Bayesian networks model for prediction of agricultural soil penetration resistance in interaction with different parameters

Inadequate and frequent use of tractors and agricultural machinery generally results to soil compaction. This compaction can cause problems such as reduced crop growth, loss of nutrients or inadequate water infiltration. Therefore, it is necessary to examine the parameters influencing soil compaction and the complex interactions between them. In this study, a Bayesian network approach was utilized to develop a decision support framework tailored for evaluating the risk of soil compaction. Raw measurement data were employed to construct a discrete Bayesian network, aiming to unveil connections among diverse factors defining soils and testing conditions. These factors encompass moisture content, tractor weight and number of passes. Bayesian network (BN) models are trained on resistance penetration (Rp) test data obtained from measured experimental values. The BN models’ predictions are compatible with the measured data, as indicated by the accuracy of the model which is 0.962. Investigating compaction phenomena entails analyzing how soil resistance responds to the mentioned factors. The generated simulated Bayesian network functions as a valuable tool for decision support, assisting users in the causal and diagnostic analysis of soil compaction under varying moisture conditions.