Dataset

In livestock- and poultry-breeding facilities, a precise, rapid, and affordable method for detecting ammonia concentrations is essential. We design and develop an electronic nose system containing a bionic chamber that imitates the nasal-cavity structure of humans and canines. The sensors are positioned based on fluid simulation results. Response data for ammonia and ethanol gases and the response/ recovery times of an ammonia sensor under three concentrations are collected using the electronic nose system. Response data for ammonia and ethanol gases collected using the electronic-nose system are classified and regressed using a sparrow search algorithm (SSA)-optimized backpropagation neural network (BPNN). The results show that the sensor has a relative mean deviation of 1.45%. The average response time of the ammonia sensor in the chamber is 13 s slower than that of the sensor directly exposed to the gas being measured, while the average recovery time is 19 s faster. In tests comparing the performance of the SSA-BPNN, support vector machine (SVM), and random forest (RF) models, the SSA-BPNN achieves a 100% classification accuracy, better than the SVM and RF models. It also outperforms the other models at regression prediction, with smaller absolute, mean absolute, and root mean square errors. Its coefficient of determination (R²) is greater than 0.99, surpassing those of the SVM and RF models. The theoretical and experimental results both indicate that the proposed electronic nose system containing a bionic chamber, when used with the SSA-BPNN, offers a promising approach for detecting ammonia in livestock- and poultry-breeding facilities.

在畜禽养殖设施中，精准、快速且成本可控的氨气浓度检测方法至关重要。本研究设计并开发了一款搭载仿生腔室的电子鼻系统，该腔室可模拟人类与犬类的鼻腔结构，传感器布局依据流体仿真结果确定。本电子鼻系统采集了氨气与乙醇气体的响应数据，以及三种浓度下氨气传感器的响应与恢复时间。针对采集到的氨气与乙醇气体响应数据，本研究采用麻雀搜索算法（Sparrow Search Algorithm, SSA）优化的反向传播神经网络（Backpropagation Neural Network, BPNN）开展分类与回归分析。实验结果表明，该传感器的相对平均偏差为1.45%。腔室内氨气传感器的平均响应时间比直接暴露于待测气体的传感器慢13秒，但其平均恢复时间则快19秒。在对比SSA-BPNN、支持向量机（Support Vector Machine, SVM）与随机森林（Random Forest, RF）模型性能的实验中，SSA-BPNN的分类准确率达100%，优于SVM与RF模型；其在回归预测任务中同样表现更优，绝对误差、平均绝对误差与均方根误差均更小，决定系数（R²）大于0.99，优于SVM与RF模型的对应指标。理论与实验结果均表明，所提出的搭载仿生腔室的电子鼻系统结合SSA-BPNN模型，为畜禽养殖设施中的氨气检测提供了一种极具应用前景的解决方案。