Perhaps you have already asked yourself why a new, idle wind turbine has been standing still for weeks despite the wind shortly after completion and commissioning.
One of the main reasons for this is the so-called tonality, the oscillation of system segments at a specific frequency that is very perceptible to the ear, which is why systems cannot receive an operating license in accordance with the IEC 61400 standard. The subsequent remediation of the causes of a decommissioned system is always associated with very cost-intensive immediate measures. The EU-funded Eureka-Alarm research project by turbine manufacturer Senvion, Novicos and other partners aims to effectively prevent such tonalities during turbine development.
If a system that has just been put into operation comes to a standstill due to tonality problems, the operator and developer must immediately solve the following questions: "Which subsystem of the system is causing the tonality and what costs and time are involved in eliminating it?"
The sound emission of wind turbines is characterized by various physical processes, such as sound excitation, sound transmission, sound radiation and sound decoupling. These processes take place at different levels of concretization, which poses a real challenge for the recording and analysis of system properties using existing measurement methods.
In the age of Industry 4.0, there was therefore a desire to exploit the possibilities of the "digital twin" to reduce costs by eliminating these tonal effects. Figure 1 shows typical sound transmission paths.
The digital twin as a solution
The digital twin is a virtual image in the computer that is identical to the real wind turbine down to the last design feature. This exact clone of a turbine and the simulation of the real behavior, primarily with regard to vibration behavior and acoustic radiation, was the project objective in order to derive corrective measures.
In order to breathe life into a system, even if it is a digital one, a description of what "the life" of a system actually is is required. "It's like a computer game in which each person is assigned a character trait and then interacts with others according to their criteria. The more descriptions and variance the characters have in the game, the more realistic their movements are in the game, the more exciting and successful the game will be on the market," explains Dr. Olgierd Zaleski, Managing Director of Novicos.
It is no different for engineers who develop such simulations. The only difference is that instead of character descriptions of people, they have access to data material from various measurements and material characteristics of the system components, some of which are newly developed. For this purpose, each important design detail whose behavior interacts with other mechanical system components is described on the basis of a design in the computer (CAD data).
In simulation language, this process is referred to as modeling. The interaction, i.e. the exchange of forces as excitation between mechanical systems, is determined in so-called multi-body simulations. Even before the start of the project, it was assumed that existing modeling methods could numerically predict sound propagation in complex systems and thus support the development process in early phases by providing important sound emission data. Unfortunately, however, at the time there was neither a sufficient database nor a suitable overall model for wind turbines. However, the consideration of the overall acoustic model behavior is one of the most important statements of the twin.
From the company's point of view, prompt results of design variants are the added value of the digital twin.
Calculation results that take months to produce are therefore only accepted in exceptional cases in today's world of ever-shorter development cycles. With the currently available algorithms and typically available hardware and software tools, simulation calculation times of several weeks would probably have been required. It was therefore the intention of the research project to develop subsystem models and improved calculation algorithms in order to greatly accelerate the calculations. In order to achieve the core property of the twin with fast delivery of results, so-called transfer functions of coupled main components of the system were used, as shown in Figure 2 and Figure 3.
These were validated with measurements of transfer functions on a real system. In simplified terms, transfer functions describe how the vibrations generated by a system, for example the gearbox, are transmitted to the tower. With these transfer functions and clustering, the division of the system into different components, the simulation could already be accelerated considerably. What was still missing was a quick method of calculating how the turbine as a whole radiated acoustically from the components. Novicos developed an acoustic model to simulate the sound radiation of a wind turbine. The boundary element method was used for this. The entire surface mesh of the rotor, the nacelle and the tower has approx. 4 million elements. A section of the BEM meshing is shown in Figure 4. The very fine meshing shown results from the frequency range targeted in the project.
Only with the new calculation methods developed by Novicos in this project can models the size of a wind turbine currently be calculated with sufficient physical accuracy.
Dr.-Ing. Marian Markiewicz, Managing Director of Novicos: "As part of the project, stable block approximation algorithms for hierarchical H-matrices were developed and implemented so that system matrices can be generated for any network with a stable approximation error. From the user's point of view, this means in simple terms that the computing memory and computing time could be minimized by a factor of 1000 thanks to these new mathematical descriptions."
Effects of the floor with regard to the sound radiation of the system could also be taken into account using so-called "impedance boundary conditions". This led to a further reduction in the computational effort and modeling processes.
The result and further prospects
With such a simulation process in the digital twin, which has been validated in practice, system manufacturers now have a way of determining the sound radiation of system components at an early stage and determining the effects of the individual components on the overall sound radiation. In the project, for example, the non-linear vibration behavior of the elastomer bearings was determined in a specific frequency range. With this knowledge, it was possible to introduce specific measures to counteract the tonality. Requirements for suppliers and vendor parts can thus be validated at an early stage. With cost reduction in mind, optimization measures on the manufacturer and supplier side will become more efficient in their application in the future. The calculation results (exemplified in Figure 5) in a period of a few days now allow this.
The consideration of quality influences on the vibration properties in the manufacturing process also allows for such close integration of development with the subsequent production process. It is another approach for saving costs at an early stage and investing in optimization measures in a more targeted manner.
The new simulation method for the radiation behaviour of large structures developed as part of the project is now being used at Novicos for the development of other digital twins, such as ships. The methods are to be further developed in the near future so that moving sources, such as aircraft, can also be calculated using them.
Do you have any questions about this article?
I will be happy to answer any questions you may have.
