A rapid detection device for the pungency intensity of zanthoxylum bungeanum and application thereof

Abstract

The application belongs to the field of quality detection of agricultural products, and relates to a rapid detection device for the pungency intensity of Chinese prickly ash and application. The device comprises a human-computer interaction core host computer interaction module for realizing instruction and data interaction between a user and the detection device; a control and signal processing central circuit board module for coordinating execution of the entire detection process and data processing; and a three-electrode detection cell module of a sensing and signal generation unit for constructing a stable electrochemical detection environment and sensing electrochemical characteristics of pungent substances in Chinese prickly ash. The device can quickly and accurately output a quantitative value of pungency, significantly improving the objectivity, resolution and efficiency of detection. Moreover, the device overcomes the inapplicability of general electrochemical equipment in this specific application scenario of Chinese prickly ash, can directly display the pungency value and grade result, simplifies the operation process, is convenient to use, is convenient to popularize in on-site environments such as raw material purchase and production quality control, and solves the problem that large laboratory equipment cannot meet the on-site rapid detection demand.

Description

Technical Field

[0001] This invention belongs to the field of agricultural product quality testing, specifically relating to a rapid detection device for the intensity of the numbing sensation of Sichuan pepper and its application. Background Technology

[0002] Sichuan peppercorn( Zanthoxylum bungeanum As a traditional and distinctive spice in my country, the unique "numbing" flavor of Sichuan pepper is a core indicator for evaluating its quality, economic value, and regional characteristics. The numbing substances mainly originate from amide compounds in Sichuan pepper, and their content and composition directly determine the intensity and persistence of the numbing flavor. In the entire industrial chain of Sichuan pepper—from breeding, harvesting, processing, storage, trading, and end-product production—rapid, objective, and accurate quantitative evaluation of the numbing intensity is crucial for achieving product quality standardization and control, ensuring consistent consumer experience, and is of great significance for promoting the modernization and high-quality development of the Sichuan pepper industry.

[0003] Currently, the assessment methods for the intensity of the numbing sensation of Sichuan peppercorns mainly rely on human sensory evaluation (GB / T 38495-2020) and instrumental analysis, but both have significant limitations. Human sensory evaluation is highly subjective, lacks stability and reproducibility, and is inefficient, making it difficult to meet the needs of rapid on-site testing. Among instrumental analysis methods, while electronic tongue technology can achieve preliminary differentiation, it is essentially a semi-quantitative method with limited detection accuracy. Furthermore, although existing electrochemical sensing technologies have revealed the correlation between electrochemical signals and numbing substances, they generally employ large, non-dedicated general-purpose electrochemical analyzers, which are complex to operate. Existing models often have limited features and fail to fully utilize the complementary information from different electrochemical technologies (such as CV and DPV), resulting in limited prediction accuracy and robustness. Moreover, there is a lack of dedicated integrated solutions from hardware to algorithms.

[0004] In conclusion, the Sichuan pepper industry urgently needs a dedicated testing device that can overcome the aforementioned shortcomings and integrate speed, accuracy, quantitative analysis, and ease of operation. Summary of the Invention

[0005] To address the lack of dedicated equipment for rapid quantitative detection of the numbing and spicy flavor intensity of Sichuan peppercorns, this invention provides a portable rapid detection device for the numbing and spicy flavor intensity of Sichuan peppercorns. Through the integrated structure of the outer shell, the host computer interaction module, the circuit board module, and the three-electrode detection cell module, rapid quantitative detection of the numbing and spicy flavor intensity of Sichuan peppercorns is achieved.

[0006] To achieve the above objectives, the present invention adopts the following technical solution.

[0007] A rapid detection device for the intensity of Sichuan pepper numbing flavor includes: a shell, a host computer interaction module, a circuit board module, and a three-electrode detection cell module; Host computer interaction module: As the core of human-computer interaction of the device, it is used to realize the interaction of instructions and data between the user and the detection device. Specifically, the host computer interaction module is used to receive and set electrochemical detection parameters, send control instructions to the circuit board module to start or stop detection, display the volt-ampere curve generated during the detection process in real time, and finally output and display the quantitative value and grade evaluation result of the intensity of Sichuan pepper numbing flavor.

[0008] The electrochemical detection parameters include the potential scan range and the scan method.

[0009] Circuit board module: As the control and signal processing center of the device, it coordinates the execution of the entire detection process and data processing. Specific functions include: generating an electrochemical excitation signal according to the instructions issued by the host computer interaction module and applying it to the three-electrode detection cell module; synchronously, acquiring the response current signal generated by the three-electrode detection cell module under the action of the excitation signal, and conditioning, amplifying and converting the signal to analog-to-digital to obtain a pre-processed electrochemical signal; further performing intelligent analysis and feature extraction on the pre-processed electrochemical signal, and finally calculating the quantitative detection result of Sichuan pepper numbing flavor through the algorithm model (CDPV-CNN numbing flavor intensity prediction model), and sending the result back to the host computer interaction module.

[0010] The electrochemical excitation signal includes the potential waveforms required for cyclic voltammetry (CV) and differential pulse voltammetry (DPV); the circuit board module has a built-in numbing flavor intensity prediction model, which can intelligently analyze and extract features from the preprocessed electrochemical signal.

