Possible factors causing wrong amplitude estimation in double integration

Hi,

I am a university student and I am currently working on the calculation/estimation for displacement from IMU's acceleration signal. I have done some studies on this topic and I got a fairly reasonable waveform but the amplitudes for the position is wrong.

My experiment setup :

IMU on a 6DOF robot arm to carry the IMU up and down (single vertical axis) at about 0.37Hz continuously for 1 minute.

• Discrete Acceleration signal is recorded in computer with 125Hz sampling frequency

• The max displacement is measured manually (100 centimeters)

• Post-processing involve 2048 data.

• Noisy Raw signal is filter with Kalman Filter and moving average filter.

• Convert unit to centimeter/Second_Square

• Next, Using FFT Filter to remove DC component in the signal

• First graph is plotted. (Please refer to the attached image)

• Integrate using trapezoidal rule to get Velocity signal

• Using FFT filter to remove DC component in the Velocity signal.

• Second graph is plotted.

• Integrate using trapezoidal rule to get displacement signal

• Using FFT filter to remove DC component in the Displacement signal.

• Plot the third graph.

Raw Acceleration and filtered raw signal is shown here


Result for Acceleration, Velocity and Displacement signals


The result clearly showed the displacement and velocity signal amplitude are having big error W.R.T the reference displacement of 100 CM and the time recorded. I am less experience in signal processing and physics. Therefore, I am hoping someone able to point out some possible factors causing this type of error so that I can debug in that direction. Thanks in advanced