基于多传感器融合的鸡蛋壳强度-厚度-刚度无损检测系统研究

Nondestructive Detection of Eggshell Strength, Thickness, and Stiffness Based on Multi-sensor Fusion System

摘要:

为了实现鸡蛋壳强度-厚度-刚度的高效、无损评估，本研究设计并构建了一套基于声振响应的多传感器激励与信号采集系统。所提出的电磁激励机构采用通电吸引-断电自由冲击的激励模式，有效克服了传统通电冲击激励在冲击控制、电参数调节及线圈寿命等方面的局限，提升了激励过程的稳定性。针对电磁激励高度参数开展系统优化，确定了兼顾激励稳定性与信号响应质量的最优激励高度。通过该系统，可稳定获取鸡蛋在激励作用下的声学与振动响应信号，并从中提取频域信号特征变量进行融合建模。基于特征级融合策略，分别构建了CARS-PLSR与PCA-PLSR回归模型，实现对蛋壳厚度、破壳强度及壳体刚度3项质量指标的预测。验证试验结果表明，CARS-PLSR模型在蛋壳厚度预测中Rp=0.847、RMSEP为0.0130 mm、RPD为1.84;PCA-PLSR模型在强度与刚度预测中Rp=0.816、RMSEP为3.56 N、RPD为1.73和Rp=0.859、RMSEP为7.34 N/mm、RPD为1.93，均达到较高预测精度。此外，对3个质量指标最佳融合模型进行了SHAP值的模型可解释性分析，揭示了振动信号在融合建模中的辅助价值。研究结果验证了声学与振动信号融合用于鸡蛋质量多指标无损检测可行性与有效性，为鸡蛋产品质量分级和检测提供了技术支撑。

Abstract:

Aiming to achieve efficient and nondestructive evaluation of eggshell strength, thickness, and stiffness, a multi-sensor excitation and signal acquisition system was designed and constructed based on acoustic-vibration responses. The proposed electromagnetic excitation mechanism employed a power-on attraction and power-off free-impact excitation mode. This effectively overcame the limitations of traditional power-on impact excitation regarding impact control, electrical parameter adjustment, and coil lifespan, thereby enhancing the stability of the excitation process. Systematic optimization of the electromagnetic excitation height parameter was conducted, identifying the optimal height that balanced excitation stability and signal response quality. Using this system, acoustic and vibration response signals of eggs under excitation can be stably acquired, from which frequency-domain signal characteristic variables were extracted for fusion modeling. Based on a feature-level fusion strategy, both CARS-PLSR and PCA-PLSR regression models were constructed to predict three quality indicators: shell thickness, breaking strength, and shell stiffness. Validation results demonstrated: the CARS-PLSR model achieved Rp=0.847, RMSEP=0.0130 mm, RPD=1.84 for shell thickness prediction;the PCA-PLSR model achieved Rp=0.816, RMSEP=3.56 N, RPD=1.73 for strength prediction, and Rp=0.859, RMSEP=7.34 N/mm, RPD=1.93 for stiffness prediction, all reaching high prediction accuracy. Furthermore, SHAP-based model interpretation was conducted on the optimal fusion models for the three quality indicators, revealing the auxiliary contribution of vibration features in the fusion modeling process. The results verified the feasibility and effectiveness of fusing acoustic and vibration signals for the non-destructive multi-indicator detection of eggshell quality, providing technical support for quality grading and inspection of egg products.