Main content area

Modeling and Simulation of a Four-Rotor Rake Loading for Predicting Accumulated Fatigue Damage: A Markov Regime-Switching Approach

Paraforos, Dimitris S., Griepentrog, Hans W., Vougioukas, Stavros G.
Applied engineering in agriculture 2018 v.34 no.2 pp. 317-325
Markov chain, Monte Carlo method, agricultural machinery and equipment, data collection, grasses, markets, models, prediction
The prediction of agricultural machinery‘s fatigue life is of increasing importance for machine developers who must produce durable and reliable machines for a globalized market with different local operating conditions. Mathematical tools that can model and simulate the variable loading of agricultural machines are necessary for fatigue life prediction. Modeling should be based on measured loads from real-world operations. In this article, the loads of a four-rotor rake were recorded during grass swathing. Markov chains were used to model the transitions between the machine‘s operating conditions (in-field swathing and headland turning) and the sequences of turning points present in the load signals. The Markov transition probabilities were trained using the recoded data and then fatigue life was predicted via executing 10,000 Monte Carlo simulations based on the trained Markov models. The differences between the accumulated fatigue damage predicted from the simulations and from the measured data had mean value and standard deviation equal to -22% and 12.8%, respectively. Evaluation of the trained model on new data (not present in the training dataset) that were recorded during swathing on a different grass field resulted in fatigue damage difference with mean and standard deviation equal to 34% and 7.5%, respectively. The fatigue damage difference was in a reasonable region considering how fatigue life is affected by high-amplitude individual cycles.