均方误差
人工神经网络
反向传播
响应面法
人工智能
纹理(宇宙学)
计算机科学
模式识别(心理学)
多层感知器
生物系统
绝对偏差
近似误差
Rprop公司
相关系数
感知器
决定系数
均方根
标准差
特征(语言学)
刚度
平均绝对误差
数学
机器学习
感觉系统
特征提取
卷积神经网络
深度学习
均方
作者
Lingqi Li,Yang Ye,Zhen Li,Zongyun Yang,Peng Wang,Xinglian Xu,Minghao Bian,Xin Chen,Qingqin Lv
标识
DOI:10.1016/j.afres.2026.102693
摘要
Currently, global experts assessing wooden breast (WB) lack uniform training standards for WB models and rely heavily on written descriptions and team experience without intuitive references. Therefore, developing WB-grade-specific hydrogel simulant sensory mimics that can mimic chicken breast texture and morphology is critical. This study used response surface methodology (RSM) to optimize the factors influencing hydrogel mimic texture (high acyl gellan, gelatin, and calcium ions) while developing and comparing the performance of a backpropagation artificial neural network (BP-ANN) with that of RSM. Hydrogel mimics of different WB grades, developed via optimized parameters, were validated by employing relevant mechanical indicators (compressive force, stress–strain, and creep) and sensory evaluations, achieving a texture highly similar to that of authentic WB. The relative error of the tail’s sagging angle between the mimics and authentic WB was controlled within 2%. Under 10-fold cross-validation, the BP-ANN model with an eight-neuron hidden layer achieved a coefficient of determination (R 2 ) of 0.9976, root mean square error (RMSE) of 150.6202, and absolute average deviation (ADD%) of 1.3515, thus outperforming the RSM model, which had an R 2 of 0.9960, RMSE of 192.1336, and ADD% of 2.9288. When full datasets were used, the BP-ANN model again demonstrated higher accuracy (R 2 = 0.9978, RMSE = 125.3869, ADD% = 1.0068) than RSM. This study provides strong evidence of the potential of hydrogels to simulate meat texture and drape, establishing a practical reference for WB morphology and texture and supporting future WB grading processes.
科研通智能强力驱动
Strongly Powered by AbleSci AI