech production today, combined with an innovative and dynamic ICT industry, make Saxony-Anhalt a high-performance region for Industry 4.0 solutions and, thus, offer a variety of cooperation opportunities.
Conclusion
The productivity of each industrial plant is largely determined by the availability of that plant. So far, this is guaranteed mainly by monitoring the current machine status (condition monitoring), which results in an emergency shutdown or change in the operating regime once that industrial plant reaches a critical state.
With the help of intelligent sensor systems and advanced networking technologies, our forecasting software can predict the future wear of technical equipment. This helps operators to organise material and personnel efficiently long before the equipment enters a critical status.
