I'm working on a project where I'm measuring the acceleration of a piece of equipment in our plant using 3 single axis accelerometers. I have 2 of them measuring the X axis (to detect when it is swaying from side to side) and one measuring the Z axis (to show when the equipment is moving up and down). I'm trying to use DIAdem to analyze the data, and I've run into a problem when I integrate the raw data to get posisition. Here's a screenshot of what I'm experiencing:
The top channel is the acceleration from one of the X axis sensors, the middle is the 1st integration, and the bottom is the 2nd. I would expect the veloctiy curve to look similar to the acceleration curve, but out of phase by 90 degrees. The position curve should also have a similar shape. What am I missing here? I'm new to using DIAdem to analyze data, so any help is appreciated. Thanks!
How are you running the calculations to determine your second and third graphs? Could you post the equations you are using to produce these?
I'm actually just using the Integrate function from the Basic Mathematics menu in the Analysis tab. Then I do a second integration on the results channel of the first. I think I may be missing an offset correction or something...
What are you using as the X channel in the integration dialog?
DIAdem Product Support Engineer
That could be the issue... I'm just using the blank time channel from the accelerometers. Should I generate a time channel in DIAdem?
You could generate a time channel in DIAdem or log the time alongside your acquisition. How are you logging the data?
I'm using a NI 9184 compact DAQ chassis with a NI 9234 signal conditioner with analog accelerometers for sensors. I'm using Signal Express to generate a .tdms log file. I've just been using the internal clock for my X-axis when graphing the data. Do I need to create a time channel or use an external clock?
So to be clear you are using the internal clock data as your Time axis?
Yes, that's what I've been using. I think I figured it out! I just used a high pass filter on the acceleration data, and then integrated that filtered signal. That resulted in the type of response curve I was expecting.
I think I'm still missing something... The velocity and position curves don't seem to match up with the actual machine motion. For example, the position curve indicates the sensor is moving up when the actual machine is moving along its Y axis, and does not indicate the correct magnitude of the Z axis change when the machine is moving up and down. Could this be due to integrating a noisy signal?