Tea withering detection method and system based on ultrasonic attenuation

Abstract

The invention relates to the field of automatic detection of tea processing, and provides a tea withering detection method and system based on ultrasonic attenuation, and the method comprises the steps: synchronously collecting echo signals, position parameters and environment temperature, generating a stacking thickness field matrix through thermodynamic sound velocity compensation and cubic spline interpolation, and outputting porosity correction according to a rheological power law function; a transmission acoustic signal is collected by a time division multiplexing mechanism, a transmission amplitude field and a propagation time field are constructed through Hilbert transform and cross-correlation time delay estimation, average sound velocity distribution is constructed, and a physiological attenuation coefficient and a damage index are generated in combination with porosity correction; a composite feature vector is constructed, and the local water content and the cell wall breakage rate are solved through a space-time gradient coupling network; and generating a global regulation and control instruction in combination with the spatial weighted average moisture content and the accumulation thickness field matrix, and generating a fixed-point targeted raking instruction in combination with the abnormal displacement coordinates. According to the method, acoustic inversion and multi-dimensional closed-loop correction are fused, and digital perception and intelligent targeted regulation and control of the tea leaf withering process are constructed.

Description

Technical Field

[0001] This invention relates to the field of automated detection in tea processing, and in particular to a method and system for detecting tea withering based on ultrasonic attenuation. Background Technology

[0002] With the rapid development of modern tea processing technology, tea withering is a crucial process that determines the quality of the finished tea. However, in the current tea processing industry, the judgment of the degree of withering still relies heavily on human sensory experience. This traditional method generally suffers from limitations such as strong subjectivity, slow response, and the ability to observe only surface characteristics, making it difficult to meet the demands of modern processing for precise perception of internal physicochemical properties. Therefore, how to achieve real-time, objective monitoring and automated closed-loop control of the entire tea withering process under the complex and ever-changing microclimate environment has become a core technical challenge that urgently needs to be solved in the process of promoting the digital and intelligent transformation of tea processing.

[0003] Chinese patent application CN119861132A discloses a method for identifying adulterated camellia oil based on ultrasonic detection and determination of characteristic indicator substances. The method includes the following steps: S1, collecting samples of pure camellia oil, pure oil from other vegetable oils, and adulterated camellia oil; S2, performing ultrasonic detection on the collected sample oils; S3, performing liquid chromatography-mass spectrometry analysis on the samples to determine the content of trace active ingredients in the sample oils; S4, establishing a PLS model between ultrasonic parameters and the proportion of adulterated oil using partial least squares method, and based on the response values ​​obtained from liquid chromatography-mass spectrometry analysis, performing principal component analysis and orthogonal partial least squares discriminant analysis to establish a camellia oil identification model and predict the types of adulterated oils.

[0004] However, current technology still faces many challenges. During the withering process of tea leaves, the local microenvironment within the accumulated layer is highly susceptible to uneven airflow, resulting in partial withering. At this time, although the surface leaves show signs of dehydration, the cell walls of the inner leaves remain intact and their moisture content is high. Traditional detection methods can only perceive the surface hardness through touch and cannot penetrate deep into the accumulated layer to capture these subtle physiological and biochemical differences. If this asynchronous withering process between the inner and outer layers is not identified in time, it is impossible to perform targeted turning or environmental compensation on the lagging areas. This will lead to uneven overall fermentation of the tea leaves, hindered aroma transformation during subsequent processing, bitter taste, and reduced quality and economic value of the finished tea. Summary of the Invention

[0005] To achieve the above objectives, this invention provides a method for detecting tea leaf withering based on ultrasonic attenuation, the specific technical solution of which is as follows:

[0006] The original echo signal, sensor array spatial position parameters and real-time ambient temperature are acquired synchronously to generate a discrete distance vector sequence. The discrete distance vector sequence is spatially integrated using a cubic spline interpolation operator to construct the packing thickness field matrix. Numerical calculations are then performed on the packing thickness field matrix based on the compressive rheological power law function to output the porosity correction factor.

[0007] The transmitted acoustic response signal is collected, and the transmission amplitude field and sound wave propagation time field are constructed using the Hilbert transform operator and cross-correlation time delay estimation logic. The average sound velocity distribution data field is constructed based on the sound wave propagation time field. The density decoupling operation is performed on the transmission amplitude field in combination with the porosity correction factor to generate the physiological attenuation coefficient and tissue damage index.

[0008] The physiological attenuation coefficient is assigned a value based on the frequency band identifier. If the frequency band identifier points to the first characteristic frequency pulse signal, the physiological attenuation coefficient is taken as the first characteristic frequency physiological attenuation coefficient. If the frequency band identifier points to the second characteristic frequency pulse signal, the physiological attenuation coefficient is taken as the second characteristic frequency physiological attenuation coefficient. The first characteristic frequency pulse signal and the second characteristic frequency pulse signal are alternately emitted by the side wall ultrasonic transmitting probe assembly.

[0009] A composite feature vector is constructed based on the average sound velocity distribution data field, the physiological attenuation coefficient of the first characteristic frequency, the tissue structure damage index and the environmental state parameter vector. A spatiotemporal gradient coupled feature extraction network is used to perform sliding convolution operation and nonlinear regression mapping on the composite feature vector to invert and solve the local water content distribution map and cell wall breakage rate distribution map.

[0010] Based on the local moisture content distribution map, cell wall breakage rate distribution map, and the calculated spatial weighted average moisture content and accumulation thickness field matrix, a global environmental control instruction set is generated according to the multivariate feedback compensation control logic mapping. Combined with the identified abnormal longitudinal displacement coordinate set, the target action parameter vector is calculated to generate a fixed-point targeted material rake instruction.

[0011] Furthermore, the method for outputting the porosity correction factor includes:

[0012] The original echo signal reflected from the surface of the withered tea stack layer, the spatial position parameters of the sensor array, and the real-time ambient temperature in the withering trough were collected simultaneously. Time-domain waveform analysis was performed on the original echo signal to determine the round-trip time difference of the sound wave. A thermodynamic sound velocity compensation mechanism was constructed based on the real-time ambient temperature to obtain the thermodynamically corrected air sound velocity. The vertical physical distance from each discrete spatial sampling point to the surface of the withered tea stack layer was calculated by combining the round-trip time difference of the sound wave. A discrete distance vector sequence was constructed based on the topological order of the spatial position parameters of the sensor array.

[0013] The preset sensor installation absolute height constant is used to perform differential operation on the discrete distance vector sequence to generate discrete medium thickness values. A cubic spline interpolation operator is introduced to construct a tridiagonal matrix equation system. The tridiagonal matrix equation system is solved using the chasing method to generate a global continuous smooth curve function. The global continuous smooth curve function is discretized and resampled to obtain the blind zone reconstructed thickness value. The discrete medium thickness values ​​are combined with the spatial sequence integration to construct the stacking thickness field matrix.

[0014] The geometric compressibility ratio of the medium is determined based on the packing thickness field matrix and the preset standard reference loose thickness. Numerical calculations are then performed on the geometric compressibility ratio of the medium according to the power law function of the compressibility rheological properties to obtain the porosity correction factor.

[0015] Furthermore, the method for calculating the vertical physical distance from each discrete spatial sampling point to the surface of the tea withering accumulation layer includes:

[0016] The thermodynamically corrected air speed is obtained by multiplying the standard air speed constant by a thermodynamic correction factor obtained by taking the square root of the ratio of real-time ambient temperature to the zero point constant of the thermodynamic absolute temperature scale plus 1.

[0017] The round-trip time difference of the sound wave is obtained by calculating the difference between the absolute arrival time of the reflected echo envelope front and the absolute start time of the excitation pulse.

[0018] The vertical physical distance is calculated by multiplying the thermodynamically corrected air speed of sound and the round-trip time difference of the sound wave, and then multiplying the product by a coefficient of one-half.

[0019] The standard state air speed constant is the reference phase velocity of sound waves propagating in dry air at standard atmospheric pressure and a thermodynamic temperature of 0°C; the thermodynamic absolute temperature scale zero-point constant is the reference value for conversion between the Celsius and Kelvin thermodynamic temperature scales; the reflected echo envelope leading edge refers to the first rising edge of the reflected echo arriving at the gas-solid two-phase acoustic reflection interface; the excitation pulse absolute start time is the zero moment of electrical pulse excitation.

[0020] Furthermore, the method for constructing the average sound velocity distribution data field includes:

[0021] Based on the time-division multiplexing mechanism, the sidewall ultrasonic transmitting probe assembly alternately transmits the first characteristic frequency pulse signal and the second characteristic frequency pulse signal, and collects the transmitted acoustic response signal after penetrating the withered tea accumulation layer. The transmitted acoustic response signal is analyzed using the Hilbert transform operator to construct the first characteristic frequency transmission amplitude field and the second characteristic frequency transmission amplitude field. The acoustic transmission delay is quantized based on the cross-correlation time delay estimation logic to construct the sound wave propagation time field.

[0022] The preset system inherent delay constant is used to perform time difference calibration on the sound wave propagation time field to obtain the net flight time. The average sound velocity distribution data field is constructed based on the ratio of the withering groove width constant to the net flight time.

[0023] Furthermore, the method for generating the physiological attenuation coefficient and the tissue damage index includes: performing a logarithmic operation on the ratio of the preset transmitted signal reference amplitude and the transmitted amplitude field according to the Beer-Lamber law; calculating the total absolute attenuation by combining the withering groove width constant; introducing a porosity correction factor to perform a density decoupling operation on the total absolute attenuation to extract the physiological attenuation coefficient; and generating the tissue damage index based on the ratio of the physiological attenuation coefficients at different frequencies.

[0024] The transmission amplitude field is assigned a value based on the frequency band identifier. If the frequency band identifier points to the first characteristic frequency pulse signal, the transmission amplitude field is taken as the first characteristic frequency transmission amplitude field. If the frequency band identifier points to the second characteristic frequency pulse signal, the transmission amplitude field is taken as the second characteristic frequency transmission amplitude field.

[0025] Furthermore, the methods for calculating the local water content distribution map and the cell wall fragmentation rate distribution map include:

[0026] The average sound velocity distribution data field, the physiological attenuation coefficient of the first characteristic frequency, the tissue damage index and the environmental state parameter vector are obtained. The environmental state parameter vector is expanded in the whole domain using the dimensional broadcasting algorithm, combined with Z-Score normalization processing, and each feature channel is spliced ​​into a composite feature vector based on discrete spatial sampling points.

[0027] The composite feature vector is input into the spatiotemporal gradient coupling feature extraction network. A one-dimensional causal convolutional layer is used to perform sliding convolution operation along the vertical displacement coordinate to extract the vertical local gradient correlation features. The vertical local gradient correlation features are then subjected to nonlinear regression mapping through a fully connected layer to invert and solve the local water content distribution map and the cell wall breakage rate distribution map.