[0011] Three-electrode detection cell module: As the sensing and signal generation unit of the device, it is used to construct a stable electrochemical detection environment and sense the electrochemical characteristics of the numbing and spicy substances in Sichuan pepper. By immersing the three-electrode system in the Sichuan pepper sample solution, under the excitation signal applied by the circuit board module, a specific electrochemical reaction for the numbing and spicy substances occurs at the working electrode interface, thereby generating a response current signal that is closely related to the type and concentration of the numbing and spicy substances.

[0012] Furthermore: the housing includes a housing body and a power switch; the housing body provides structural support and physical protection for the entire device; the power switch is located on the surface of the housing body and is used to control the device to be powered on and off; The host computer interaction module includes a host computer display screen; the host computer display screen is embedded in the surface of the outer shell and is used for parameter setting, control command issuance, real-time display of detection curves and quantitative results of numbing flavor intensity, and is displayed through the host computer display screen; Circuit board module: The circuit board module is a printed circuit board integrated inside the housing, including a main control unit, potentiostat unit one, potentiostat unit two, signal conditioning unit, microcurrent detection unit, and power supply; wherein, the main control unit is used to generate control commands to drive potentiostat unit one and potentiostat unit two to generate specific electrochemical excitation signals and apply them to the three-electrode detection cell module; the microcurrent detection unit and signal conditioning unit are used to collect the response current signal generated in the three-electrode detection cell module and transmit it back to the main control unit; the main control unit can process and analyze the received signals and finally obtain the quantitative detection result of Sichuan pepper numbing flavor; Three-electrode detection cell module: The three-electrode detection cell module includes a support and a three-electrode system; the support is used to fix the three-electrode system to ensure that it is stably immersed in the solution to be tested; the three-electrode system is used to collect the electrochemical signal corresponding to the numbing flavor characteristic of Sichuan pepper.

[0013] Preferably, the three-electrode system is provided in two groups, namely the CV three-electrode system and the DPV three-electrode system; the CV three-electrode system includes a RE reference electrode one, a CE counter electrode one, and a WE working electrode one; the DPV three-electrode system includes a RE reference electrode two, a CE counter electrode two, and a WE working electrode two.

[0014] Preferably, the first WE working electrode and the second WE working electrode are one of glassy carbon electrode, gold electrode or platinum electrode.

[0015] Preferably, the bracket includes a support column and two electrode fixing disks: a first electrode fixing disk and a second electrode fixing disk. The first electrode fixing disk and the second electrode fixing disk have precise positioning holes for fixing the three-electrode system.

[0016] Preferably, the host computer display screen is a touch screen, which integrates parameter configuration, detection and startup, data storage and historical query functions; the microcontroller used in the main control unit is an STM32 series microcontroller.

[0017] This invention also provides an operating method for a rapid detection device for the numbing sensation of Sichuan peppercorns, characterized by comprising the following steps: S1 Sample Preparation: A quantitative sample of Sichuan pepper to be tested was crushed, sieved, and then extracted with ethanol solution by ultrasonication. After centrifugation, the Sichuan pepper extract was prepared. The Sichuan pepper extract was mixed with phosphate buffer solution in a certain proportion to prepare the test solution. Preferably, the volume ratio of Sichuan pepper extract to phosphate buffer is 1:9.

[0018] S2 Electrode Installation and Pretreatment: The glassy carbon working electrode is subjected to surface grinding, cleaning and ultrasonic pretreatment to remove the oxide layer on the electrode surface; the two sets of RE reference electrode 1, WE working electrode 1, CE counter electrode 1 and RE reference electrode 2, WE working electrode 2, CE counter electrode 2 used for cyclic voltammetry and differential pulse voltammetry are fixed by electrode fixing disk 1 and electrode fixing disk 2 respectively and immersed below the surface of the solution to be tested; S3 Signal Acquisition and Processing: CV+DPV synchronous scanning is initiated via the host computer display screen. The main control unit generates a corresponding excitation waveform and applies it to the three-electrode detection cell, generating a volt-ampere signal of the Sichuan pepper numbing substance at the working electrode interface in the detection cell. The micro-current detection unit converts the response current in the three-electrode detection cell into a voltage signal. After filtering, amplification, and analog-to-digital conversion, the signal is transmitted back to the main control unit for signal processing and analysis. S4 Numbing Flavor Intensity Prediction and Display: The main control unit calls the numbing flavor intensity prediction model (i.e., CDPV-CNN numbing flavor intensity prediction model) to intelligently analyze the collected volt-ampere curve data, output the quantitative value of numbing flavor intensity and the level evaluation result, and display it on the upper computer display screen; the CDPV-CNN numbing flavor intensity prediction model integrates the dynamic characteristics of the CV curve and the peak characteristics of the DPV curve, and after weighted fusion through a cross-modal attention mechanism, outputs the numbing flavor intensity value through a regression network.

[0019] The construction of the CDPV-CNN numbing flavor intensity prediction model includes the following steps: S4-1, Data Preparation and Preprocessing: Collect Sichuan pepper samples (including green and red Sichuan peppers), select a total of n groups of Sichuan pepper samples (n is a positive integer, preferably 1584 groups), and perform simultaneous CV and DPV scanning data to obtain the corresponding n groups of CV data and n groups of DPV data.

