Abstract: Unlike passenger vehicles, mobile farming units are constantly subjected to shock loads while working in the field. The frame of a corn harvester supports all the essential components including the engine, cab, tanker, and header, etc. Because fatigue performance is critical to the reliability of the entire harvester, it is therefore important to apply an appropriate method to evaluate the corresponding fatigue life. A novel fatigue theory named Power Density was explored in this paper to estimate the fatigue life of the frame with significantly better precision. Power density is a quantitative measure of the stress gradient in time, thus having a physical unit of power per cubic meter volume as in W/m³. The power density of a material can be established using S-N curve. Short-time Fourier transform(STFT) is employed to resolve the frequency content of the input signal which is inherently random in time. The dominant frequency components of the induced stress variation in time along with the corresponding transformation coefficients at a certain time can be extracted in the STFT time-frequency domain. Hamming window was chosen and properly configured to extract the salient features unique to the stress signal acquired. Assuming that the induced stress variation was a time harmonic function, the power density of each frequency can thus be expressed by stress amplitude, frequency and time according to Fourier theory. Using the assumption, the damage rate at a certain time can be determined. The accumulated damage within a time period was then calculated by summing up the damage collected at each time interval. Consequently the fatigue life can be estimated according to the accumulated damage and the time period. An experimental plan was also developed to obtain the stress profiles of the critical points in the harvester frame under the real-world working condition. The critical test points were determined by physical investigation and finite element analysis. Strain gauges were mounted to each test point and subsequently hooked up to a Wheatstone bridge for outputting strain readings in real-time. The output voltage was amplified by a signal amplifier module and acquired by a computer program running on an NI DAQ card. The response rate of the strain gauge was 50 000 Hz. Sampling frequency was then set at 5 000 Hz to ensure no aliasing of the test data would occur. The dynamic stress-time curves of all the test points were stored as voltage signals. Finally, using the test data and the power density theory, the fatigue life of the corn harvester frame was estimated at 394 h, as opposed to 845 h that was obtained by applying the Miner rule. The fatigue life estimate given by the new theory was in close agreement with the real-world operation life of the corn harvester frame which was between 400 and 500 h. In this new approach, we considered both the amplitude and oscillation frequency of the stress variation simultaneously, the fatigue estimate was physically more credible, particularly so when the working conditions involved loading inputs that were nonstationary, broadband, and random. In addition to providing a feasible alternative to all current fatigue theories, the method developed in the paper demonstrated to be applicable to the fatigue analysis of farming equipment.