[0028] Furthermore, the method for generating the fixed-point targeted material scraping command includes:

[0029] A non-uniform flow field spatial weighting function is introduced to perform hydrodynamic correction on the local water content distribution map to calculate the spatial weighted average water content. The global water loss rate is obtained through the differential operation of the sampling period index, and a state deviation vector is constructed by combining the preset withering target evolution trajectory. Based on the multivariate feedback compensation control logic, the state deviation vector is mapped to a global environmental regulation instruction set.

[0030] The system iterates through local moisture content distribution maps, cell wall breakage rate distribution maps, spatially weighted average moisture content and accumulation thickness field matrices to identify abnormal longitudinal displacement coordinate sets that meet preset physiological deviation thresholds, geometric accumulation thresholds and biochemical statistical deviation thresholds. Combined with the fan speed parameter constraints of the global environmental control instruction set, the system calculates the target action parameter vector and generates a fixed-point targeted material scraping instruction.

[0031] Furthermore, the steps of the multivariable feedback compensation control logic include:

[0032] When the spatial weighted average moisture content is detected to exceed the target threshold set by the standard trajectory, and the absolute value of the global water loss rate is less than the preset minimum water loss slope, it is determined that the current environment has insufficient vapor pressure deficit. A ventilation acceleration command containing the fan speed parameter and an auxiliary dehumidification command containing the auxiliary heating parameter are generated, and the ventilation acceleration command and auxiliary dehumidification command are encapsulated into the global environment control command set.

[0033] When the spatially weighted average moisture content falls into the preset critical range of the wilting endpoint threshold, and the mean value of the cell wall breakage rate distribution map exceeds the preset moderate standard, a dynamic temperature and humidity balance instruction containing parameters for reducing wind speed and maintaining constant temperature is generated and encapsulated into the global environmental control instruction set.

[0034] Furthermore, the method for identifying the abnormal longitudinal displacement coordinate set includes:

[0035] Extract the real-time local moisture content element value at the longitudinal displacement coordinate corresponding to each discrete spatial sampling point in the local moisture content distribution map. If the real-time local moisture content element value exceeds the spatial weighted average moisture content and the numerical deviation between the two exceeds the physiological deviation threshold, determine that the local tea accumulation unit indexed by the longitudinal displacement coordinate is a moisture evaporation obstruction area, and mark the corresponding longitudinal displacement coordinate and classify it into the abnormal longitudinal displacement coordinate set.

[0036] Extract the real-time stacking thickness element value at the longitudinal displacement coordinate of each discrete spatial sampling point in the stacking thickness field matrix. If the real-time stacking thickness element value exceeds the geometric stacking threshold, determine that the local tea stacking unit indexed by the longitudinal displacement coordinate is the airflow penetration resistance zone, and mark the corresponding longitudinal displacement coordinate and classify it into the abnormal longitudinal displacement coordinate set.

[0037] Extract the real-time cell wall breakage rate element value at the longitudinal displacement coordinate corresponding to each discrete spatial sampling point in the cell wall breakage rate distribution map. If the real-time cell wall breakage rate element value is less than the global cell wall breakage rate statistical benchmark, determine that the local tea accumulation unit indexed by the longitudinal displacement coordinate is a mechanical intervention compensation area, and mark the corresponding longitudinal displacement coordinate and classify it into the abnormal longitudinal displacement coordinate set.

[0038] The calculation method for the global cell wall breakage rate statistical benchmark is as follows: perform spatial domain integration on the cell wall breakage rate distribution map to solve the mean value of the cell wall breakage rate of the entire tank, calculate the difference between the mean value of the cell wall breakage rate of the entire tank and the preset biochemical statistical deviation threshold, and determine the difference as the global cell wall breakage rate statistical benchmark.

[0039] The tea withering detection system based on ultrasonic attenuation is used to implement the above-mentioned tea withering detection method based on ultrasonic attenuation. The system includes a geometric field correction module, an acoustic attenuation feature extraction module, a parameter inversion module, and a closed-loop correction module.

[0040] The geometric field correction module is used to synchronously acquire the original echo signal, the spatial position parameters of the sensor array, and the real-time ambient temperature to generate a discrete distance vector sequence. It uses a cubic spline interpolation operator to integrate the discrete distance vector sequence into a spatial sequence to construct the packing thickness field matrix. It then performs numerical calculations on the packing thickness field matrix based on the power law function of compressibility rheology and outputs a porosity correction factor.

[0041] The acoustic attenuation feature extraction module is used to collect transmitted acoustic response signals, construct the transmission amplitude field and sound wave propagation time field using Hilbert transform operator and cross-correlation time delay estimation logic, construct the average sound velocity distribution data field based on the sound wave propagation time field, and perform density decoupling operation on the transmission amplitude field in combination with porosity correction factor to generate physiological attenuation coefficient and tissue damage index.

[0042] The parameter inversion module constructs a composite feature vector based on the average sound velocity distribution data field, the physiological attenuation coefficient of the first characteristic frequency, the tissue structure damage index, and the environmental state parameter vector. It then uses a spatiotemporal gradient coupling feature extraction network to perform sliding convolution and nonlinear regression mapping on the composite feature vector to invert and solve the local water content distribution map and the cell wall breakage rate distribution map.

[0043] The closed-loop correction module: based on the local moisture content distribution map, cell wall breakage rate distribution map, calculated spatial weighted average moisture content and accumulation thickness field matrix, generates a global environmental control instruction set according to the multivariate feedback compensation control logic mapping, and calculates the target action parameter vector by combining the identified abnormal longitudinal displacement coordinate set, and generates a fixed-point targeted material rake instruction.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0045] This invention introduces a thermodynamic sound velocity compensation mechanism to perform time-domain analysis and correction on the acquired raw echo signal, eliminating the geometric measurement deviation caused by the thermal drift of the local microclimate in the withering workshop on the air sound velocity scale. It also uses a cubic spline interpolation operator to construct a stacking thickness field matrix that includes the time dimension, reconstructing the global digital geometry of the tea withering stacking layer, effectively solving the detection blind zone problem caused by the spatial sampling discreteness of the sensor array.

[0046] This invention introduces a medium porosity correction factor to perform density decoupling calculation on the transmission amplitude field, and combines multi-frequency ultrasonic detection technology to construct a tissue structure damage index. This effectively removes non-physiological mechanical attenuation interference caused by increased physical packing density and pore closure, and realizes a deep quantitative inversion of the moisture migration characteristics and micro-skeleton structure damage state inside the tea withering and accumulation layer.

[0047] This invention uses a spatiotemporal gradient coupled feature extraction network to perform feature fusion and deep convolution operations on heterogeneous physical parameters including mechanical, acoustic and environmental dimensions, and establishes a nonlinear mapping relationship from physical measurement parameters to physiological and biochemical indicators. This effectively inverts the local moisture content distribution and cell wall breakage rate distribution inside the withered tea accumulation layer, solving the problem that traditional single-indicator detection cannot perceive the internal microscopic physiological state and structural evolution.

[0048] This invention constructs a two-layer asynchronous control mechanism that combines global environmental benchmark regulation with local anomaly targeted intervention by performing multi-dimensional state constraint analysis and anomaly threshold discrimination on real-time inverted local moisture content distribution map, cell wall breakage rate distribution map and accumulation thickness field matrix. This effectively eliminates local over-withering or under-withering phenomena caused by uneven distribution of environmental field in the withering tank, and realizes closed-loop correction of physical field and physiological field and quality homogenization throughout the tea withering process. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the principle of the tea withering detection method based on ultrasonic attenuation of the present invention.

[0051] Figure 2 This is a schematic diagram illustrating the ranging principle of sound waves on the surface of the withered tea leaf accumulation layer of the present invention.

[0052] Figure 3This is a schematic diagram illustrating the principle of the sidewall ultrasonic transceiver probe assembly of the present invention performing multi-band penetration scanning on the withered tea accumulation layer;

[0053] Figure 4 This is a functional block diagram of the tea withering detection system based on ultrasonic attenuation of the present invention.

[0054] Reference numerals: 21. Top scanning radar; 22. Discrete spatial sampling point; 23. Ultrasonic transceiver probe assembly; 24. Gas-solid two-phase acoustic reflection interface; 25. Withering groove; 31. Side wall ultrasonic transmitting probe assembly; 32. Side wall ultrasonic receiving probe assembly. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Example 1:

[0057] Please see Figure 1 As shown, this embodiment provides a method for detecting tea leaf withering based on ultrasonic attenuation, including:

[0058] Step S1000: Synchronously acquire the raw echo signal. Sensor array spatial position parameters and real-time ambient temperature To generate discrete distance vector sequences Using cubic spline interpolation operators to analyze discrete distance vector sequences Spatial sequence integration is performed to construct the stacking thickness field matrix. And based on the power-law function of compressible rheological properties, the accumulation thickness field matrix is... Perform numerical calculations and output the porosity correction factor. .

[0059] Specifically, this step aims to use top-scan radar to acquire the raw echo signal reflected from the surface of the withered tea leaf accumulation layer. Simultaneously, it collects the spatial position parameters of the sensor array that carries the movement of the top scanning radar. The geometric deformation of the tea leaf withering accumulation layer during water loss, shrinkage, and gravity settling was quantified by time-domain analysis, and an accumulation thickness field matrix containing the time dimension was constructed. And the porosity correction factor characterizing the compactness of tea leaf gaps. This eliminates the false attenuation interference caused by the dynamic changes in the thickness and density of the tea withering accumulation layer in subsequent steps, thereby achieving real-time compensation for the distortion of the physical field of the tea withering detection environment.

[0060] Further, step S1000 includes:

[0061] Step S1100: Synchronously acquire the raw echo signal reflected from the surface of the withered tea leaf accumulation layer. Sensor array spatial position parameters and the real-time ambient temperature inside the withering tank For the original echo signal Perform time-domain waveform analysis to determine the round-trip time difference of the sound wave, based on the real-time ambient temperature. A thermodynamic sound velocity compensation mechanism is constructed to obtain the thermodynamically corrected air sound velocity. The vertical physical distance from each discrete spatial sampling point to the surface of the withered tea leaf accumulation layer is calculated by combining the round-trip time difference of the sound wave. This is then applied based on the spatial position parameters of the sensor array. Constructing a discrete distance vector sequence from the topological order .

[0062] Specifically, this step aims to drive the top scanning radar and sensor array installed above the withering trough, driven by the central control unit, to perform spatial scanning, and to collect in real time the raw echo signals reflected from the surface of the tea withering accumulation layer. Simultaneously, it collects the spatial position parameters of the sensor array carrying the motion trajectory of the top scanning radar. And the real-time ambient temperature inside the withering tank as reported by the environmental monitoring unit. The data processing unit processes the original echo signal. Analyze the bidirectional propagation time characteristics of the ultrasonic pulse beam and reflected echo in the air medium, combined with the real-time ambient temperature. A thermodynamic sound velocity compensation mechanism for the air medium is constructed to eliminate the scaling error of sound wave propagation scale caused by thermal convection effect in the withering workshop. This is achieved by calculating the spatial position parameters of the sensor array. Given the determined vertical physical spacing between each discrete spatial sampling point, a discrete distance vector sequence is generated after temperature drift correction. .