[0020] Each set of data includes CV voltammetry curves (potential range -0.2V to +1.2V, 512 sampling points) and DPV voltammetry curves (potential range -0.2V to +1.2V, 512 sampling points), and corresponds to the sensory evaluation calibration benchmark value of numbing flavor intensity (using the gLMS method); the raw CV data and DPV data are preprocessed to obtain preprocessed data; The preprocessing steps are as follows: First, a Savitzky-Golay filter (window size 9, polynomial order 3) is applied to eliminate high-frequency noise; second, wavelet packet transform (wavelet basis 'db4', decomposition level 5) is used for signal denoising; finally, normalization is performed using the formula... To achieve feature normalization, where μ is the mean and σ is the standard deviation; S4-2, Dataset Partitioning: The preprocessed data from S4-1 is divided into training and testing sets proportionally to ensure the statistical reliability of model training and evaluation.

[0021] Ideally, the training set should comprise 65-75% of the data, with the remaining data serving as the test set. S4-3, Model Architecture Design: The CDPV-CNN numbing flavor intensity prediction model uses a two-branch (CV branch and DPV branch) one-dimensional convolutional structure to process CV and DPV data respectively: CV branch: Input the CV voltammetry curve and extract redox kinetic features, including electron transfer number (α), peak potential difference (ΔEp), and integrated charge (Q) through a one-dimensional convolutional layer (convolutional kernel size of 5 and stride of 1). DPV branch: Input DPV current-voltage curve, extract peak features through parallel one-dimensional convolutional layer (convolutional kernel size of 5, stride of 1), including peak current (ip), half width at half maximum (FWHM) and relaxation time (τ). Feature fusion layer: Introduces a cross-modal attention mechanism to adaptively weight and fuse CV and DPV features; the attention weight calculation formula is as follows: Among them, w i This represents the cross-modal attention weight of the i-th element; f cv , i and f dpv , i Let represent the value of the i-th element in the CV feature vector and the DPV feature vector, respectively; j is the summation traversal index. f cv , j and f dpv , j Let represent the value of the j-th element in the CV and DPV feature vectors, respectively; Through the above feature fusion layer operation, the synergistic response behavior of hydroxy-α-sanshool and hydroxy-β-sanshool is enhanced, resulting in a weighted fusion feature mapping vector; Preferably, the fused feature vector outputs a 32-dimensional feature vector. The regression output layer maps the weighted fused feature vector to a quantitative value of numbing sensation (range 0-100%) through a fully connected network (containing 64 nodes) and outputs the level evaluation result; the Huber loss function is used to optimize the model's robustness, and its formula is: Where L δ This represents the Huber loss function value. y This represents the true intensity of the numbing sensation. y ^ represents the predicted value. δSet as the threshold parameter (set to 1.0); This step completes the architectural design of the prediction model and yields the initial numbing flavor intensity prediction model.

[0022] S4-4, Model Training and Optimization: Using the training set data obtained in S4-2, the initial numbing sensation intensity prediction model designed in S4-3 was trained. During the training process, the loss function and validation set performance were monitored to ensure model convergence. Through training, the preliminary CDPV-CNN numbing sensation intensity prediction model was obtained. The optimal training parameters included: using the Adam optimizer with an initial learning rate of 0.001; a batch size of 32; dynamic control of the training cycle through an early stopping strategy; and the introduction of Dropout mechanism (ratio of 0.2) and L2 regularization (coefficient of 0.01) to suppress overfitting. S4-5, Model Validation and Performance Evaluation: Using the test set obtained from S4-2, the performance of the preliminary CDPV-CNN numbing flavor intensity prediction model trained in S4-4 was evaluated; the coefficient of determination (R²) was calculated. 2 ) and root mean square error (RMSE), where R 2 A value not lower than 0.95 and an RMSE not higher than 2.5% indicate that the model has high accuracy and generalization ability. Therefore, this model is determined as the final prediction model for the numbing flavor intensity (i.e., CDPV-CNN numbing flavor intensity prediction model), which is suitable for rapid quantitative detection of the numbing flavor intensity of Sichuan pepper.

[0023] Through the above steps, a dedicated CDPV-CNN numbing flavor intensity prediction model was constructed and integrated into the device of this invention to realize the quantitative value and grade evaluation of the numbing flavor intensity of Sichuan pepper.

[0024] Beneficial effects: (1) Highly specialized, solving the problem that general electrochemical equipment cannot directly detect the intensity of Sichuan pepper numbing flavor: This device is designed for Sichuan pepper numbing flavor intensity detection from hardware structure to software algorithm, realizing the whole process integration from Sichuan pepper extract to electrolyte solution to Sichuan pepper numbing flavor intensity display, overcoming the inapplicability of general electrochemical equipment in the specific application scenario of Sichuan pepper.

[0025] (2) It achieves rapid and accurate quantitative detection, overcoming the limitations of manual evaluation and semi-quantitative methods: traditional methods rely on human sensory evaluation, which is highly subjective and inefficient. This invention integrates dual electrochemical fingerprint information through CV+DPV synchronous scanning mode and combines it with CDPV-CNN fusion algorithm model, which can quickly and accurately output quantitative values ​​of numbing flavor intensity, significantly improving the objectivity, resolution and efficiency of detection.

