In this study, artificial neural networks (ANNs) technique was used in the prediction of abrasive wear rate of high-Cr white cast irons (HCCIs) after subcritical heat treatment at different temperatures. High Cr WCI alloys with different compositions were tested at sliding speed of 1.04 m s⁻¹ under the normal load of 30 N and different sliding distances of 500, 1000 and 1500 m. The abrasive wear rates obtained from wear tests were used in the formation of the data sets of the ANN. A multilayer perceptron model has been constructed with back-propagation algorithm using the input parameters of load, tempering temperature, and Cr/C ratio. The output parameter of the model is abrasive wear rate. Experimental results showed that abrasive wear rate of high-Cr WCI was significantly increased with the increasing of Cr/C ratio. High-Cr WCI alloys with higher volume fraction of carbides and structures with martensitic matrix at lower Cr/C ratio exhibited lower abrasive weight losses. The increasing of both sliding distance and load increases the abrasive weight losses. The HCCI-2 alloy exhibited the lower abrasive weight losses as compared with the other alloys in both as-cast and heat-treated conditions. In addition, the abrasive weight losses for all investigated alloys with different Cr/C ratios after destabilization heat treatment are lower than alloys in the as-cast state. This is may be due to the presences of stronger martensitic matrix structure rather than austenitic or pearlitic matrix structures. Correlation coefficients between the experimental data and outputs from the ANN established the feasibility of ANNs to effectively model and predict the wear rate of high Cr WCI. From the sensitivity analysis, it is concluded that the tempering temperature had the most influence on the wear rate, while the applied load and Cr/C ratio had a small influence on the wear rate of high-Cr WCI.

在这项研究中，使用人工神经网络（ANN）技术预测在不同温度下亚临界热处理后高铬白口铸铁（HCCI）的磨料磨损率。在1.04 m s ^-1的滑动速度下测试了不同成分的高Cr WCI合金在30 N的正常载荷下以及500、1000和1500 m的不同滑动距离下。从磨损测试获得的磨料磨损率被用于ANN数据集的形成。使用负载，回火温度和Cr / C比的输入参数，通过反向传播算法构建了多层感知器模型。该模型的输出参数是磨料磨损率。实验结果表明，随着Cr / C比的增加，高Cr WCI的磨料磨损率显着增加。碳化物的体积分数较高的高Cr WCI合金和较低的Cr / C比的具有马氏体基体的组织表现出较低的磨料失重。滑动距离和载荷的增加都会增加磨料的重量损失。在铸造和热处理条件下，HCCI-2合金的磨料重量损失均低于其他合金。另外，所有研究的去稳定热处理后具有不同Cr / C比的合金的磨料重量损失均低于铸态状态的合金。这可能是由于存在较强的马氏体基体结构，而不是奥氏体或珠光体基体结构。实验数据与人工神经网络输出之间的相关系数确定了人工神经网络对高Cr WCI磨损率进行有效建模和预测的可行性。从敏感性分析可以得出结论，回火温度对磨损率的影响最大，而施加的载荷和Cr / C比对高Cr WCI的磨损率的影响较小。在失稳热处理后，所有研究的具有不同Cr / C比的合金的磨料重量损失均低于铸态状态的合金。这可能是由于存在较强的马氏体基体结构，而不是奥氏体或珠光体基体结构。实验数据与人工神经网络输出之间的相关系数确定了人工神经网络对高Cr WCI磨损率进行有效建模和预测的可行性。从敏感性分析可以得出结论，回火温度对磨损率的影响最大，而施加的载荷和Cr / C比对高Cr WCI的磨损率的影响较小。在失稳热处理后，所有研究的具有不同Cr / C比的合金的磨料重量损失均低于铸态状态的合金。这可能是由于存在较强的马氏体基体结构，而不是奥氏体或珠光体基体结构。实验数据与人工神经网络输出之间的相关系数确定了人工神经网络对高Cr WCI磨损率进行有效建模和预测的可行性。从敏感性分析可以得出结论，回火温度对磨损率的影响最大，而施加的载荷和Cr / C比对高Cr WCI的磨损率的影响较小。这可能是由于存在较强的马氏体基体结构，而不是奥氏体或珠光体基体结构。实验数据与人工神经网络输出之间的相关系数确定了人工神经网络对高Cr WCI磨损率进行有效建模和预测的可行性。从敏感性分析可以得出结论，回火温度对磨损率的影响最大，而施加的载荷和Cr / C比对高Cr WCI的磨损率的影响较小。这可能是由于存在较强的马氏体基体结构，而不是奥氏体或珠光体基体结构。实验数据和人工神经网络输出之间的相关系数确定了人工神经网络对高Cr WCI磨损率进行有效建模和预测的可行性。从敏感性分析可以得出结论，回火温度对磨损率的影响最大，而施加的载荷和Cr / C比对高Cr WCI的磨损率的影响较小。