[0063] In the specific implementation process, this step involves the central control unit sending a command to the top scanning radar installed above the withering trough, driving its ultrasonic transceiver probe assembly to perform periodic spatial scanning along the longitudinal axis of the withering trough, and simultaneously acquiring real-time coordinate data fed back by the servo encoder to generate spatial position parameters of the sensor array. Based on the spatial position parameters of the sensor array At each discrete spatial sampling point with a unique index, the transmitting unit in the ultrasonic transceiver probe assembly radiates an ultrasonic pulse beam vertically downwards. After passing through the air medium, the ultrasonic pulse beam is reflected at the gas-solid two-phase acoustic reflection interface formed between the air and the surface of the withered tea leaves. The resulting reflected sound pressure wave is then captured by the receiving unit in the ultrasonic transceiver probe assembly and converted into an electrical signal, generating the original echo signal. .

[0064] Further, please refer to Figure 2 As shown, Figure 2 This is a schematic diagram illustrating the ranging principle of sound waves on the surface of the withered tea leaf accumulation layer according to the present invention. Figure 2 As shown, the top scanning radar 21, installed above the withering trough 25, performs periodic spatial scanning operations along the longitudinal axis of the withering trough 25 under the drive of the central control unit. Figure 2 The horizontally extending black line located at the top and passing through the top scanning radar 21 represents the movement trajectory of the top scanning radar 21. Each specific position on this trajectory is determined by the spatial position parameters of the sensor array. A unique index is used to define discrete spatial sampling points 22. Each sampling point is determined by the spatial location parameters of the sensor array. At the determined discrete spatial sampling point 22, the ultrasonic transceiver probe assembly 23 mounted on the top scanning radar 21 radiates energy vertically downwards. Figure 2 The solid blue arrow represents the ultrasonic pulse beam radiated by the transmitting unit, which travels through the air medium and reaches the surface of the withered tea leaf accumulation layer below. Figure 2 The red dashed arrow represents the reflected sound pressure wave after reflection at the gas-solid two-phase acoustic reflection interface 24. This reflected sound pressure wave returns along the original path and is captured by the receiving unit, thus being converted into the original echo signal. .

[0065] The data processing unit processes the acquired raw echo signals. Performing time-domain waveform analysis to address the fluctuations in reflected wave amplitude on the surface of withered tea leaves caused by leaf curling, this step employs a constant-ratio timing discriminator (CFD) to track the original echo signal. The moment when the envelope amplitude reaches a fixed percentage of the peak value is used to pinpoint the absolute start time of the excitation pulse for the transmitting unit's operation from the noise background, i.e., the zero moment of electrical pulse excitation, and the absolute arrival time when the receiving unit captures the leading edge of the reflected echo envelope. The leading edge of the reflected echo envelope refers to the first rising edge of the reflected echo arriving after reflection from the gas-solid two-phase acoustic reflection interface. Based on the original echo signal... The data processing unit calculates the time span between the two absolute moments determined by the time-domain tracking, thereby quantifying the round-trip time difference of the ultrasonic pulse beam between the acoustic radiation end face of the transmitting unit and the surface of the withered tea accumulation layer.

[0066] Given that tea withering is essentially a thermodynamic process involving continuous temperature and humidity control, local microclimate fluctuations within the withering trough can cause changes in the thermophysical properties of the air, which acts as the medium for sound wave propagation. If a fixed sound velocity constant at standard atmospheric pressure is used in this calculation, the thermal drift of air density will lead to geometric measurement deviations at the millimeter or even centimeter level, thus masking the subtle volume shrinkage characteristics of tea leaves due to physiological water loss. Therefore, this step introduces a thermodynamic sound velocity compensation mechanism based on the ideal gas law. The data processing unit extracts the original echo signal from the environmental monitoring unit. Real-time ambient temperature fed back during the sampling period of the data acquisition action synchronization Using this real-time ambient temperature The phase velocity of the sound wave in the air medium is reconstructed to obtain the thermodynamically corrected air sound wave. Coupled with the round-trip time difference of the sound wave, the vertical physical distance from the sound wave radiation end face of the transmitting unit at the current discrete spatial sampling point to the surface of the withered tea accumulation layer is calculated.

[0067] The specific calculation logic for the vertical physical distance from the sound wave radiating end face of the transmitting unit to the surface of the tea withering accumulation layer at each discrete spatial sampling point is as follows: First, perform a sound velocity calibration calculation based on the thermophysical properties of the air medium, that is, multiply the standard state air sound velocity constant by the real-time ambient temperature. The dimensionless thermodynamic correction coefficient is obtained by taking the square root of the ratio of the absolute thermodynamic temperature scale zero-point constant to the absolute thermodynamic temperature scale zero-point constant, thereby obtaining the thermodynamically corrected air velocity under the current thermodynamic environment. Subsequently, the difference between the absolute arrival time of the reflected echo envelope front and the absolute start time of the excitation pulse is calculated to obtain the round-trip time difference of the sound wave, characterizing the bidirectional transmission of the ultrasonic pulse beam in physical space. Finally, the thermodynamically corrected air velocity and the round-trip time difference are multiplied, and the resulting product is multiplied by a coefficient of half to eliminate the multiplication effect of the bidirectional transmission path, thus calculating the vertical physical distance reflecting the physical settling characteristics of tea leaves during withering. The standard air velocity constant is the reference phase velocity of sound waves propagating in dry air at standard atmospheric pressure and a thermodynamic temperature of 0°C, and its value is fixed at 331.4 meters per second (m / s); the thermodynamic absolute temperature scale zero-point constant is the reference value for conversion between the Celsius temperature scale and the Kelvin thermodynamic temperature scale, and its value is fixed at 273.15 Kelvin (K); the round-trip time difference is the physical time delay required for the ultrasonic pulse beam to complete one bidirectional transmission between the sound wave radiation end face of the transmitting unit and the surface of the withered tea leaf accumulation layer.

[0068] This step iterates through all parameters determined by the spatial location of the sensor array. After indexing each discrete spatial sampling point and completing the above calculations, based on the spatial position parameters of the sensor array... The topological order is used to encapsulate all calculated vertical physical distances into a discrete distance vector sequence corrected for temperature drift. .in, This represents the sampling period index corresponding to each complete spatial scan performed by the top scanning radar, and its value increases monotonically as the withering process continues.

[0069] Step S1200: Call the preset sensor installation absolute height constant. For discrete distance vector sequences Differential operations are performed to generate discrete medium thickness values. A cubic spline interpolation operator is introduced to construct a tridiagonal matrix equation system. The tridiagonal matrix equation system is solved using the pursuit method to generate a globally continuous smooth curve function. The globally continuous smooth curve function is discretized and resampled to obtain the blind zone reconstructed thickness values. The discrete medium thickness values ​​are then combined with spatial sequence integration to construct the packing thickness field matrix. .

[0070] Specifically, this step aims to utilize the discrete distance vector sequence output in step S1100. and the system's preset sensor installation absolute height constant By employing coordinate mapping operations based on physical geometric constraints, the relative observation distance within the sensor's field of view is mapped to the accumulation thickness of the tea withering accumulation layer, thus resolving the detection blind zone problem caused by the discrete spatial distribution of the sensor array. Furthermore, by introducing a cubic spline interpolation operator from numerical analysis, a continuous accumulation thickness field matrix containing the time dimension is reconstructed. This process aims to establish a digital global geometry of the tea withering accumulation layer, eliminating the loss of geometric information caused by the interval between sampling points.

[0071] In the specific implementation process, this step sets the bearing bottom surface of the withering trough as the zero potential energy reference surface, and calls the fixed vertical distance between the acoustic radiation end face of the top scanning radar and the bearing bottom surface of the withering trough, i.e., the constant absolute height of the sensor installation. Install the sensor at an absolute height constant. and discrete distance vector sequence The vertical physical distance of each discrete spatial sampling point is differentially calculated to obtain the discrete medium thickness value of the tea accumulation at each discrete spatial sampling point, thus realizing the mapping of the data dimension from the vertical distance from the sensor to the surface of the withered tea accumulation layer to the physical thickness of the medium.

[0072] Since the discrete medium thickness values ​​are composed of only a finite number of discrete spatial sampling points, they cannot characterize the true geometric undulations of the tea withering accumulation layer surface within the sampling interval region, i.e., the detection blind zone not covered by the sensor array. Furthermore, the tea withering accumulation layer, as a viscoelastic porous medium subjected to gravitational settling, has its surface contour physically constrained by the nonlinear angle of repose characteristics of the bulk material, exhibiting continuity and second-order differentiable smoothness in its spatial distribution. Therefore, this step introduces a cubic spline interpolation operator to continuously reconstruct the discrete thickness data set composed of the discrete medium thickness values. A piecewise cubic polynomial is used to construct a numerical fitting curve between adjacent discrete spatial sampling points, and a boundary smoothing constraint is applied. This boundary smoothing constraint requires that the first derivative representing the tangent slope and the second derivative representing the curvature of the numerical fitting curve remain continuous at each discrete spatial sampling point that serves as an interpolation node. This eliminates non-differentiable geometric abrupt changes that may occur due to traditional linear interpolation algorithms, i.e., non-physically realistic inflection points on broken lines. Based on the aforementioned boundary smoothness constraint, this step uses a cubic spline interpolation operator to calculate the blind zone reconstruction thickness value within the physical sampling interval region of the sensor array. This blind zone reconstruction thickness value and the measured discrete medium thickness value are then mapped together into a spatiotemporal coordinate system, ultimately constructing a continuously distributed accumulation thickness field matrix along the longitudinal axis of the withering groove. .in, The longitudinal displacement coordinates on the longitudinal axis of the withering trough are based on the spatial position parameters of the sensor array. Determined physical location indices distributed along the length of the withering trough, i.e., the longitudinal displacement coordinates of each discrete spatial sampling point. .