[0026] (3) The device is portable and easy to use, which improves the convenience and scalability of testing: The device has a compact structure and a built-in host computer interactive interface, which can directly display the numbing intensity value and grade result, simplifying the operation process, reducing the professional requirements for operators, and making it easy to promote and use in on-site environments such as raw material acquisition and production quality control, thus solving the problem that large laboratory equipment cannot meet the needs of rapid on-site testing. Attached Figure Description

[0027] Figure 1 This is a front view of the detection device of the present invention.

[0028] Figure 2 This is an isometric view of the internal structure of the detection device of the present invention.

[0029] Figure 3 This is a front view of the internal structure of the detection device of the present invention.

[0030] Figure 4 This is a top view and connection diagram of the three-electrode detection cell of the present invention.

[0031] Figure 5 This is an isometric view of the control circuit board of the present invention.

[0032] Figure 6 This is a block diagram of the core control unit structure of the detection system of the present invention.

[0033] Figure 7 This is a schematic diagram of the pulse signal conditioning circuit of the present invention.

[0034] Figure 8 This is a circuit diagram of the potentiostat of the present invention.

[0035] Figure 9 This is a schematic diagram of the potential boosting circuit of the present invention.

[0036] Figure 10 This is a schematic diagram of the feedback resistor gating circuit for adjusting the current detection range according to the present invention.

[0037] Figure 11 This is a schematic diagram of the operation and data display process of the host computer software of the present invention.

[0038] Figure 12 This is a flowchart of the signal processing of the hardware circuit of the present invention.

[0039] Figure 13 This is a schematic diagram of the UI layout of the host computer interface of the present invention.

[0040] Reference numerals: 1-Outer casing, 2-Host computer display screen, 3-Power switch, 4-CV three-electrode system, 5-DPV three-electrode system, 6-Hardware circuit board, 7-Power supply, 8-Sample solution detection stage, 9-Support, 10-Beaker, 11-Terminal 1, 12-Negative voltage connection, 13-Positive voltage connection, 14-Main control board power supply line, 15-GND ground, 16-Electrode fixing disk 1, 17-RE reference electrode 1, 18-WE working electrode 1, 19-CE counter electrode 1, 20-Main control unit, 21-Signal conditioning unit, 22-Potential rise unit, 23-Filtering unit, 24-Micro current detection unit, 25-Potentialometer unit 1, 26-Potentialometer unit 2, 27-Receiver power supply terminal, 28-RE reference electrode 2, 29-WE working electrode 2, 30-CE counter electrode 2, 31-Electrode fixing disk 2, 32-Terminal 2. Detailed Implementation

[0041] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments, but the implementation of the present invention is not limited thereto.

[0042] Example 1: This embodiment, in conjunction with the accompanying drawings, details the specific hardware structure and parameter configuration of the rapid detection device for the numbing sensation of Sichuan pepper described in this invention.

[0043] like Figures 1 to 6 As shown, the outer casing 1 of the portable rapid testing device is made of ABS engineering plastic using injection molding. Its dimensions are 280mm long, 200mm wide, and 150mm high. Internally, the hardware circuit board 6 and power supply 7 are rationally arranged and fixed using molded reinforcing ribs. A 7-inch TFT-LCD capacitive touchscreen is embedded in the front of the outer casing 1 as the host computer display screen 2, with a resolution of 800*480, used to display testing parameters, volt-ampere curves, and numbing odor intensity results in real time. The power switch 3 is located on the right side of the display screen and uses a self-locking button.

[0044] The core of the device is the main control unit 20, which has a built-in numbing sensation intensity prediction model. It can process and analyze the received signals and finally obtain the quantitative detection result of the numbing sensation of Sichuan peppercorns. The main control unit 20 uses STMicroelectronics' STM32F103ZET6 as the microcontroller. This chip is based on the ARM-Cortex-M3 core, with a main frequency of 72MHz, and has sufficient computing power and I / O resources to coordinate system operation. The power supply 7 adopts a wide voltage input (AC100-240V) switching power supply, which internally converts the output to a stable DC voltage of ±5V and 3.3V. One path supplies power to the STM32F103ZET6 chip and its peripheral circuits through the main control board power supply line 14, and the other path supplies power to the analog circuit part on the hardware circuit board 6 through the positive voltage connection 13, the negative voltage connection 12 and the ground line GND15.

[0045] In the three-electrode detection cell module, both the CV three-electrode system 4 and the DPV three-electrode system 5 employ RE reference electrode 17 and RE reference electrode 28, WE working electrode 18 (3mm in diameter) and WE working electrode 29 (3mm in diameter), and CE counter electrode 19 and CE counter electrode 20. The electrodes are precisely fixed by electrode fixing disk 16 and electrode fixing disk 21 to ensure stable and repeated immersion in the sample solution during detection.

