Abstract An agricultural backhoe is an important machine designed for multiple assignments in agriculture and livestock. Due to severe working conditions, agricultural backhoe elements are subjected to high loads. Therefore, a structural design must provide a safe machine under all loading conditions at minimum weight and cost. In this work, a 3D model of an agricultural backhoe was proposed, to be used in tractors category II according to the classification of the ASABE S217 standard. In the structural design of the agricultural backhoe Theoretical Analysis (TA), Finite Element Analysis (FEA), Response Surface Methodology (RSM) and Artificial Neural Network (ANN) were used. A finite element model of the agricultural backhoe in the critical position was developed, considering a maximum breakout force according to SAE J1179 standard. The finite element model was theoretically validated through a comparison between numerical and theoretical normal stresses at twelve strategic points of the agricultural backhoe components, finding a maximum absolute difference of 7.0 %. Also, a mass reduction of the principal backhoe components (bucket, arm, boom and links) was done, using Central Composite Design (CCD) under RSM in a commercial FE software and ANN technique with Neural Lab software. A mass reduction of the initial agricultural backhoe model of 24.8 % (from 446.3 kg to 335.4 kg) was achieved using the RSM technique, with an increment on the maximum von Mises stress of 6.8 % (from 117.4 MPa to 125.4 MPa), as well as a reduction of the minimum safety factor of 4.8 % (from 2.94 to 2.80 ). ANN allowed predicting the results obtained by RSM to reduce the boom mass with a correlation coefficient of 0.96 , using 80.0 % of data and around 13.0 % less time. This study showed that a combination of RSM and ANN techniques with TA and FEA provides useful results to reduce the structural mass of agricultural equipments, thus it is recommended to decrease the number of numerical case studies and the solution time with satisfactory results.

中文翻译：

---

使用 TA、FEA、RSM 和 ANN 的农业反铲结构设计

摘要 农用反铲是为农牧业多种作业而设计的重要机械。由于恶劣的工作条件，农用反铲元件承受高负载。因此，结构设计必须以最小的重量和成本在所有负载条件下提供安全的机器。在这项工作中，根据 ASABE S217 标准的分类，提出了一种农用反铲的 3D 模型，用于 II 类拖拉机。在农业反铲理论分析（TA）的结构设计中，使用了有限元分析（FEA）、响应面方法（RSM）和人工神经网络（ANN）。根据 SAE J1179 标准，考虑到最大挖掘力，开发了关键位置农用反铲的有限元模型。通过在农业反铲部件的十二个关键点处的数值和理论法向应力之间的比较，从理论上验证了有限元模型，发现最大绝对差异为 7.0%。此外，还使用商业有限元软件中的 RSM 中央复合设计 (CCD) 和神经实验室软件的 ANN 技术，完成了主要反铲部件（铲斗、臂、动臂和连杆）的质量减少。使用 RSM 技术使初始农用反铲模型的质量减少了 24.8%（从 446.3 kg 到 335.4 kg），最大 von Mises 应力增加了 6.8%（从 117.4 MPa 到 125.4 MPa），以及将最小安全系数降低 4.8%（从 2.94 到 2.80）。人工神经网络允许预测 RSM 获得的结果，以减少动臂质量，相关系数为 0.96，使用 80.0% 的数据和大约 13.0% 的时间。本研究表明，RSM 和 ANN 技术与 TA 和 FEA 的结合提供了有用的结果来减少农业设备的结构质量，因此建议减少数值案例研究的数量和求解时间，并获得令人满意的结果。