Scholarlink

<p>Uneven intra-row seed placement remains a major constraint in precision sorghum planting, reducing stand uniformity, water-use efficiency, and ultimately yield. This study aimed to develop, validate, and embed a Simulated Annealing–optimized Adaptive Neuro-Fuzzy Inference System (SA-ANFIS) for predicting sorghum intra-row seed spacing as a function of soil moisture and planter forward speed. A two-factor factorial experiment (4 speed levels × 6 moisture levels × 5 replications = 120 observations) was conducted using an instrumented smart-planter prototype equipped with a capacitive soil-moisture sensor and two rotary encoders. Data were screened against predefined exclusion rules, normalized using training-set statistics only, and partitioned 70:30 (84 training, 36 testing). A first-order Sugeno ANFIS with 9 rules was trained using hybrid learning, after which its antecedent (premise) parameters were tuned by Simulated Annealing. Performance was assessed by MSE, RMSE, coefficient of determination (R²), residual analysis, Bland–Altman agreement analysis, and paired statistical tests. The optimized rule base was subsequently deployed on an Arduino Mega 2560 controller. Optimization reduced normalized training MSE from 0.0028 to 0.0011. On the independent test set, RMSE reduced prediction error from 0.8171 cm (R² = 0.830) for the baseline to 0.4595 cm (R² = 0.898) for SA-ANFIS, corresponding to a 43.8% reduction in prediction error. Paired t-tests confirmed the improvement to be statistically significant (p &lt; 0.05). Residual analysis showed randomly distributed errors with no apparent systematic trend, while Bland–Altman analysis demonstrated near-zero mean bias and more than 95% of observations within the limits of agreement. The optimized controller executed reliably on the Arduino Mega 2560 within a 41 ms control cycle. The proposed SA-ANFIS framework provides a reproducible, statistically validated, and embedded-ready approach to intelligent sorghum seed-spacing control. Although validation was limited to prototype-scale trials at a single site, the demonstrated accuracy and real-time feasibility establish a sound foundation for subsequent hardware-in-the-loop testing and multi-location field validation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development of a simulated annealing–adaptive neuro-fuzzy inference system (SA-ANFIS) model for sorghum seed intra-row spacing optimization

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月24日

查看全文
  • Ibrahim Abubakar,
  • Abdullahi Ibrahim Mohammed,
  • Balami Audu Ayuba,
  • Idah Peter Abah

摘要

Uneven intra-row seed placement remains a major constraint in precision sorghum planting, reducing stand uniformity, water-use efficiency, and ultimately yield. This study aimed to develop, validate, and embed a Simulated Annealing–optimized Adaptive Neuro-Fuzzy Inference System (SA-ANFIS) for predicting sorghum intra-row seed spacing as a function of soil moisture and planter forward speed. A two-factor factorial experiment (4 speed levels × 6 moisture levels × 5 replications = 120 observations) was conducted using an instrumented smart-planter prototype equipped with a capacitive soil-moisture sensor and two rotary encoders. Data were screened against predefined exclusion rules, normalized using training-set statistics only, and partitioned 70:30 (84 training, 36 testing). A first-order Sugeno ANFIS with 9 rules was trained using hybrid learning, after which its antecedent (premise) parameters were tuned by Simulated Annealing. Performance was assessed by MSE, RMSE, coefficient of determination (R²), residual analysis, Bland–Altman agreement analysis, and paired statistical tests. The optimized rule base was subsequently deployed on an Arduino Mega 2560 controller. Optimization reduced normalized training MSE from 0.0028 to 0.0011. On the independent test set, RMSE reduced prediction error from 0.8171 cm (R² = 0.830) for the baseline to 0.4595 cm (R² = 0.898) for SA-ANFIS, corresponding to a 43.8% reduction in prediction error. Paired t-tests confirmed the improvement to be statistically significant (p < 0.05). Residual analysis showed randomly distributed errors with no apparent systematic trend, while Bland–Altman analysis demonstrated near-zero mean bias and more than 95% of observations within the limits of agreement. The optimized controller executed reliably on the Arduino Mega 2560 within a 41 ms control cycle. The proposed SA-ANFIS framework provides a reproducible, statistically validated, and embedded-ready approach to intelligent sorghum seed-spacing control. Although validation was limited to prototype-scale trials at a single site, the demonstrated accuracy and real-time feasibility establish a sound foundation for subsequent hardware-in-the-loop testing and multi-location field validation.