[0046] The six core modules of the hardware circuit board are integrated on a single PCB. The micro-current detection unit 24 is crucial for signal acquisition; it employs a multi-stage feedback resistor network based on the ADG408 analog switch chip (see...). Figure 10 The main control chip's GPIO pins control the selection of feedback resistors with different resistance values ​​(their correspondence is shown in Table 2), converting a wide range of response currents from nA to mA into voltage signals. These signals are then amplified by a two-stage amplifier circuit composed of a high-precision operational amplifier OPA2277PA (see Table 2). Figure 7 9) Perform signal conditioning to ultimately ensure high signal-to-noise ratio and accuracy throughout the entire measurement range. Potentiostat Unit 1 (25) and Potentiostat Unit 2 (26) are constructed using the operational amplifier OPA2228UA as the core circuit (see...). Figure 8 This ensures that the potential applied to the electrolytic cell strictly follows the waveform settings of CV or DPV, with a potential control accuracy better than ±1mV. The overall signal processing flow of the hardware circuit board 6 is as follows: Figure 12 As shown.

[0047] Through the above-described dedicated hardware design and system integration, this device achieves the system performance indicators shown in the table below: The overall system design specifications for this device are as follows:

[0048]

[0049] Table 2 Sensitivity levels, feedback resistors, and corresponding current ranges

[0050] Example 2: This embodiment details the application process of the device of the present invention in the detection of the numbing intensity of Sichuan pepper, and fully elaborates on the construction process of the core CDPV-CNN numbing intensity prediction model.

[0051] Accurately weigh 5.00 g of Sichuan pepper sample (accurate to 0.01 g), grind it using a grinder, and pass it through a 20-mesh sieve. Place the Sichuan pepper powder in a 150 mL brown stoppered conical flask, add 75 mL of ethanol, shake well, and then extract ultrasonically at 40 kHz for 20 min at 20℃. Transfer the mixture to a centrifuge tube and centrifuge at 2000 r / min for 5 min. Collect the supernatant in a 200 mL brown volumetric flask. Wash the precipitate with a small amount of ethanol, vortex for 1 min, and centrifuge again at 2000 r / min for 5 min. Combine the supernatants. Repeat this process once. Finally, bring the volume to 200 mL with ethanol to prepare the Sichuan pepper extract, which is stored at -18℃ for later use.

[0052] The reagents and instruments used in the experiment included: 95% food-grade ethanol, ultrapure water, 0.1 M PBS phosphate buffer (pH=7.4), KQ-500DE CNC ultrasonic cleaner, TGL-16MS benchtop high-speed centrifuge, DFY-500C grinder, and the rapid detection device for numbing flavor intensity involved in this invention.

[0053] The information on the pepper samples used in the experiment is shown in Table 3 below, covering the main producing areas and varieties.

[0054] Table 3. Information on varieties and origins of Sichuan pepper.

[0055] Take 2 mL of Sichuan pepper extract and mix it with 18 mL of 0.1 M PBS phosphate buffer to prepare the test sample, and place it in a 50 mL beaker 10. Before the test, the glassy carbon WE working electrode 18 and WE working electrode 29 were pretreated: they were polished on chamois leather with 0.3 μm and 0.05 μm alumina powder respectively, then rinsed thoroughly with ultrapure water, and finally ultrasonically cleaned for 1 min each in anhydrous ethanol and ultrapure water to thoroughly remove surface adsorbates and oxide layers.

[0056] The pretreated CV and DPV three-electrode system is correctly installed using electrode fixing disk 16 and electrode fixing disk 2, and stably immersed below the surface of the sample liquid. The device power switch 3 is turned on, and the host computer display screen 2 is activated. Serial communication parameters are set (baud rate 9600 bps), the "CV+DPV synchronous scanning" detection mode is selected, and the potential scanning range is set to -0.2 V to +1.2 V. The host computer software operation and data display process is as follows: Figure 11 As shown. Clicking the acquisition button will automatically complete waveform application, current signal acquisition, and data transmission. After acquisition, the data will be displayed on the host computer interface (its UI layout can be found in [reference]). Figure 13 The results show the intensity of the numbing sensation.

[0057] To objectively assess the accuracy of the device's predictive results, 15 rigorously trained sensory evaluators (with over 200 hours of experience) were organized. The sensory analysis required the following sample preparation process: First, 2 g of Sichuan pepper essential oil was dissolved in 20 mL of ethanol solution. The mixture was then heated in a 60°C constant-temperature water bath for 10 minutes. Subsequently, 20 mL of 70°C distilled water was slowly added. After the mixture cooled naturally to room temperature, it was brought to a final volume using ultrapure water to form a 1 L standard stock solution. Then, using drinking water as the dilution medium, the standard stock solution was serially diluted using the General Labeled Magnitude Scale (gLMS) method for sensory analysis. The specific parameters were: 1.38 (no perceptible threshold), 5.75 (weak), 16.22 (medium), 33.11 (strong), 50.12 (very strong), and 95.5 (the strongest perceptible sensation). After multiple rounds of taste testing by the sensory evaluators, several gradient concentrations covering different numbing sensation intensity ranges were selected as reference samples. Sensory evaluation was conducted on the numbing intensity of the 10 pepper samples mentioned above. Each sample underwent three rounds of randomized, balanced evaluation, and the final numbing intensity value was the average of the three rounds' scores. Based on the sensory evaluation results, the prediction accuracy of this device for the 10 pepper samples was calculated, and the results are shown in Table 4 below. The average prediction accuracy for all samples exceeded 90%, verifying the effectiveness and reliability of the device and method of this invention in practical applications.

