Pigs exposed to stressors might change their daily typical feeding intake pattern. The objective of this study was to develop a method for the early identification of deviations from an individual pig’s typical feeding patterns. In addition, a general approach was proposed to model feed intake and real-time individual nutrient requirements for pigs with atypical feeding patterns. First, a dynamic linear model (DLM) was proposed to model the typical daily feed intake (DFI) and daily gain (DG) patterns of pigs. Individual DFI and DG dynamics are described by a univariate DLM in conjunction with Kalman filtering. A standardized tabular cumulative sum (CUMSUM) control chart was applied to the forecast errors generated by DLM to activate an alarm when a pig showed deviations from its typical feeding patterns. The relative feed intake (RFI) during a challenge period was calculated. For that, the forecasted individual pig DFI is expressed as its highest DFI relative to the intake during pre-challenge period. Finally, the DLM and RFI approaches were integrated into the actual precision-feeding model (original model) to estimate real-time individual nutrient requirements for pigs with atypical feeding patterns. This general approach was evaluated with data from two studies (130 pigs, at 35.25 ± 3.9 kg of initial BW) that investigated during 84 days the effect of precision-feeding systems for growing-finishing pigs. The proposed general approach to estimating real-time individual nutrient requirements (updated model) was evaluated by comparing its estimates with those generated by the original model. For 11 individuals out of 130, the DLM did not fit the observed data well in a specific period, resulting in an increase in the sum of standardized forecast errors and in the number of time steps that the model needed to adapt to the new patterns. This poor fit can be identified by the increase in the CUMSUM with a consequent alarm generated. The results of this study show that the updated model made it possible to reduce intra-individual variation for the estimated lysine requirements in comparison with the original model, especially for individuals with atypical feeding patterns. In conclusion, the DLM in conjunction with CUMSUM could be used as a tool for the online monitoring of DFI for growing-finishing pigs. Moreover, the proposed general approach allows the estimation of real-time amino acid requirements and accounts for the reduced feed intake and growth potential of pigs with atypical feeding patterns.

暴露于应激源下的猪可能会改变其日常典型的采食方式。这项研究的目的是开发一种方法，以便及早发现与个体猪的典型喂养方式之间的偏差。此外，提出了一种通用方法来模拟具有非典型喂养方式的猪的采食量和实时个体营养需求。首先，提出了动态线性模型（DLM），以模拟典型的猪日采食量（DFI）和日增重（DG）模式。单个DFI和DG动态由单变量DLM结合卡尔曼滤波来描述。标准化表格累加总和（CUMSUM）将控制图应用于DLM产生的预测误差，以在猪显示出偏离其典型喂养方式的警报时触发警报。相对饲料摄入量（RFI）在挑战期间进行了计算。为此，将预测的个体猪DFI表示为其在挑战前期相对于采食量的最高DFI。最后，将DLM和RFI方法集成到实际的精确喂养模型（原始模型）中，以估计具有非典型喂养模式的猪的实时个体营养需求。通过两项研究（130头猪，初始体重为35.25±3.9千克）的数据对这种通用方法进行了评估，该研究在84天的时间内研究了精确饲喂系统对育肥猪的影响。通过将其估算值与原始模型生成的估算值进行比较，评估了估算实时个体营养需求量（更新模型）的一般方法。对于130个人中的11个人，DLM在特定时期内无法很好地拟合观察到的数据，从而导致标准预测误差的总和以及模型适应新模式所需的时间步数增加。这种不适感可以通过CUMSUM的增加来识别，并随之产生警报。这项研究的结果表明，与原始模型相比，更新的模型使减少赖氨酸需求量的个体内部差异成为可能，特别是对于具有非典型喂养方式的个体。总之，DLM与CUMSUM一起可以用作在线监测生长肥育猪的DFI的工具。此外，