[0073] The stacking thickness field matrix The specific construction logic is as follows: The data processing unit extracts the spatial position parameters contained in the sensor array. The longitudinal displacement coordinates of each discrete spatial sampling point in the data The two-dimensional data pairs, consisting of the corresponding discrete medium thickness values, are input into a cubic spline interpolation operator. To describe the piecewise cubic polynomial of each segmented interval's geometry, the cubic spline interpolation operator establishes a tridiagonal matrix equation system describing the global smoothness constraint based on the boundary smoothness constraint. This tridiagonal matrix equation system transforms the functional relationship between adjacent discrete spatial sampling points into a numerically solvable linear algebraic problem, ensuring that the entire tea withering accumulation layer surface satisfies the smooth characteristics of mechanical equilibrium at any local micro-element. Next, the chasing method in numerical analysis is used to solve the tridiagonal matrix equation system, calculating the interpolation polynomial coefficients required for the numerical fitting curve within each segmented interval. Finally, based on the interpolation polynomial coefficients, a globally continuous smooth curve function covering the longitudinal displacement coordinates of the entire withering trough is generated, and the globally continuous smooth curve function is discretized and resampled according to a preset spatial resolution. This resampling operation calculates the blind zone reconstructed thickness value within the physical sampling interval region of the sensor array, and integrates this blind zone reconstructed thickness value with the discrete medium thickness values ​​used as interpolation nodes into a spatial sequence, mapping them together to a unified spatiotemporal coordinate system to construct a continuously distributed accumulation thickness field matrix along the longitudinal axis of the withering groove. The preset spatial resolution is higher than the sensor's physical sampling density.

[0074] Step S1300, based on the stacking thickness field matrix and preset standard reference loose thickness The geometric compressibility ratio of the medium is determined, and numerical calculations are performed on the geometric compressibility ratio of the medium based on the power-law function of the compressibility rheological properties to calculate the porosity correction factor. .

[0075] Specifically, this step aims to utilize the stacking thickness field matrix constructed in step S1200. Based on the rheological compaction theory of porous media, this study addresses the self-weight consolidation effect of tea withering layers as viscoelastic granular materials under gravity, utilizing the aforementioned packing thickness field matrix. The evolution characteristics of the internal pore structure of the medium are deduced by the power-law function of the compressible rheological properties, and the porosity correction factor is calculated and output. This process aims to generate a dimensionless physical correction quantity that is strictly mapped to the spatial location and time dimension of the withering tank, thereby quantifying and removing the mechanical attenuation component caused by the increase in physical packing density, preventing the mechanical attenuation component from confusing and masking the physiological acoustic attenuation characteristics caused by cell dehydration, and realizing quantitative compensation for the acoustic transmission field distortion caused by the thickness change of the tea withering packing layer.

[0076] In the specific implementation process, the data processing unit reads the stacking thickness field matrix. According to the rheological compaction theory of porous media, the vertical stress at the bottom of the withered tea layer is positively correlated with the cumulative physical thickness of the overlying tea layer. An increase in vertical stress leads to compression of the free volume between the discrete tea units in the media framework, resulting in a decrease in media porosity. In ultrasonic transmission physics, a decrease in porosity directly alters the acoustic properties of the medium, causing an increase in the scattering cross-section of the ultrasonic pulse beam during penetration and exacerbating acoustic impedance mismatch, thus producing non-physiological mechanical attenuation.

[0077] To eliminate mechanical attenuation interference caused by the evolution of the geometric packing morphology of the tea leaves during withering at the signal processing level, this step constructs a nonlinear acoustic impedance correction mechanism based on the rheological compaction theory of porous media. This mechanism does not directly output the physical value of density, but instead outputs a dimensionless parameter, namely the porosity correction factor, used to correct the total ultrasonic attenuation coefficient through numerical calculation. The data processing unit processes the stacking thickness field matrix. The longitudinal displacement coordinates of each discrete spatial sampling point in the index and sampling period index The preset power-law function for compressible rheological properties is invoked to perform numerical calculations, and the porosity correction factor is calculated. .

[0078] The execution logic of the power-law function of the compressibility rheological property is as follows: Extract the current longitudinal displacement coordinate. The stacking thickness field matrix below loose thickness compared to the preset standard reference The process involves performing division to obtain the dimensionless geometric compressibility ratio of the medium; then, using a preset compaction index as a power, an exponential operation is performed on the geometric compressibility ratio to derive a nonlinear rheological characteristic value reflecting the degree of closure of the pore structure inside the medium; finally, the nonlinear rheological characteristic value is multiplied by a preset geometric attenuation reference coefficient to calculate the porosity correction factor characterizing the compactness of the medium. The standard reference loose thickness is mentioned above. The physical threshold constant is a preset value based on the initial bulk density characteristics of the specific tea variety to be tested. Its value is typically set as the critical height when the medium is in a naturally loosely packed state without significant rheological compaction. The compaction index is a dimensionless rheological parameter characterizing the sensitivity of tea, as a viscoelastic porous medium, to deformation of its internal pore structure under vertical stress. Its value range is typically between 1.2 and 1.5, calibrated through porous media mechanics experiments. The geometric attenuation reference coefficient is a system intrinsic constant calibrated through standard media transmission experiments, used to balance the conversion ratio between geometric and acoustic dimensions. The porosity correction factor... It is the longitudinal displacement coordinate of the discrete spatial sampling point on the longitudinal axis of the withering trough. and sampling period index Dimensionless physical scalars that are strictly mapped.

[0079] Step S2000: Acquire the transmitted acoustic response signal, and construct the transmission amplitude field using the Hilbert transform operator and cross-correlation time delay estimation logic. Harmony sound wave propagation time field Based on the sound wave propagation time field Constructing the average sound speed distribution data field Combined with porosity correction factor For transmission amplitude field Perform density decoupling operations to generate physiological attenuation coefficients. and tissue damage index .

[0080] Specifically, this step aims to utilize a preset first characteristic frequency pulse signal. Second characteristic frequency pulse signal and the porosity correction factor output in step S1300. Combined with the system's inherent withering groove width constant and the reference amplitude of the transmitted signal The transmitted acoustic response signal of the tea withering accumulation layer was obtained using sidewall ultrasonic transmission technology. Furthermore, the porosity correction factor was then... As a physical field decoupling operator, the transmitted acoustic response signal undergoes dual corrections at both geometric and physical levels, ultimately calculating and outputting the average sound velocity distribution data field. Tissue damage index and the physiological attenuation coefficient of the first characteristic frequency This process aims to eliminate non-physiological scattering interference on sound wave energy caused by the physical accumulation morphology of the tea leaves during withering, such as increased density and pore closure. It extracts acoustic fingerprints that purely reflect the physiological state of the tea leaves within the withering accumulation layer, as well as the microscopic fracture characteristics of the solid-phase skeleton structure of the tea leaves, which characterize the physical structure.

[0081] The first characteristic frequency pulse signal is set to a frequency of 50 kHz, which has a relatively long wavelength and interacts with the liquid phase moisture in the withered tea accumulation layer to characterize the change in liquid phase moisture content. The second characteristic frequency pulse signal is set to a frequency of 200 kHz, which has a relatively short wavelength and interacts with the solid phase skeleton structure of the tea in the withered tea accumulation layer, namely the cell wall and leaf vein microstructure, to characterize the integrity of the solid phase skeleton.

[0082] Further, step S2000 includes:

[0083] Step S2100: Based on the time-division multiplexing mechanism, control the sidewall ultrasonic transmitting probe assembly to alternately transmit the first characteristic frequency pulse signal. Second characteristic frequency pulse signal The transmitted acoustic response signal after penetrating the withered tea leaf accumulation layer was collected, and the transmitted acoustic response signal was analyzed using the Hilbert transform operator to construct the transmission amplitude field at the first characteristic frequency. Second characteristic frequency transmission amplitude field Based on cross-correlation delay estimation logic, acoustic transmission delay is quantized to construct the sound wave propagation time field. .

[0084] Specifically, this step aims to control the timing of the input sensor array based on the input sensor array timing commands. The drive sidewall ultrasonic transmitting probe assembly performs multi-band penetration scanning on the withered tea leaf accumulation layer, transmitting preset first characteristic frequency pulse signals into the withered tea leaf accumulation layer. Second characteristic frequency pulse signal This study utilizes the differential scattering cross-sectional characteristics generated by the acoustic energy interaction between liquid moisture and solid skeleton structure in the withered tea leaf accumulation layer using ultrasonic waves of different frequencies. The transmitted acoustic response signals after penetration are collected and processed to calculate and output the transmission amplitude field at the first characteristic frequency. Second characteristic frequency transmission amplitude field and the sound wave propagation time field This process provides orthogonalized data support for subsequent steps by simultaneously constructing the amplitude of the transmitted signal, which includes the characteristics of transmitted energy attenuation, and the sound wave propagation time, which includes the characteristics of acoustic transmission velocity. This solves the technical problem that single-frequency detection cannot distinguish between moisture attenuation and structural scattering attenuation.

[0085] In the specific implementation process, the central control unit follows the timing control instructions of the sensor array. A drive signal is sent to the sidewall ultrasonic transmitting probe assembly arranged on the sidewall of the withering trough. To avoid aliasing of acoustic signals of different frequency bands in the time domain and to eliminate standing wave interference, this step adopts a time-division multiplexing mechanism to control the sidewall ultrasonic transmitting probe assembly to alternately transmit the first characteristic frequency pulse signal at millisecond intervals. Second characteristic frequency pulse signal These correspond to the first characteristic frequency pulse signal, respectively. Second characteristic frequency pulse signal The ultrasonic pulse beams all penetrate the withered tea leaves horizontally in the horizontal direction, and the ultrasonic receiving probe assembly on the opposite side wall synchronously collects the transmitted acoustic response signal after penetrating the withered tea leaves.

[0086] Further, please refer to Figure 3 As shown, Figure 3 This is a schematic diagram illustrating the principle of the sidewall ultrasonic transceiver probe assembly of the present invention performing multi-band penetration scanning on the withered tea leaf accumulation layer. Figure 3 As shown, in terms of hardware layout, the two side walls of the withering trough 25 are respectively provided with side wall ultrasonic transmitting probe assembly 31 and side wall ultrasonic receiving probe assembly 32. Figure 3 The image shows the distribution of two alternately emitted ultrasonic pulse beams in the horizontal space based on a time-division multiplexing mechanism. Figure 3 The waveform curves of different colors and frequencies represent the detection mechanisms for different physical properties of tea. Specifically, Figure 3 The blue curve represents the pulse signal with the first characteristic frequency. Its set frequency is 50kHz. This frequency band has a relatively long wavelength and mainly interacts with the liquid moisture in the withering and accumulation layer of tea leaves. The degree of signal attenuation after transmission is used to characterize the change in liquid moisture content inside the tea leaves. Figure 3 The red curve represents the second characteristic frequency pulse signal. Its set frequency is 200kHz. This frequency band has a relatively short wavelength and mainly interacts with the solid skeleton structure of tea leaves in the withered tea accumulation layer through sound energy scattering. Its transmission response is used to characterize the integrity of the solid skeleton.