[0058] Table 4. Accuracy of different varieties of Sichuan pepper in the numbing sensation intensity detection system

[0059] The circuit board module processes the CV-DPV volt-ampere curve data collected in the above steps in real time, and performs intelligent analysis using a pre-set CDPV-CNN numbing flavor intensity prediction model. This device invented and uses the CDPV-CNN (Collaborative Dual-Pulse Voltammetry Convolutional Neural Network) fusion algorithm as the core modeling technology. The construction of the CDPV-CNN numbing flavor intensity prediction model of the present invention includes the following steps: S1. Data Preparation and Preprocessing: In this embodiment, simultaneous CV and DPV scanning data were collected from 22 core Sichuan pepper producing areas across the country (covering both green and red Sichuan pepper), totaling 1584 sets. Each set of data includes CV voltammetry curves (potential range -0.2V to +1.2V, 512 sampling points) and DPV voltammetry curves (potential range -0.2V to +1.2V, 512 sampling points), and corresponds to the sensory evaluation benchmark value of numbing flavor intensity (using the gLMS method). The raw data underwent preprocessing: first, a Savitzky-Golay filter (window size 9, polynomial order 3) was applied to eliminate high-frequency noise; second, wavelet packet transform (wavelet basis 'db4', decomposition level 5) was used for signal denoising; finally, standardization was performed using the formula... Feature normalization is achieved, where μ is the mean and σ is the standard deviation; S2, Dataset Partitioning: The preprocessed data is divided into training and testing sets proportionally. In this specific implementation, the training set contains 1108 data sets (70%), and the testing set contains 476 data sets (30%) to ensure the statistical reliability of model training and evaluation.

[0060] S3, Model Architecture Design: The CDPV-CNN model uses a two-branch one-dimensional convolutional structure to process CV and DPV data separately: CV branch: Input the CV voltammetry curve and extract redox kinetic features, including electron transfer number (α), peak potential difference (ΔEp), and integrated charge (Q) through a one-dimensional convolutional layer (convolutional kernel size of 5 and stride of 1). DPV branch: Input DPV current-voltage curve, extract peak features through parallel one-dimensional convolutional layer (convolutional kernel size of 5, stride of 1), including peak current (ip), half width at half maximum (FWHM) and relaxation time (τ). Feature fusion layer: Introduces a cross-modal attention mechanism to adaptively weight and fuse CV and DPV features; the attention weight calculation formula is as follows: Among them, wi This represents the cross-modal attention weight of the i-th element; f cv , i and f dpv , i Let represent the value of the i-th element in the CV feature vector and the DPV feature vector, respectively; j is the summation traversal index. f cv , j and f dpv , j Let represent the value of the j-th element in the CV and DPV feature vectors, respectively; Through the above feature fusion layer operation, the synergistic response behavior of hydroxy-α-sanshool and hydroxy-β-sanshool was enhanced, resulting in a 32-dimensional weighted fusion feature mapping vector. Regression Output Layer: The weighted fused feature vectors described above are mapped to quantitative values ​​of numbing sensation intensity (range 0-100%) through a fully connected network (containing 64 nodes), and the level evaluation results are output; the Huber loss function is used to optimize the robustness of the model, and its formula is: L δ This represents the Huber loss function value. y This represents the true intensity of the numbing sensation. y ^ represents the predicted value. δ Set as the threshold parameter (set to 1.0); This step completes the architecture design of the CDPV-CNN model.

[0061] S4. Model Training and Optimization: Using the training set data partitioned in S2 of this embodiment, the CDPV-CNN model designed in S3 of this embodiment is trained. The training parameters include: using the Adam optimizer with an initial learning rate of 0.001; a batch size of 32; dynamic control of the training cycle through an early stopping strategy; and the introduction of Dropout (ratio of 0.2) and L2 regularization (coefficient of 0.01) to suppress overfitting. During training, the loss function and validation set performance are monitored to ensure model convergence. Through this training step, a preliminary CDPV-CNN model for predicting numbing sensation intensity is obtained.

[0062] S5. Model Validation and Performance Evaluation: Using the test set obtained in S2 of this embodiment, the performance of the preliminary CDPV-CNN numbing flavor intensity prediction model trained in S4 of this embodiment is evaluated. The coefficient of determination (R²) is calculated. 2 The model's R² and root mean square error (RMSE) were measured. The model was validated on the test set.2 The accuracy reached 0.962, and the RMSE was 2.15%, indicating that the model has high accuracy and generalization ability. Therefore, this model was determined to be the final CDPV-CNN numbing flavor intensity prediction model, which is suitable for rapid quantitative detection of Sichuan pepper numbing flavor intensity.

[0063] Through the above steps, a dedicated CDPV-CNN numbing flavor intensity prediction model was constructed and integrated into the device of this invention to realize the quantitative value and grade evaluation of the numbing flavor intensity of Sichuan pepper.

[0064] Finally, the intensity and grade evaluation results of the Sichuan pepper numbing flavor are displayed on the upper computer display screen 2, completing the entire process of rapid detection of the intensity of the Sichuan pepper numbing flavor.

[0065] The embodiments described above are merely some embodiments of the present invention. The accompanying drawings show preferred embodiments of the present invention, but these do not limit the scope of this application. Their purpose is to enable those skilled in the art to gain a more thorough and comprehensive understanding of the disclosure of this application. All equivalent embodiments based on the content of this application's specification and drawings, whether directly or indirectly applied to other related technical fields, are within the scope of patent protection of this application.

