Thank you for attending our Webinar on Effective Defect Analysis: Maximizing your Vibration Setup by Mitch Stansloski, PhD, PE – Pioneer Engineering
If you missed our Webinar, you can view it on demand at any time.
We hope that you found your time with us to be informative and interesting. Here are the answers to your questions:
Q: Does the Nyquist sampling theorem come into play with your analysis?
A: The Nyquist Sampling Theorem states that there must be at least two samples per cycle in order to prevent aliasing. If less than two samples per cycle are collected, false low frequencies develop. In practice, the sampling frequency is set to 2.56 times the maximum frequency desired or higher for some devices. Some devices may sample at a higher rate to provide more detail in the data being acquired. This additional data can increase the accuracy of the measurements and calculations as a result.
Q: Does vibration help on hydraulic pump problems?
A: Yep! For the sliding vane type or gear type, the pumping frequency is the number of vanes or gear teeth multiplied by the running speed.
Q: For collecting data on shovels, that run 24/7, would you still follow the interval settings that you presented or is there a different method to use?
A: Since these are so critical and have such massive loads, I would set the intervals closer than recommended until experience dictated otherwise.
Q: For resolution, question is 2 hz to 1.23 hz acceptable for how the ludeca signal processing works? In regards to envelope bearing defection, do you recommend high pass to low pass filters to be no more of a difference of 2 to 4 khz?
A: Not sure I understand the first part of this question. For envelope bearing detection, I would set one to find stage one bearing defects. This would put the filters at 5kHz to 40 or 50kHz. I would set another to find stage two bearing defects (bearing resonances). This would make the filters at 500 Hz to 5kHz.
Q: For resolution, in regards of delta F value to stay between 2 hz to 1.23hz, this value is f max divided by # lines?
A: Resolution is the frequency range divided by the number of lines.
Q: Looking at a reduction gearcase, wouldn’t the slower output with fewer cpm’s need more averages in order to get a better sampling rate verse the high speed input which is turning faster and would require fewer averages?
A: The number of averages won’t affect the sampling rate. In general, we recommend more averages for higher frequency ranges. But, you may still want a few more averages for the output of a gearbox since there are many transients that are low frequency. A faster turning shaft wouldn’t require fewer averages.
Q: How would you develop a collection interval with varying loads?
A: Choose the worst case load, as a function of load and duty cycle.
Q: How did you calculate the belt frequency please?
A: Belt Freq = Input Speed x pi x Sheave Diameter / Belt Length
Q: How do I determine monitoring cycle based on load?
A: The bearing L10 life equation does show that the life is inversely proportional to the load cubed. So, as load exceeds recommend, collection intervals should dramatically shorten.
Q: How do you ensure adequate resolution on a time waveform?
A: Inadequate resolution on a time waveform will cause erroneous low amplitudes. Keep improving the resolution until the amplitudes are maximum and repeatable.
Q: How do you find the load on a particular machine…Is there a formula?
A: Convert the current draw to horsepower is likely the easiest way.
Q: How do you set up the proper resolution? How parameters you use to decide what resolution to use?
A: If using a Hanning window, the resolution should be set so that it is 1/3 of the smallest peak separation needed to correctly distinguish between defect frequencies.
Q: How many averages do you recommend for vibration data collection?
A: A rule of thumb is to start at 4 averages with a frequency range of 120,000 CPM and then go up or down from there.
Q: What is a good Fmin setting?
A: If collecting a velocity spectrum using an accelerometer for collection, the Fmin should eliminate the integration error found in the first frequency bin. So, the Fmin should filter at least the first bin.
Q: What’s the best way to deal with variable speed?
A: Orders based will work if the speeds are variable but don’t change constantly. In addition, multi-level alarms based on running speed could be used. For machines whose speed varies constantly, order tracking is required in order to obtain a clear spectrum.
Q: Will it not be messy to have a number of frequency range for different equipments is the same route of measurement?
A: The data collector doesn’t care how you’ve set up your collection specification. And, neither does your trending software, Better to have a more detailed database and valuable data than one that is simple but ineffective.
Vibration Analysis by Mitch Stansloski PhD PE