[0087] The data processing unit performs a Hilbert transform operation on the transmitted acoustic response signal acquired by the sidewall ultrasonic receiving probe assembly to extract the instantaneous envelope signal, thereby eliminating the interference of carrier phase fluctuations caused by multipath effects during the transmission of ultrasonic pulse beams within the tea withering and stacking layer on energy measurement. This step extracts the pulse signal corresponding to the first characteristic frequency. Second characteristic frequency pulse signal The instantaneous envelope peak value of the transmitted acoustic response signal at the sidewall ultrasonic receiving probe assembly is obtained, and the instantaneous envelope peak value is mapped to the longitudinal displacement coordinate. and sampling period index In the spatiotemporal coordinate system, construct the transmission amplitude field of the first characteristic frequency. Second characteristic frequency transmission amplitude field .

[0088] The first characteristic frequency transmission amplitude field Second characteristic frequency transmission amplitude field The specific construction logic is as follows: The data processing unit identifies the frequency band it is currently processing. Frequency band identifier This is a variable index used to distinguish different frequencies. Subsequently, the transmitted acoustic response signal of the corresponding frequency band output by the sidewall ultrasonic receiving probe assembly is extracted, and a Hilbert transform operator is performed on it to obtain an analytic signal. Then, the complex modulus of this analytic signal is calculated to obtain the instantaneous envelope signal. Within a preset complete time window of a single ultrasonic transmission sampling, the global maximum value of the instantaneous envelope signal is searched, and this global maximum value is determined as the index within the sampling period. Longitudinal displacement coordinates Transmission amplitude field at the location When the frequency band is identified Pointing to the first characteristic frequency pulse signal At that time, the transmitted amplitude field Assigned the value as the transmission amplitude field of the first characteristic frequency When frequency band identifier Pointing to the second characteristic frequency pulse signal At that time, the transmitted amplitude field Assigned the value as the second characteristic frequency transmission amplitude field .

[0089] To capture the acoustic transmission delay required for an ultrasonic pulse beam to penetrate the withered tea layer in the industrial environment of tea withering, which is subject to mechanical vibration and electromagnetic interference, this step abandons the traditional first-wave threshold detection method, which is susceptible to amplitude fluctuations, and adopts a cross-correlation delay estimation logic with better noise resistance. The data processing unit extracts the reference pulse signal stored in the system memory that is used to drive the ultrasonic transmitting probe assembly on the side wall. , and the transmitted acoustic response signal acquired by the sidewall ultrasonic receiving probe assembly.

[0090] The specific execution of the cross-correlation delay estimation logic is as follows: First, the data unit will reference the pulse signal. The transmitted acoustic response signal and the transmitted acoustic response signal are placed in the same time-domain integration interval, and a sliding convolution integral operation is performed to construct a cross-correlation function that reflects the change in the similarity of the two signal waveforms with time displacement. Subsequently, within a preset physical propagation time window, a time lag variable that causes the cross-correlation function to reach its global maximum value is searched, and this time lag variable is determined as the sound wave propagation time field in the current spatiotemporal coordinates. The sound wave propagation time field This directly characterizes the absolute physical time delay required for an ultrasonic pulse beam to penetrate the withered tea leaf accumulation layer, providing a temporal characteristic basis for subsequent steps.

[0091] Step S2200: Call the preset system inherent delay constant to adjust the sound wave propagation time field. Perform time difference calibration calculations to obtain the net flight time, based on the withering trough width constant. The average sound speed distribution data field is constructed by the ratio of the net flight time to the average sound speed. .

[0092] Specifically, this step aims to utilize the sound wave propagation time field output in step S2100. Combined with the system's preset withering groove width constant Based on the system's inherent delay constant, a data field representing the average sound velocity distribution within the withered tea leaf accumulation layer is constructed using time difference calculations. Given the significant positive physical correlation between sound wave propagation velocity and its equivalent elastic modulus and density in porous viscoelastic media, and the fact that the physical hardness of tea leaves, mainly maintained by cell turgor pressure, gradually decreases during withering and water loss, leading to a dynamic decay of the media's equivalent elastic modulus, this step aims to indirectly quantify the macroscopic hardness state of tea leaves as they evolve during withering by calculating the sound velocity field. This addresses the misjudgment problem that can easily arise from using a single decay index under complex temperature and humidity conditions.

[0093] In the specific implementation process, the data processing unit first calls the withering slot width constant stored in the system memory. The width of the withering trough is constant. The linear geometric path length between the acoustic radiation surface of the sidewall ultrasonic transmitting probe assembly and the acoustic receiving surface of the sidewall ultrasonic receiving probe assembly is defined as the physical acoustic path reference for acoustic wave transmission.

[0094] Simultaneously, to eliminate non-medium transmission errors introduced by the hardware system, this step introduces a pre-calibrated system inherent delay constant. This system inherent delay constant encompasses the delay of the ultrasonic pulse beam during transmission through the transmitting loop cable and the digital-to-analog conversion delay, the acoustic transmission delay of the ultrasonic transducer's acoustic impedance matching layer, and the total fixed time lag generated by the induced response electrical signal during analog-to-digital conversion in the receiving loop, measured in seconds (s). The data processing unit traverses the input acoustic wave propagation time field. The net transit time of the ultrasonic pulse beam is obtained by subtracting the inherent delay constant of the system from the value of each discrete sound wave transmission time point in the index.

[0095] Finally, the width constant of the withering groove is... Divide by the net flight time to calculate the average sound speed distribution data field in the corresponding spatiotemporal coordinates. And map it to the longitudinal displacement coordinates. and sampling period index In the matrix space formed.

[0096] Step S2300: Apply the preset transmit signal reference amplitude according to Beer-Lambert law. and transmission amplitude field The ratio is logarithmically calculated, combined with the withering trough width constant. To calculate the total absolute attenuation, a porosity correction factor is introduced. Perform density decoupling operation on the total absolute attenuation to extract the physiological attenuation coefficient. Based on the physiological attenuation coefficient at different frequencies The ratio of the two values ​​is used to generate the tissue damage index. .

[0097] Specifically, this step aims to utilize the first characteristic frequency transmission amplitude field from step S2100. Second characteristic frequency transmission amplitude field Combined with the system's preset transmission signal reference amplitude and the constant width of the withering trough The total energy loss of the inverted sound wave along the transmission path is calculated; thus, a porosity correction factor from step S1300 is forcibly introduced. As a key quantitative decoupling parameter, the first characteristic frequency transmission amplitude field, which includes a mixture of physical scattering and physiological absorption effects, is used. Second characteristic frequency transmission amplitude field Normalization is performed. This process aims to eliminate non-physiological mechanical attenuation interference induced by the increased physical packing density and closure of internal pores in the tea leaf withering layer. It extracts acoustic fingerprint features that purely reflect the evolution of viscous damping determined by moisture content and acoustic scattering cross-sectional characteristics determined by the integrity of the leaf structure. The final output is the tissue structure damage index. This enables the accurate and reliable detection of the withering quality of tea leaves.

[0098] In the specific implementation process, the data processing unit quantifies the transmission loss of the ultrasonic pulse beam's sound pressure amplitude during the penetration of the tea leaf withering and accumulation layer, based on the acoustic simulation of Beer-Lambert Law (BLL). This step calls upon a preset reference amplitude for the transmitted signal. The incident energy reference is the standard output voltage of the sidewall ultrasonic transmitting probe assembly when unloaded. Simultaneously, the transmission amplitude field at the end of the lateral physical path between the sidewall ultrasonic transmitting probe assembly and the sidewall ultrasonic receiving probe assembly is extracted. Among them, when the frequency band identifier Pointing to the first characteristic frequency pulse signal At that time, the transmission amplitude field The value is taken as the transmission amplitude field of the first characteristic frequency. When frequency band identifier Pointing to the second characteristic frequency pulse signal At that time, the transmitted amplitude field Assigned the value as the second characteristic frequency transmission amplitude field .

[0099] The data processing unit calculates the reference amplitude of the transmitted signal. and transmission amplitude field The natural logarithm of the ratio is then divided by the constant width of the withering trough. This allows us to calculate the total absolute attenuation of ultrasonic waves penetrating the withered tea leaf accumulation layer at the current frequency band.

[0100] To eliminate the interference of the geometric packing morphology of the tea withering layer on the quantitative inversion of physiological acoustic characteristics, this step introduces a porosity correction factor. As a dimensionless denominator, density decoupling is performed on the total absolute attenuation mentioned above. This is due to the porosity correction factor. The additional mechanical attenuation factor caused by the excessively dense packing and reduced internal porosity of the tea withering layer due to its own weight consolidation effect was quantified. This additional mechanical attenuation factor can be offset at the signal processing level through division, thereby obtaining the corrected physiological attenuation coefficient after removing the interference of physical packing density. Based on frequency band identification The difference in frequency band identification Specifically, it is divided into the physiological attenuation coefficient of the first characteristic frequency. Second characteristic frequency physiological attenuation coefficient When the frequency band is identified Pointing to the first characteristic frequency pulse signal Physiological attenuation coefficient The value is taken as the physiological attenuation coefficient of the first characteristic frequency. When frequency band identifier Pointing to the second characteristic frequency pulse signal Physiological attenuation coefficient The value is taken as the physiological attenuation coefficient of the second characteristic frequency. Wherein, the first characteristic frequency physiological attenuation coefficient Used to characterize the viscous absorption attenuation caused by water in the liquid phase; the second characteristic frequency physiological attenuation coefficient Used to characterize scattering attenuation caused by the destruction of the solid-phase framework integrity.

[0101] Furthermore, to quantify the tissue mechanical structure degradation characteristics of tea leaf tissue during the evolution of physicochemical properties, specifically the transformation of cell walls from a state of tension and integrity supported by intracellular turgor pressure to a state of wrinkling and rupture due to dehydration, this step utilizes ultrasound of different frequencies to measure the differences in scattering sensitivity at mesoscopic scattering interfaces within the tea tissue, such as the cell walls and leaf vein network, and calculates the physiological attenuation coefficient of the second characteristic frequency. and the physiological attenuation coefficient of the first characteristic frequency The ratio of and generates a dimensionless tissue damage index. .

[0102] Step S3000, based on the average sound velocity distribution data field Physiological attenuation coefficient of the first characteristic frequency Tissue damage index and environmental state parameter vector To construct composite feature vectors Using a spatiotemporal gradient coupled feature extraction network to extract composite feature vectors Perform sliding convolution and nonlinear regression mapping to invert and solve the local moisture content distribution map. Distribution of cell wall fragmentation rate .

[0103] Specifically, this step aims to utilize the average sound velocity distribution data field, which characterizes macroscopic mechanical stiffness, output from step S2200. The physiological attenuation coefficient of the first characteristic frequency characterizing the water content of the liquid phase output in step S2300. And the tissue damage index, which characterizes the integrity of tissue structure. Combined with the environmental state parameter vector collected by the environmental monitoring unit Spatiotemporal alignment and data fusion are performed. This is achieved by constructing composite feature vectors that incorporate mechanical, acoustic, and environmental dimensions. The system utilizes a pre-trained spatiotemporal gradient coupled feature extraction network to perform nonlinear feature mapping, thereby calculating the local moisture content distribution map characterizing the physiological water loss of tea leaves. Distribution of cell wall fragmentation rate characterizing fermentation precursor conditions This process enables the quantitative inversion from physical acoustic measurement parameters to physiological and biochemical process indicators, providing a direct digital decision-making basis for the closed-loop control of tea withering process.

