Low-cost, accurate monitoring of macronutrient ions in soils, plants, and water is highly desired to improve fertilizer management for maximum profitability and minimum negative environmental impacts. Traditional ion-selective electrodes (ISEs) suffer from interference from non-target ions. This paper reports the integration of artificial neural networks (ANNs) and a miniature sensor containing an array of three ISE-based sensing elements to improve accuracy of the sensor in detecting and quantifying target nitrate ( ${\mathrm {NO}}_{3}^{-}$ ), phosphate (H ₂ ${\mathrm {PO}}_{4}^{-}$ ), and potassium (K ⁺ ) ions in the environment. The sensor outputs of ${\text{NO}}_{3}^{-}$ , H ₂ ${\text{PO}}_{4}^{-}$ , and K ⁺ ion concentrations are used to train and optimize ANNs. The optimized neural networks are applied to classify and estimate concentrations of the target ions in the presence of interfering ions. The ANN-assisted array of sensing elements reduces cross-sensitivity between these elements. The present sensor is validated with measurements of ${\text{NO}}_{3}^{-}$ , $\text{H}_{2} {PO}_{4}^{-}$ , and K ⁺ ions in soil solution, plant sap, and tile drainage water from crop fields.

中文翻译：

---

集成人工神经网络的微型多离子传感器

非常需要对土壤、植物和水中的常量营养素离子进行低成本、准确的监测，以改善肥料管理，以实现最大的盈利能力和最小的对环境的负面影响。传统的离子选择性电极 (ISE) 会受到非目标离子的干扰。本文报告了人工神经网络 (ANN) 和包含三个基于 ISE 的传感元件阵列的微型传感器的集成，以提高传感器在检测和量化目标硝酸盐方面的准确性。 ${\mathrm {NO}}_{3}^{-}$ ), 磷酸盐 (H ₂ ${\mathrm {PO}}_{4}^{-}$ ) 和环境中的钾 (K ⁺ ) 离子。传感器输出 ${\text{NO}}_{3}^{-}$ , H ₂ ${\text{PO}}_{4}^{-}$ 和 K ⁺离子浓度用于训练和优化人工神经网络。优化的神经网络用于在存在干扰离子的情况下对目标离子的浓度进行分类和估计。ANN 辅助的传感元件阵列降低了这些元件之间的交叉敏感性。本传感器通过以下测量进行验证 ${\text{NO}}_{3}^{-}$ , $\text{H}_{2} {PO}_{4}^{-}$ , 以及土壤溶液、植物汁液和农田瓷砖排水中的K ⁺离子。