Claims

1. A rapid detection device for the intensity of Sichuan pepper numbing flavor, characterized in that, The rapid detection device for the intensity of Sichuan pepper numbing sensation includes: The outer casing, host computer interaction module, circuit board module, and three-electrode detection cell module; The host computer interaction module serves as the core of the device's human-computer interaction, enabling users to interact with the detection device through commands and data. The host computer interaction module is used to receive and set electrochemical detection parameters, send control commands to the circuit board module to start or stop the detection, dynamically display the volt-ampere curve generated during the detection process in real time, and finally output and display the quantitative value and grade evaluation result of the intensity of the Sichuan pepper numbing flavor. Circuit board module: As the control and signal processing center of the device, it coordinates the execution of the entire detection process and data processing. The circuit board module can generate an electrochemical excitation signal according to the instructions issued by the host computer interaction module and apply it to the three-electrode detection cell module. Simultaneously, it collects the response current signal generated by the three-electrode detection cell module under the action of the electrochemical excitation signal, and conditions, amplifies and converts the signal into digital form to obtain a pre-processed electrochemical signal. It further performs intelligent analysis and feature extraction on the pre-processed electrochemical signal, and finally obtains the quantitative detection result of Sichuan pepper numbing flavor through the CDPV-CNN numbing flavor intensity prediction model, and sends the result back to the host computer interaction module. Three-electrode detection cell module: As the sensing and signal generation unit of the device, it is used to construct a stable electrochemical detection environment and sense the electrochemical characteristics of the numbing and spicy substances in Sichuan pepper. By immersing the three-electrode system in the Sichuan pepper sample solution, under the excitation signal applied by the circuit board module, a specific electrochemical reaction for the numbing and spicy substances occurs at the working electrode interface, thereby generating a response current signal that is closely related to the type and concentration of the numbing and spicy substances.

2. The rapid detection device for the numbing sensation of Sichuan pepper according to claim 1, characterized in that, The outer casing includes an outer casing (1) and a power switch (3); the outer casing (1) provides structural support and physical protection for the entire device; the power switch (3) is located on the surface of the outer casing (1) and is used to control the device to be powered on and off. Upper computer interaction module: including upper computer display screen (2); the upper computer display screen (2) is embedded in the surface of the outer shell (1) and is used for parameter setting, control command issuance, real-time display of detection curve and quantitative results of numbing flavor intensity, and is displayed through the upper computer display screen (2); The circuit board module is a printed circuit board integrated inside the outer casing (1), including a main control unit (20), a potentiostat unit one (25), a potentiostat unit two (26), a signal conditioning unit (21), a microcurrent detection unit (24), and a power supply (7); wherein, the main control unit (20) is used to generate control commands to drive the potentiostat unit one (25) and the potentiostat unit two (26) to generate specific electrochemical excitation signals and apply them to the three-electrode detection cell module; the microcurrent detection unit (24) and the signal conditioning unit (21) are used to collect the response current signal generated in the three-electrode detection cell module and transmit it back to the main control unit (20); the main control unit (20) can process and analyze the received signals and finally obtain the quantitative detection result of the Sichuan pepper numbing flavor; The three-electrode detection cell module includes a support (9) and a three-electrode system; the support (9) is used to fix the three-electrode system to ensure that it is stably immersed in the solution to be tested; the three-electrode system is used to collect the electrochemical signal corresponding to the numbing flavor of Sichuan pepper.

3. The rapid detection device for the numbing sensation of Sichuan pepper according to claim 2, characterized in that, The three-electrode system is provided in two groups, namely the CV three-electrode system (4) and the DPV three-electrode system (5); the CV three-electrode system (4) includes RE reference electrode one (17), CE counter electrode one (19) and WE working electrode one (18); the DPV three-electrode system (5) includes RE reference electrode two (28), CE counter electrode two (30) and WE working electrode two (29).

4. The rapid detection device for the numbing sensation of Sichuan peppercorns according to claim 2, characterized in that, The WE working electrode one (18) and WE working electrode two (29) are one of glassy carbon electrode, gold electrode or platinum electrode.

5. The rapid detection device for the numbing sensation of Sichuan pepper according to claim 2, characterized in that, The bracket (9) includes a support column and an electrode fixing disk one (16) and an electrode fixing disk two (31); the electrode fixing disk one (16) and the electrode fixing disk two (31) have precise positioning holes for fixing the three-electrode system.

6. The rapid detection device for the numbing sensation of Sichuan pepper according to claim 1, characterized in that, The host computer display screen (2) is a touch screen, which integrates parameter configuration, detection start, data storage and historical query functions; the microcontroller used by the main control unit (20) is an STM32 series.