[0104] Further, step S3000 includes:

[0105] Step S3100: Obtain the average sound velocity distribution data field. Physiological attenuation coefficient of the first characteristic frequency Tissue damage index and environmental state parameter vector The dimensional broadcasting algorithm is used to analyze the environmental state parameter vector. Perform global spatial expansion, combine Z-Score standardization to eliminate dimensional differences, and concatenate each feature channel into a composite feature vector based on discrete spatial sampling points. .

[0106] Specifically, this step aims to resolve the average sound velocity distribution data field output by step S2200. The physiological attenuation coefficient of the first characteristic frequency output in step S2300 and tissue damage index and the environmental state parameter vector collected by the environmental monitoring unit This addresses the heterogeneity issue in terms of physical dimensions and data dimensions. By performing spatiotemporal alignment operations and numerical normalization, the mechanical characteristics representing the macroscopic hardness of the medium, i.e., the average sound velocity distribution data field, are obtained. Acoustic characteristics characterizing the evolution of the microstructure of a medium, namely the physiological attenuation coefficient at the first characteristic frequency. and tissue damage index And the thermodynamic boundary conditions characterizing external stress conditions, i.e., the environmental state parameter vector. By integrating these elements into a unified mathematical expression, a composite feature vector is constructed that includes information on the endogenous physiological state and external environment of the tea leaf withering accumulation layer. This eliminates the order-of-magnitude differences between different physical parameters.

[0107] In the specific implementation process, given the environmental state parameter vector collected by the environmental monitoring unit... Includes real-time ambient temperature and relative ambient humidity Typically, this is one-dimensional time-series data that varies over time; while the physical feature data matrix to be fused includes the average sound velocity distribution data field, which serves as mechanical feature data. Physiological attenuation coefficient, the first characteristic frequency of acoustic feature data and tissue damage index All are two-dimensional spatiotemporal matrices that vary with time and longitudinal displacement space. This step employs a dimensionality broadcasting algorithm to distribute the environmental state parameter vectors... Longitudinal displacement coordinates along the withering trough The direction is extended and mapped globally to ensure that its spatial dimension is consistent with the physical feature data matrix, thereby ensuring that there is a corresponding environmental state parameter vector associated with any discrete spatial sampling point, and establishing a spatial correspondence between the environmental field and the mechanical and acoustic fields.

[0108] In order to eliminate the average sound velocity distribution data field Physiological attenuation coefficient of the first characteristic frequency Tissue damage index and the environment state parameter vector after global expansion mapping To mitigate the negative impact of numerical differences caused by varying physical dimensions on the convergence speed and feature weight allocation bias in subsequent nonlinear mapping, this step performs Z-Score standardization on the four sets of heterogeneous physical parameters. This process maps the original discrete sampled values ​​of each physical quantity to a dimensionless distribution with a mean of 0 and a standard deviation of 1, thus achieving statistical normalization of the heterogeneous data.

[0109] After completing the spatiotemporal alignment and standardization processes, the data processing unit traverses the longitudinal displacement coordinates of each discrete spatial sampling point. and sampling period index Four sets of heterogeneous physical parameters, after standardization, are spliced ​​together as independent feature channels to construct a composite feature vector representing the holographic state of the tea leaf withering accumulation layer. .

[0110] Step S3200, combine the feature vectors The input spatiotemporal gradient coupled feature extraction network utilizes a one-dimensional causal convolutional layer to perform longitudinal displacement coordinates. Sliding convolution operations are performed to extract vertical local gradient correlation features. A fully connected layer is then used to perform nonlinear regression mapping on these features to inversely calculate the local moisture content distribution map. and cell wall fragmentation rate distribution map .

[0111] Specifically, this step aims to transform the composite feature vector derived from the output of step S3100. A pre-defined spatiotemporal gradient coupled feature extraction network is input, and a composite feature vector is established through tensor operations within the network. Local moisture content distribution map characterizing the degree of physiological water loss in tea leaves And a distribution of cell wall fragmentation rate characterizing fermentation precursor conditions. The quantitative mapping relationship between them enables parameter inversion from multidimensional physical field observations to physiological and biochemical indicators that cannot be directly observed, providing core decision data for the digital monitoring of tea processing.

[0112] The spatiotemporal gradient coupling feature extraction network is not a general function fitter, but an inverse inversion mechanism built to address the asynchronous nature of the sound wave propagation speed, which characterizes the macroscopic mechanical modulus of the medium, and the acoustic attenuation coefficient evolution rate, which characterizes the microstructure of the medium, during the tea withering process.

[0113] In the specific implementation process, this step will combine the feature vectors. The input is fed into a spatiotemporal gradient coupling feature extraction network. The specific computational logic of this network is as follows:

[0114] The first step is the extraction of local correlation features along the longitudinal gradient. Unlike conventional image recognition, this step utilizes the longitudinal displacement coordinates along the withering groove from a one-dimensional causal convolutional layer pre-installed in the spatiotemporal gradient coupling feature extraction network. For composite feature vectors Perform a sliding convolution operation. This operation aims to extract the features contained in the composite feature vector. The longitudinal local gradient correlation features of the average sound velocity distribution data field component and the physiological attenuation coefficient component of the first characteristic frequency in the spatial distribution are identified, namely, the rate of change of mechanical and acoustic physical parameters along the length of the withering trough and their mutual coupling relationship. The spatiotemporal gradient coupling feature extraction network analyzes the correlation between the sound velocity gradient and the attenuation gradient through convolution kernel operations, thereby locating the spatial distribution anomalies of physical parameters caused by local packing density anomalies or undercooking phenomena in the tea withering accumulation layer. For example, when the convolution kernel in the one-dimensional causal convolution layer captures a feature pair with a gentle sound velocity gradient (i.e., the rate of change of the sound velocity gradient is lower than a preset threshold for the rate of change of the sound velocity gradient) but a steep attenuation gradient (i.e., the rate of change of the attenuation gradient is higher than a preset threshold for the rate of change of the attenuation gradient), the spatiotemporal gradient coupling feature extraction network maps this feature pair to represent a non-uniform withering state in which the tea withering accumulation layer is in a state of rapid surface water loss but unchanged internal hardness.

[0115] Second, the nonlinear mapping of the longitudinal local gradient correlation feature space. The longitudinal local gradient correlation features, after passing through a one-dimensional causal convolutional layer, are input into a fully connected layer in the spatiotemporal gradient coupling feature extraction network. Using a nonlinear activation function, such as Leaky ReLU, this fully connected layer maps the physical feature representation space represented by the longitudinal local gradient correlation features to the target physiological indicator space defined by water content and breakage rate, thereby fitting data including the average sound velocity distribution. Physiological attenuation coefficient of the first characteristic frequency and tissue damage index The complex nonlinear constitutive relationship between physical measurement parameters and biochemical indicators including local water content and cell wall breakage rate.

[0116] Third, the generation of the physiological and biochemical state distribution matrix. The output layer of the spatiotemporal gradient coupling feature extraction network generates two parallel matrices corresponding to the longitudinal displacement coordinates through nonlinear regression operations. and sampling period index A state distribution matrix with consistent topological dimensions, i.e., a local water content distribution map characterizing the degree of water loss. Distribution of cell wall fragmentation rate characterizing fermentation precursor conditions .

[0117] Step S4000, based on the local moisture content distribution map Cell wall fragmentation rate distribution map The spatially weighted average moisture content was calculated. and the stacking thickness field matrix A global environmental control instruction set is generated based on the multivariable feedback compensation control logic mapping. Combined with the identified set of abnormal longitudinal displacement coordinates Solve the target motion parameter vector to generate a fixed-point targeted material scraping command. .

[0118] Specifically, this step aims to solve the technical problem of localized over-withering or under-withering caused by uneven distribution of the environmental field in traditional tea withering processes. It utilizes the local moisture content distribution map representing the overall moisture state from step S3200. And a distribution map of cell wall fragmentation rate characterizing the state of the entire fermentation precursor. And the continuous stacking thickness field matrix characterizing the global physical stacking state from step S1200. By transforming the spatiotemporal distribution characteristics of the detection data into action commands for the actuators, a global environmental control command set is calculated and output to control the macroscopic operating parameters of ventilation fans, heaters, and humidifiers. And a fixed-point targeted rake command that includes parameters such as the moving target position, the depth of the rake teeth, and the mixing intensity. This achieves uniform withering quality and optimization of the process path.

[0119] Further, step S4000 includes:

[0120] Step S4100: Introduce a non-uniform flow field spatial weighting function to the local water content distribution map. Perform fluid dynamics corrections to calculate the space-weighted average water content. via sampling period index Differential operations are used to obtain the global water loss rate. And combined with the preset withering target evolution trajectory Construct a state deviation vector, and map the state deviation vector to a global environmental control instruction set based on multivariable feedback compensation control logic. .

[0121] Specifically, this step aims to utilize the local moisture content distribution map from step S3200. and cell wall fragmentation rate distribution map Combined with the pre-set withering target evolution trajectory By using a spatial domain weighted integral transformation based on fluid dynamics weights, the state distribution matrix, which serves as a high-dimensional feature input, is reduced in dimensionality and mapped to a global state representative value that characterizes the overall drying degree of all tea leaves, namely, the spatially weighted average moisture content. Furthermore, based on multivariable feedback compensation control logic, the instantaneous withering state of all tea leaves in the trough at the current moment and the evolution trajectory of the withering target are calculated. The state deviation vector between them generates a global environment control instruction set. This process aims to establish the basic thermodynamic driving potential energy conditions within the withering tank, such as vapor pressure deficit, to provide the necessary environmental kinetic basis for the transmembrane transport of moisture in tea leaves and prevent irreversible deviations from the withering process path caused by deviations in macroscopic temperature and humidity boundary conditions.

