Predictive Maintenance is the anticipatory servicing of machinery and plants and German-based company Indalyz Monitoring & Prognostics (IM&P) has developed a piece of prognostic software that uses artificially intelligent algorithms to predict when a component on a machine will break down and have to be replaced.
Wind energy plants are already predominantly constantly monitored by sensors. Such sensors measure, for example, the machine noise produced by the running plant. These sound waves are characteristic of each individual plant and reveal what technical condition their single parts are in. If a specified tolerance has been exceeded, the plant must be serviced or in the worst case switched off. It must be switched off entirely, regardless of whether the point in time is unfavourable because the stoppage takes place during the most wind rich days in spring or autumn, for example. The common maintenance strategy, which is based on monitoring of the current state of machinery and plant, thus has disadvantages. The solution is so-called ‘predictive maintenance’. This is now a key term of Industry 4.0.
IMP develops predictive software-based maintenance strategies for individual machines, complex plants and machine clusters. Malfunctions are thus detected while they are arising by artificially intelligent algorithms long before the critical machine state is reached. The user can thus organise service, material and personnel optimally, which in turn reduces the operating costs.
“Our software can predict the wearing of technical systems. At the moment we can predict for the period of half a year whether a component of a wind park system has to be replaced, for example”, commented Professor Michael Schulz of IM&P. For many years, the physicist has researched the subject of artificial intelligence (AI) and developed his Predictive Maintenance Software Solution at the University of Ulm and the university Technische Universität (TU) Chemnitz. The IM&P programme uses artificial intelligence for the maintenance prognoses of machines for the first time. Previously, this was only usual in space travel or military technology. “Our prognosis software is based on self-learning, artificially intelligent core algorithms, which are themselves combined with different intelligent filters and projective methods. Past and current sensor data are compared with engineering, technical and manufacturer specific parameters and supplemented with other relevant information such as the climatic conditions at the site. “During this, the prognostic system trains itself. This did not previously exist in this level of complexity”, commented Professor Michael Schulz. Other prognostic procedures are based above all on the analysis of statistics and findings gained from many years of experience.
The prognostic software can be used in many other fields, if individually adapted. These include power stations, rail vehicles such as goods wagons, engines and special purpose vehicles. In addition to this, there are already outline proposals for the use of the prognostic software in small and medium sized hydroelectric power plants, and for pumping and piping systems of the natural gas and crude oil industry - such as in the monitoring of leaks as they arise.