Alcohol content and original wort concentration are key indicators of beer quality. The detection of these metrics and the authentication of beer authenticity are crucial for protecting consumer rights. To this end, this study investigates quantitative detection methods for beer alcohol content and original wort concentration based on near-infrared spectroscopy (NIRS), as well as authenticity verification methods for craft, industrial, and non-fermented beers. Convolutional neural networks combined with a long short-term memory networks (CNN-LSTM) feature extraction method was proposed for establishing multiple regression models and partial least squares discriminant analysis (PLS-DA) model. The results indicate that the CNN-LSTM combined with the support vector machine regression demonstrates optimal performance, with coefficients of determination exceeding 0.99 for the alcohol content calibration, validation, and independent test sets, and all relative root mean square errors below 2.67%. For original wort concentration, the coefficients of determination exceeded 0.97 across the calibration, validation, and independent test sets, with relative root mean square errors below 4.05%. The CNN-LSTM combined with the PLS-DA approach exhibited the lowest variable dimension while achieving 100% classification accuracy. This method offers rapid, non-destructive, and efficient advantages, making it suitable for beer quality control and market regulation.

To address traditional beer detection limitations, this study proposes a rapid NIRS-based method for detecting key indicators and verifying authenticity. Designing Single-task (STL) and Multi-task learning (MTL) strategies, it employs Variable Importance in Projection for wavelength selection. Deep spectral features were extracted utilizing a Multi-Head Attention (MHA)-fused Convolutional Neural Network (CNN-MHA), Long Short-Term Memory (LSTM-MHA), and hybrid CNN-LSTM-MHA networks. To further enhance model performance, the Bayesian Optimization Algorithm globally optimized network hyperparameters in STL, alongside hyperparameters and multi-task loss weights in MTL. Partial least squares regression, support vector machine regression, and partial least squares discriminant analysis models were established using these features. Results indicate that the MTL-based CNN-LSTM-MHA network effectively learns shared features across multiple tasks, significantly improving model generalization. Specifically, the coefficients of determination ($R^2$) for alcohol content and original wort concentration in the validation set were 0.996 and 0.997, respectively, with relative root mean square errors (rRMSE) of 2.024% and 2.515%. In the independent test set, the $R^2$ were 0.995 and 0.991, with rRMSE of 2.515% and 2.087%, respectively. Furthermore, 100% classification accuracy was achieved across all datasets. This method provides an efficient technical solution for beer market regulation and real-time detection in production processes.

To address the time-delay and nonlinear characteristics of liquid level control in near-infrared spectroscopybased liquid phase detection equipment, as well as the pipeline cavitation issues caused by improper sample pump speed settings during sample delivery-which may result in air bubble retention within the cuvette and subsequently degrade spectral data quality-a dual-buffer bottle sample delivery system model was established. A Proportional-Integral-Derivative (PID) controller was designed, and an enhanced hybrid algorithm integrating the Particle Swarm Optimization (PSO) algorithm and the Sparrow Search Algorithm (SSA) was proposed. The hybrid algorithm, referred to as the Adaptive Chaotic Mapping Particle Swarm Sparrow Algorithm (ACM-PSSA), incorporates Tent chaotic mapping for population initialization, a nonlinear cosine-based adaptive sparrow classification strategy, and a master-slave optimization mechanism wherein SSA performs global exploration and PSO executes local exploitation to optimize PID parameters. Simulation results demonstrate that ACM-PSSA outperforms traditional SSA and PSO across six benchmark test functions in terms of convergence speed, accuracy, and stability. When applied to the liquid level control of the dual-buffer bottle system, the optimized controller achieved a rise time of 0.188 seconds, a settling time of 1.211 seconds, and an overshoot reduced to 20.98%. By leveraging chaotic mapping, adaptive classification, and a master-slave optimization framework, ACM-PSSA effectively overcomes the limitations of conventional SSA and PSO, significantly enhancing the efficiency of PID parameter optimization and the overall control performance of the dual-buffer bottle sample delivery system.