[0122] In the specific implementation process, considering the inherent physical differences in airflow field distribution and wind pressure gradient in different sections of the withering trough, such as the air inlet and outlet, this step introduces a non-uniform flow field spatial weighting function based on computational fluid dynamics simulation to adjust the local moisture content distribution map. Perform fluid dynamics correction. The specific execution logic of the fluid dynamics correction is as follows: The data processing unit indexes the current sampling period... Below, the longitudinal displacement coordinates of the withering trough. As the integration variable, within the integration domain determined by the total physical length of the withering trough, the local moisture content distribution map is... The product of the product with the non-uniform flow field spatial weighting function is subjected to definite integral operation, and the resulting integral is normalized by multiplying it by the reciprocal of the total physical length of the withering trough. This process yields the spatially weighted average moisture content, which accurately reflects the overall moisture state of the tea leaves in the entire trough. Subsequently, based on the currently executing sampling period index... Spatial weighted average moisture content Perform differentiation to obtain the index of the current sampling period. Global water loss rate The non-uniform flow field spatial weighting function is a dimensionless correction coefficient preset based on the aerodynamic characteristics inside the withering trough; the total physical length of the withering trough is a geometric constant that defines the upper limit of the integration domain for spatial domain integration operations, and its unit is meters (m). Its value is determined by the mechanical design drawings of the withering trough equipment.

[0123] Real-time calculated spatial weighted average moisture content and overall water loss rate Evolution trajectory of the pre-set withering target Real-time comparisons are performed to construct a state deviation vector containing the difference in moisture content and the difference in water loss rate. Based on multivariable feedback compensation control logic, the data processing unit maps the state deviation vector into a global environmental control instruction set for adjusting the thermodynamic boundary conditions within the withering tank. .

[0124] The specific control logic of the multivariable feedback compensation control logic is as follows:

[0125] When the spatially weighted average moisture content is detected The absolute value of the total water loss rate exceeds the target threshold set by the standard trajectory. When the vapor pressure is less than the preset minimum water loss slope, the current environment is determined to have insufficient vapor pressure deficit. An auxiliary dehumidification command, including parameters to increase fan speed and to activate auxiliary heating, is generated and encapsulated into the global environmental control command set. The aim is to increase the water potential difference between the leaf surface and the ambient air, thereby forcibly accelerating water evaporation;

[0126] When the spatially weighted average moisture content is detected The cell wall fragmentation rate distribution map shows that the cell falls within the preset critical range of the wilting endpoint threshold. When the average value exceeds the preset moderate standard, a dynamic temperature and humidity balance instruction is generated, which includes parameters for reducing wind speed and maintaining constant temperature, and encapsulated into the global environmental control instruction set. The control mechanism gradually reduces the wind speed and maintains a constant temperature, allowing the tea leaves to enter a balanced state of moisture loss and fermentation, preventing excessive withering due to inertia.

[0127] Step S4200: Traverse the local moisture content distribution map Cell wall fragmentation rate distribution map Spatial weighted average moisture content and the stacking thickness field matrix To identify those that meet a preset physiological deviation threshold Geometric stacking threshold and biochemical statistical bias threshold abnormal longitudinal displacement coordinate set Combined with the global environment control instruction set The fan speed parameter constraint is used to calculate the target motion parameter vector and generate a fixed-point targeted material scraping command. .

[0128] Specifically, this step aims to utilize the local water content distribution map from step S3200, which characterizes the spatial heterogeneity of physiological water in the medium. Distribution of cell wall fragmentation rate characterizing spatial synchronicity of biochemical transformation Spatial weighted average moisture content from step S4100 And the packing thickness field matrix from step S1200, which characterizes the physical boundary of the geometric packing morphology of the medium. Combined with the global environmental control instruction set from step S4100, which serves as an environmental constraint parameter for the motion intensity of the mechanical actuator. It performs clustering identification and targeted action calculation for localized processing defect areas. This process is achieved by controlling the instruction set by the global environment. The global environmental thermodynamic driving force provided by the fixed-point targeted material scraping command By combining the triggered local mechanical disturbances, a target motion parameter vector is calculated, which includes parameters such as the moving target position, the depth of the rake teeth, and the mixing intensity. Finally, a fixed-point targeted rake command is generated. This is a targeted material-raking instruction. The aim is to perform point-to-point physical turning of the withering lag physical space area due to abnormal longitudinal displacement coordinate index, reconstruct the internal pore channel structure of the tea accumulation layer in this area and the gas-solid exchange contact area of ​​the leaf surface, eliminate local processing defects caused by uneven material accumulation density and dead angle of airflow field, and achieve uniformity of tea withering quality throughout the whole trough.

[0129] In the specific implementation process, this step traverses the local moisture content distribution map. Cell wall fragmentation rate distribution map and the stacking thickness field matrix Based on a preset physiological deviation threshold Geometric stacking threshold and biochemical statistical bias threshold Identify and mark the set of abnormal longitudinal displacement coordinates that require mechanical intervention. Among them, the physiological deviation threshold This is based on statistical calibration of historical high-quality tea withering data. The specific method for obtaining this data is as follows: Select withering process data of several batches of high-quality tea, calculate the spatial distribution standard deviation of moisture content at each time point, and then set the physiological deviation threshold. Set to 1.5 to 2.0 times this standard deviation. The geometric stacking threshold. The determination of the critical penetration capability of the withering trough ventilation system is based on the following method: Under rated fan power conditions, the outlet wind speed after the airflow vertically penetrates tea leaf stacks of different heights is measured using a hot-wire anemometer. The critical stacking height corresponding to the point where the outlet wind speed decays to the minimum effective evaporation critical wind speed is determined as the geometric stacking threshold. The biochemical statistical deviation threshold. The acquisition is based on the real-time discrete adaptive calculation of the global cell wall fragmentation rate. The specific acquisition method is as follows: Real-time calculation of the cell wall fragmentation rate distribution map at the current moment. Spatial distribution standard deviation and mean cell wall fragmentation rate of the whole tank The spatial distribution standard deviation and the mean cell wall breakage rate of the entire tank were then compared. The ratio is defined as the coefficient of variation. A positive correlation mapping relationship is constructed between the coefficient of variation and the adaptive proportional coefficient, such that the adaptive proportional coefficient dynamically fluctuates within the numerical range of 0.10 to 0.15 as the coefficient of variation increases. Finally, the biochemical statistical deviation threshold is... Set as the average cell wall breakage rate of the entire tank The product of the adaptive scaling factor and the adaptive scaling factor.

[0130] The abnormal longitudinal displacement coordinate set The specific identification and labeling logic is as follows:

[0131] Extracting local moisture content distribution map The real-time local moisture content element value corresponding to the longitudinal displacement coordinate of a certain discrete spatial sampling point is defined as follows: if the real-time local moisture content element value exceeds the spatial weighted average moisture content... Furthermore, the numerical deviation between the two exceeds the physiological deviation threshold. The data processing unit determines that the local tea leaf accumulation unit indexed by the longitudinal displacement coordinate is a region where moisture evaporation is hindered, and marks the longitudinal displacement coordinate and includes it in the abnormal longitudinal displacement coordinate set. ;

[0132] Extracting the packing thickness field matrix The real-time stacking thickness element value corresponds to the longitudinal displacement coordinate of a discrete spatial sampling point. If this real-time stacking thickness element value exceeds the geometric stacking threshold... The data processing unit determines that the local tea leaf accumulation unit indexed by the longitudinal displacement coordinate is an airflow penetration resistance zone. This determination directly depends on the accumulation thickness field matrix. Provide physical boundary constraints, and mark the longitudinal displacement coordinate and classify it into the abnormal longitudinal displacement coordinate set. ;

[0133] Distribution of cell wall fragmentation rate The data processing unit determines the local tea leaf accumulation unit indexed by the longitudinal displacement coordinate as a mechanical intervention compensation zone when the real-time cell wall breakage rate element value at the longitudinal displacement coordinate of a discrete spatial sampling point is lower than the global cell wall breakage rate statistical benchmark. This longitudinal displacement coordinate is then marked and included in the abnormal longitudinal displacement coordinate set. The calculation logic for the statistical benchmark of the global cell wall breakage rate is as follows: First, the cell wall breakage rate distribution map is analyzed. Perform spatial domain integration to calculate the average cell wall fragmentation rate across the entire cell tank. Then, the average cell wall breakage rate of the entire tank was calculated. Deviation from preset biochemical statistical threshold The difference is used as the statistical benchmark for the global cell wall breakage rate.

[0134] For each identified abnormal longitudinal displacement coordinate, this step generates a corresponding fixed-point targeted harrowing instruction. This is a targeted material-raking instruction. It is not a simple switching quantity, but a targeted action parameter vector that includes specific moving target position parameters, rake tooth penetration depth parameters, and mixing intensity parameters. Its generation logic fuses the stacking thickness field matrix. Geometric field physical constraints and global environment control instruction set Global environmental operating condition constraints.

[0135] The fixed-point targeted material scraping command The specific generation logic is as follows:

[0136] First, the calculation of the moving target position parameters: the extracted abnormal longitudinal displacement coordinates are converted into the moving target position parameters of the servo drive system of the rake device, which aims to drive the rake device to move quickly along the guide rail to the withering lag physical space region indexed by the abnormal longitudinal displacement coordinates.

[0137] Second, the calculation of the rake tooth penetration depth parameter: extracting the accumulation thickness field matrix. The real-time accumulation thickness element value corresponding to the abnormal longitudinal displacement coordinate is used to dynamically calculate the rake tooth penetration depth parameter of the rake tooth motor, combined with the preset mechanical safety margin. The rake tooth penetration depth parameter serves as a geometric field physical constraint, ensuring effective stirring of the bottom tea leaves while avoiding contact with the bottom surface of the withering trough, thus maximizing operation within the geometric safety boundary.

[0138] Third, the calculation of the turning intensity parameter: This involves calculating the targeted action intensity of the rake motor, i.e., the turning intensity parameter. This turning intensity parameter depends not only on the local moisture content deviation and the real-time buildup thickness, but also on the global environmental control command set. The global environmental constraints, i.e., the weighted influence of the wind turbine speed parameter, include, for example, appropriately reducing the mechanical tumbling frequency under high global wind speed conditions to prevent dust and blade breakage. The local moisture content deviation is the difference between the real-time local moisture content element value and the spatially weighted average moisture content at the abnormal longitudinal displacement coordinate. The resulting difference is used to quantify the degree of moisture lag in the withering lag physical space region indexed by the abnormal longitudinal displacement coordinates.

[0139] The specific calculation logic for the turning intensity parameter is as follows: The product of the local moisture content deviation and the moisture deviation weight gain coefficient is calculated as the moisture dimension driving component; the product of the real-time stack thickness element and the thickness weight gain coefficient is simultaneously calculated as the geometric dimension driving component; subsequently, the moisture dimension driving component and the geometric dimension driving component are added to obtain the basic action intensity, and this basic action intensity is multiplied by the environmental constraint weighting coefficient for global working condition correction, thereby solving for the final turning intensity parameter. Here, the moisture deviation weight gain coefficient is a preset proportional constant used to balance the influence of moisture differences on the turning intensity parameter; the thickness weight gain coefficient is a preset proportional constant used to balance the influence of stack thickness on the turning intensity parameter, ensuring that a larger mechanical torque is applied in areas with thicker stacks to overcome material resistance; the environmental constraint weighting coefficient is based on a global environmental control instruction set. The dynamic dimensionless factor, calculated from the fan speed parameters, is usually negatively correlated with the real-time wind speed and is used to coordinate the coupling relationship between mechanical stirring action and macroscopic airflow field.

