Ship builder Wärtsilä was getting complaints about the nighttime noise in a residential area of Vassa, Finland near one of the company's main large 4-stroke engine factories. The facility produces about 500 ship engines per year, and every engine is subjected to a test run. When they measured sound levels, they found that the likely source of the disturbance was exhaust noise produced by a 1,000-hour endurance test which runs continuously at 750 rpm. That's a lot of noise, and the noise had to be reduced at night while the test still ran nonstop.
To begin the process of finding the source of the loudest sounds -- and to lower the volume -- engineers took measurements at three locations between the test facility and the residential area. Sound pressure level measurements indicated a peak in the sound spectrum at the 100 Hz 1/3 octave band at approximately 94 Hz in narrowband, which corresponds to the crankshaft rotation frequency (CRF) order of 7.5. Typically, the exhaust noise of the engine should have the highest peaks at CRF orders 3 and 4.5. The measured noise was found to be well under the nighttime environmental noise limit. Thus, the annoyance levels were produced by the dominant half order modulated low frequency noise.
Additional measurements were taken to identify the exact cause of the 100 Hz 1/3 octave band noise taken near the engine exhaust system on the factory roof. High noise levels were noted near the first-stage silencer, which is a double expansion chamber reactive silencer. Measurements were taken at several points near the silencer surface. The noise peak in the 100 Hz 1/3 octave band dropped more than 10 dB when the engine load went from 100% to 75% while the total noise level dropped by nearly 8 db. Speed sweeps were performed under varying loads to determine the resonances of the entire exhaust system.
Using simulation to identify the problem
Company engineers decided to use Actran, a software program produced by MSC Software to help identify the problem. "Wärtsilä used Actran to address this problem because they wanted to evaluate the software's effectiveness in solving vibro-acoustic problems," Erin Komi, research scientist for VTT Technical Research Centre in Finland, the group that performed the simulation, told Design News in an email. Actran was used to create a vibro-acoustic model of the silencer, including both internal and external air volumes, as well as infinite elements for sound radiation computations. A zero admittance boundary which characterizes the surface as a rigid wall was applied on the laboratory roof. Field points were positioned in and around the silencer for comparison with physical measurements.
The engineers turned to simulation because it offered an efficient and intuitive solution, especially for non-experts of exhaust system acoustics. "The problem may also have been solved without simulation by someone who is very experienced in acoustics of exhaust systems, but extra testing -- which means a higher cost and more time consuming solution -- may have been necessary without simulation," Joe Zhou, senior application engineer for MSC Software, told us. "The value Actran brought was to simulate the realistic vibro-acoustic physics that was happening in the exhaust system, and to offer visual results -- color maps of acoustic pressure -- to help pinpoint the root of the noise problem. We were also able to virtually prototype the new design and predict the sound of the new design."
Simulation showed that the primary acoustic resonances predicted by the model matched theory extremely well but failed to explain the source of the 94 Hz noise. Engineers noticed, however, that a curved duct leading to the silencer inlet was of a length that could potentially have a first axial resonant frequency near 94 Hz. So they expanded the model to include the ductwork and quickly identified the root cause. "The problem was actually found outside the silencer. After the first Actran simulation, the noise problem was found at a small duct connected to the silencer," said Zhou. "Without the light shed by the first simulation results, engineers would have still been looking for problems in the silencer design itself. In that case, any design change in the silencer would not have fixed the noise issue."
The results were quiet
Extending the silencer inlet by 1.4m to the centerline of the first chamber solved the noise problem. Lengthening the horizontal ductwork leading to the silencer changed the resonant frequency so it no longer matched the engine order at 94 Hz. Also, the extended inlet was at the nodal line of the longitudinal and cross-channel modes of the first chamber, substantially reducing the effects of both modes.
The simulation predicted that this change would reduce the sound level at 94 Hz by 7 dB, which put it within the acceptable levels for the residential area. The physical silencer was modified to match the changes made to the Actran model and sound pressure level measurements were repeated, thus solving the problem. "Actran's application in exhaust systems is not unusual. There are other cases of mufflers and silencers," aid Zhou. "The uniqueness in this case is how Actran simulation helped the engineer identify the cause of the root problem quickly."
Tell us your experience in solving a knotty engineering problem. Send stories to Lauren Muskett for Sherlock Ohms.