7. The operating method of the rapid detection device for the numbing sensation of Sichuan pepper according to any one of claims 1-6, characterized in that, The steps are as follows: S1 Sample Preparation: A quantitative sample of Sichuan pepper to be tested was crushed, sieved, and then extracted with ethanol solution by ultrasonication. After centrifugation, the Sichuan pepper extract was prepared. The Sichuan pepper extract was mixed with phosphate buffer solution in a certain proportion to prepare the test solution. S2 Electrode Installation and Pretreatment: The glassy carbon working electrode is subjected to surface grinding, cleaning and ultrasonic pretreatment to remove the oxide layer on the electrode surface; the two sets of RE reference electrode 1 (17), WE working electrode 1 (18), CE counter electrode 1 (19) and RE reference electrode 2 (28), WE working electrode 2 (29), CE counter electrode 2 (30) used for cyclic voltammetry and differential pulse voltammetry are fixed by electrode fixing disk 1 (16) and electrode fixing disk 2 (31) respectively and immersed below the surface of the solution to be tested; S3 Signal Acquisition and Processing: CV+DPV synchronous scanning is started through the upper computer display screen (2). The main control unit (20) generates the corresponding excitation waveform and applies it to the three-electrode detection cell. The volt-ampere signal of the Sichuan pepper numbing substance at the working electrode interface is generated in the detection cell. The micro current detection unit (24) converts the response current in the three-electrode detection cell into a voltage signal. After filtering, amplification and analog-to-digital conversion, the signal is transmitted back to the main control unit (20) for signal processing and analysis. S4 Numbness Intensity Prediction and Display: The main control unit (20) calls the CDPV-CNN numbing intensity prediction model to perform intelligent analysis on the collected voltammetric curve data, outputs the quantitative value of numbing intensity and the grade evaluation result, and displays it on the upper computer display screen (2); the CDPV-CNN numbing intensity prediction model integrates the dynamic characteristics of the CV curve and the peak characteristics of the DPV curve, and after weighted fusion through the cross-modal attention mechanism, outputs the numbing intensity value through the regression network.

8. According to the operation method of claim 7, the volume ratio of Sichuan pepper extract to phosphate buffer in step S1 is 1:

9.

9. The operating method according to claim 7, characterized in that, The construction of the CDPV-CNN numbing flavor intensity prediction model includes the following steps: S4-1, Data Preparation and Preprocessing: Collect pepper samples, select a total of n groups of pepper samples, and perform simultaneous CV and DPV scanning data to obtain the corresponding n groups of CV data and n groups of DPV data; n is a positive integer; Each set of data includes CV voltammetry curves and DPV voltammetry curves, and corresponds to the baseline value of numbing flavor intensity calibrated by sensory evaluation; the raw CV data and DPV data are preprocessed to obtain preprocessed data; The preprocessing steps are as follows: First, a Savitzky-Golay filter is applied to eliminate high-frequency noise; second, wavelet packet transform is used for signal denoising; finally, normalization is performed using the formula... Feature normalization is achieved, where μ is the mean and σ is the standard deviation; S4-2, Dataset Partitioning: The preprocessed data from S4-1 is divided into training and testing sets proportionally to ensure the statistical reliability of model training and evaluation. S4-3, Model Architecture Design: The CDPV-CNN model for predicting the intensity of numbing sensation uses a two-branch one-dimensional convolutional structure to process CV and DPV data separately. CV branch: Input the CV voltammetric curve and extract redox kinetic features through a one-dimensional convolutional layer, including electron transfer number α, peak potential difference ΔEp and integrated charge Q; DPV branch: Input the DPV current-voltage curve and extract peak features through parallel one-dimensional convolutional layers, including peak current ip, full width at half maximum (FWHM), and relaxation time τ. Feature fusion layer: Introduces a cross-modal attention mechanism to adaptively weight and fuse CV and DPV features; the attention weight calculation formula is as follows: ； in f cv , i and f dpv , i Let i and ii represent the i-th elements of the CV and DPV feature vectors, respectively. Through the above feature fusion layer operation, the synergistic response behavior of hydroxy-α-sanshool and hydroxy-β-sanshool is enhanced, resulting in a weighted fusion feature mapping vector; Regression output layer: The weighted fused feature vector is mapped to the quantitative value of numbing sensation intensity through a fully connected network, and the level evaluation result is output; the Huber loss function is used to optimize the robustness of the model, and its formula is: ； in y This represents the true intensity of the numbing sensation. y ^ represents the predicted value. δ Set the threshold parameter to 1.0; This step completes the architecture design of the prediction model and yields the initial numbing flavor intensity prediction model. S4-4, Model Training and Optimization: Using the training set data obtained in S4-2, the initial numbing flavor intensity prediction model designed in S4-3 was trained; during the training process, the loss function and validation set performance were monitored to ensure model convergence; through training, a preliminary CDPV-CNN numbing flavor intensity prediction model was obtained. S4-5, Model Validation and Performance Evaluation: The performance of the preliminary CDPV-CNN numbing flavor intensity prediction model trained in S4-4 was evaluated using the test set obtained in S4-2; the coefficient of determination R was calculated. 2 and root mean square error, where R 2 A value of not less than 0.95 and an RMSE of not more than 2.5% indicate that the model has high accuracy and generalization ability, thus determining this model as the final prediction model for the intensity of numbing sensation.

10. The operating method according to claim 9, characterized in that, In step S4-2, the training set accounts for 65-75% of the total data, and the remaining data is the test set. The training parameters in S4-4 include: using the Adam optimizer with an initial learning rate of 0.001; a batch size of 32; dynamic control of the training cycle through an early stopping strategy; and the introduction of Dropout mechanism and L2 regularization to suppress overfitting.

Images