| 关罗浩 1，战 磊 1，张 玺 1，饶宇宁 1，刘秋红 2，高 攀 2，刘建书 2，刘程炜 3.响应面法和 BP 神经网络在烟叶拉力测定中的应用[J].广东农业科学,2023,50(11):146-154 |
|
查看全文
HTML
响应面法和 BP 神经网络在烟叶拉力测定中的应用 |
| Application of Response Surface Method and BP Neural Network in the Determination of Tobacco Leaves Tensile Force |
| |
| DOI：10.16768/j.issn.1004-874X.2023.11.015 |
| 中文关键词: 烟叶拉力 质构仪 响应面法 参数优化 含水率 BP 神经网络 |
| 英文关键词: tobacco leaves tensile force texture analyzer response surface method parameter optimization moisture content BP neural network |
| 基金项目:广东中烟科技项目（粤烟工科任〔2021〕-008 号） |
|
| 摘要点击次数: 565 |
| 全文下载次数: 523 |
| 中文摘要: |
| 【目的】烟叶拉力是烟叶物理特性之一，反映烟叶的耐加工性，对烟叶拉力特性进行研究可为烟叶打叶复烤参数设置提供参考，进一步提高烟叶加工的经济效益。【方法】为提高质构仪测定烟叶拉力的稳定性和准确性，利用 Box-Behnken 原理设计三因素三水平参数，应用响应面法分析各参数对结果变异系数的影响，得到烟叶拉力测定的最优参数组合。研究含水率对烟叶拉力的影响，进一步建立烟叶含水率 X- 拉力 Y 的 BP 神经网络预测模型。【结果】响应面法分析结果显示，拉力变异系数受样品宽度的影响极显著，受测试速率影响显著，受触发力影响不明显，拉力测定的最优参数组合为样品宽度 10 mm、测试速率 0.5 mm/s、触发力 0.1 N，以此参数测定拉力的变异系数显著下降为 13.8%。烟叶的抗张强度随含水率的增大呈先增大后减小趋势，景东 C3F 含水率为 18.41% 时的抗张强度最大、为 0.456 N/mm，景东 C1F 含水率为 18.46% 时抗张强度仅为 0.288 N/mm；红河 C3F、普洱 C3F 含水率分别为 20.64%、18.47% 时的最大抗张强度分别为 0.447、0.310 N/mm；不同地区、等级烟叶的拉力存在差异。建立的烟叶含水率 X- 拉力 Y BP 神经网络预测模型，预测值与真实值吻合度较高，均方误差 MSE 为 0.04761，均方根误差 RMSE 为 0.2182。【结论】响应面分析法可用于分析烟叶拉力测定参数对结果的影响，优化后的结果稳定性提高；不同地区、等级烟叶的拉力差异显著，并且随含水率增大呈先增大后减小变化，可根据该规律选择合适的含水率，使烟叶耐加工性最佳；建立的 BP 神经网络模型的预测值误差较小、精度较好，可用于对烟叶拉力的预测。 |
| 英文摘要: |
| 【Objective】The tensile force of tobacco leaves is one of the physical characteristics of tobacco leaves,which reflects the processing resistance of tobacco leaves. Studying the tensile force characteristics of tobacco leaves can provide reference for the setting of processing parameters of threshing and redrying tobacco leaves, and further improve the economic benefits of tobacco processing. 【Method】In order to improve the stability and accuracy of measuring the tensile force of tobacco leaves by texture analyzer, three factors and three levels parameters were designed by Box-Behnken principle, and the influence of each parameter on the coefficient of variation of the results was analyzed by response surface method, and the optimal parameter combination for measuring the tensile force was obtained. The effect of moisture content on tobacco leaf tension was studied. Further, the BP neural network prediction model of moisture content X- tension Y of tobacco leaves was established. 【Result】The analysis results of response surface method show that it can be seen that the sample width has a significant influence on the coefficient of variation of tensile force, and the test rate has a significant influence, but the trigger force has no obvious influence. The optimal parameter combination was obtained: the sample width was 10 mm, the test rate was 0.5 mm/s, the trigger force was 0.1 N. The coefficient of variation of the tensile force measured by these parameters decreased significantly to 13.8%. With the increase of moisture content, the tensile strength of tobacco leaves first increased and then decreased. When the moisture content of Jingdong C3F was 18.41%, the tensile strength reached the maximum, which was 0.456 N/mm. The tensile strength of Jingdong C1F was only 0.288 N/mm, when the moisture content was 18.46%. When the moisture content of Honghe C3F and Pu 'er C3F were 20.64% and 18.47%, the maximum tensile strength were 0.447 N/mm and 0.310 N/mm respectively. There are differences in the tension of tobacco leaves in different regions and grades. The BP neural network prediction model of moisture content X- tension Y of tobacco leaves was established. The predicted value was in good agreement with the real value, with the mean square error MSE of 0.04761 and the root mean square error RMSE of 0.2182. 【Conclusion】 Response surface analysis can be used to analyze the influence of parameters on the results of tobacco tensile test, and the stability of the results is improved after the parameters are optimized. The tensile force of tobacco leaves in different regions and grades is significantly different, and it first increases and then decreases with the increase of moisture content. According to this law, the appropriate moisture content can be selected to make tobacco leaves have the best processing resistance. The established BP neural network model has small error and good accuracy, and can be used to predict the tensile force of tobacco leaves. |
|
查看/发表评论 下载PDF阅读器 |