[0140] Finally, the calculated moving target position parameters, rake tooth penetration depth parameters, and mixing intensity parameters are encapsulated and constructed into a fixed-point targeted rake command. And send it to the implementing agency.

[0141] Example 2:

[0142] This embodiment, based on Embodiment 1, provides a tea withering detection system based on ultrasonic attenuation, such as... Figure 4 As shown, the system includes a geometric field correction module, a sound attenuation feature extraction module, a parameter inversion module, and a closed-loop correction module;

[0143] The geometric field correction module is used to synchronously acquire the original echo signal. Sensor array spatial position parameters and real-time ambient temperature To generate discrete distance vector sequences Using cubic spline interpolation operators to analyze discrete distance vector sequences Spatial sequence integration is performed to construct the stacking thickness field matrix. And based on the power-law function of compressible rheological properties, the accumulation thickness field matrix is... Perform numerical calculations and output the porosity correction factor. ;

[0144] The acoustic attenuation feature extraction module is used to acquire the transmitted acoustic response signal and construct the transmission amplitude field using the Hilbert transform operator and cross-correlation time delay estimation logic. Harmony sound wave propagation time field Based on the sound wave propagation time field Constructing the average sound speed distribution data field Combined with porosity correction factor For transmission amplitude field Perform density decoupling operations to generate physiological attenuation coefficients. and tissue damage index .

[0145] The parameter inversion module is based on the average sound velocity distribution data field. Physiological attenuation coefficient of the first characteristic frequency Tissue damage index and environmental state parameter vector To construct composite feature vectors Using a spatiotemporal gradient coupled feature extraction network to extract composite feature vectors Perform sliding convolution and nonlinear regression mapping to invert and solve the local moisture content distribution map. Distribution of cell wall fragmentation rate ;

[0146] The closed-loop correction module is based on the local moisture content distribution map. Cell wall fragmentation rate distribution map The spatially weighted average moisture content was calculated. and the stacking thickness field matrix A global environmental control instruction set is generated based on the multivariable feedback compensation control logic mapping. Combined with the identified set of abnormal longitudinal displacement coordinates Solve the target motion parameter vector to generate a fixed-point targeted material scraping command. .

[0147] The parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0148] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting tea leaf withering based on ultrasonic attenuation, characterized in that, include: The original echo signal, sensor array spatial position parameters and real-time ambient temperature are acquired synchronously to generate a discrete distance vector sequence. The discrete distance vector sequence is spatially integrated using a cubic spline interpolation operator to construct the packing thickness field matrix. Numerical calculations are then performed on the packing thickness field matrix based on the compressive rheological power law function to output the porosity correction factor. The transmitted acoustic response signal is collected, and the transmission amplitude field and sound wave propagation time field are constructed using the Hilbert transform operator and cross-correlation time delay estimation logic. The average sound velocity distribution data field is constructed based on the sound wave propagation time field. The density decoupling operation is performed on the transmission amplitude field in combination with the porosity correction factor to generate the physiological attenuation coefficient and tissue damage index. The physiological attenuation coefficient is assigned a value based on the frequency band identifier. If the frequency band identifier points to the first characteristic frequency pulse signal, the value of the first characteristic frequency physiological attenuation coefficient is taken; if the frequency band identifier points to the second characteristic frequency pulse signal, the value of the second characteristic frequency physiological attenuation coefficient is taken. The first characteristic frequency pulse signal and the second characteristic frequency pulse signal are alternately emitted by the side wall ultrasonic transmitting probe assembly. A composite feature vector is constructed based on the average sound velocity distribution data field, the physiological attenuation coefficient of the first characteristic frequency, the tissue structure damage index and the environmental state parameter vector. A spatiotemporal gradient coupled feature extraction network is used to perform sliding convolution operation and nonlinear regression mapping on the composite feature vector to invert and solve the local water content distribution map and cell wall breakage rate distribution map. Based on the local moisture content distribution map, cell wall breakage rate distribution map, and the calculated spatial weighted average moisture content and accumulation thickness field matrix, a global environmental control instruction set is generated according to the multivariate feedback compensation control logic mapping. Combined with the identified abnormal longitudinal displacement coordinate set, the target action parameter vector is calculated to generate a fixed-point targeted material rake instruction.

2. The method for detecting tea withering based on ultrasonic attenuation according to claim 1, characterized in that, The method for outputting the porosity correction factor includes: The original echo signal reflected from the surface of the withered tea stack layer, the spatial position parameters of the sensor array, and the real-time ambient temperature in the withering trough were collected simultaneously. Time-domain waveform analysis was performed on the original echo signal to determine the round-trip time difference of the sound wave. A thermodynamic sound velocity compensation mechanism was constructed based on the real-time ambient temperature to obtain the thermodynamically corrected air sound velocity. The vertical physical distance from each discrete spatial sampling point to the surface of the withered tea stack layer was calculated by combining the round-trip time difference of the sound wave. A discrete distance vector sequence was constructed based on the topological order of the spatial position parameters of the sensor array. The preset sensor installation absolute height constant is used to perform differential operation on the discrete distance vector sequence to generate discrete medium thickness values. A cubic spline interpolation operator is introduced to construct a tridiagonal matrix equation system. The tridiagonal matrix equation system is solved using the chasing method to generate a global continuous smooth curve function. The global continuous smooth curve function is discretized and resampled to obtain the blind zone reconstructed thickness value. The discrete medium thickness values ​​are combined with the spatial sequence integration to construct the stacking thickness field matrix. The geometric compressibility ratio of the medium is determined based on the packing thickness field matrix and the preset standard reference loose thickness. Numerical calculations are then performed on the geometric compressibility ratio of the medium according to the power law function of the compressibility rheological properties to obtain the porosity correction factor.

3. The method for detecting tea withering based on ultrasonic attenuation according to claim 2, characterized in that, The method for calculating the vertical physical distance from each discrete spatial sampling point to the surface of the tea withering accumulation layer includes: The thermodynamically corrected air speed is obtained by multiplying the standard air speed constant by a thermodynamic correction factor obtained by taking the square root of the ratio of real-time ambient temperature to the zero point constant of the thermodynamic absolute temperature scale plus 1. The round-trip time difference of the sound wave is obtained by calculating the difference between the absolute arrival time of the reflected echo envelope front and the absolute start time of the excitation pulse. The vertical physical distance is calculated by multiplying the thermodynamically corrected air speed of sound and the round-trip time difference of the sound wave, and then multiplying the product by a coefficient of one-half. The standard state air speed constant is the reference phase velocity of sound waves propagating in dry air at standard atmospheric pressure and a thermodynamic temperature of 0°C; the thermodynamic absolute temperature scale zero-point constant is the reference value for conversion between the Celsius and Kelvin thermodynamic temperature scales; the reflected echo envelope leading edge refers to the first rising edge of the reflected echo arriving at the gas-solid two-phase acoustic reflection interface; the excitation pulse absolute start time is the zero moment of electrical pulse excitation.

4. The method for detecting tea withering based on ultrasonic attenuation according to claim 1, characterized in that, The method for constructing the average sound velocity distribution data field includes: Based on the time-division multiplexing mechanism, the sidewall ultrasonic transmitting probe assembly alternately transmits the first characteristic frequency pulse signal and the second characteristic frequency pulse signal, and collects the transmitted acoustic response signal after penetrating the withered tea accumulation layer. The transmitted acoustic response signal is analyzed using the Hilbert transform operator to construct the first characteristic frequency transmission amplitude field and the second characteristic frequency transmission amplitude field. The acoustic transmission delay is quantized based on the cross-correlation time delay estimation logic to construct the sound wave propagation time field. The preset system inherent delay constant is used to perform time difference calibration on the sound wave propagation time field to obtain the net flight time. The average sound velocity distribution data field is constructed based on the ratio of the withering groove width constant to the net flight time.

5. The method for detecting tea withering based on ultrasonic attenuation according to claim 4, characterized in that, The method for generating the physiological attenuation coefficient and the tissue damage index includes: performing a logarithmic operation on the ratio of the preset transmitted signal reference amplitude and the transmitted amplitude field according to the Beer-Lamber law; calculating the total absolute attenuation by combining the withering groove width constant; introducing a porosity correction factor to perform a density decoupling operation on the total absolute attenuation to extract the physiological attenuation coefficient; and generating the tissue damage index based on the ratio of the physiological attenuation coefficients at different frequencies. The transmission amplitude field is assigned a value based on the frequency band identifier. If the frequency band identifier points to the first characteristic frequency pulse signal, the transmission amplitude field is taken as the first characteristic frequency transmission amplitude field. If the frequency band identifier points to the second characteristic frequency pulse signal, the transmission amplitude field is taken as the second characteristic frequency transmission amplitude field.

6. The method for detecting tea withering based on ultrasonic attenuation according to claim 1, characterized in that, The methods for calculating the local water content distribution map and the cell wall fragmentation rate distribution map include: The average sound velocity distribution data field, the physiological attenuation coefficient of the first characteristic frequency, the tissue damage index and the environmental state parameter vector are obtained. The environmental state parameter vector is expanded in the whole domain using the dimensional broadcasting algorithm, combined with Z-Score normalization processing, and each feature channel is spliced ​​into a composite feature vector based on discrete spatial sampling points. The composite feature vector is input into the spatiotemporal gradient coupling feature extraction network. A one-dimensional causal convolutional layer is used to perform sliding convolution operation along the vertical displacement coordinate to extract the vertical local gradient correlation features. The vertical local gradient correlation features are then subjected to nonlinear regression mapping through a fully connected layer to invert and solve the local water content distribution map and the cell wall breakage rate distribution map.

7. The method for detecting tea withering based on ultrasonic attenuation according to claim 1, characterized in that, The method for generating the fixed-point targeted material scraping command includes: A non-uniform flow field spatial weighting function is introduced to perform hydrodynamic correction on the local water content distribution map to calculate the spatial weighted average water content. The global water loss rate is obtained through the differential operation of the sampling period index, and a state deviation vector is constructed by combining the preset withering target evolution trajectory. Based on the multivariate feedback compensation control logic, the state deviation vector is mapped to a global environmental regulation instruction set. The system iterates through local moisture content distribution maps, cell wall breakage rate distribution maps, spatially weighted average moisture content and accumulation thickness field matrices to identify abnormal longitudinal displacement coordinate sets that meet preset physiological deviation thresholds, geometric accumulation thresholds and biochemical statistical deviation thresholds. Combined with the fan speed parameter constraints of the global environmental control instruction set, the system calculates the target action parameter vector and generates a fixed-point targeted material scraping instruction.

8. The method for detecting tea withering based on ultrasonic attenuation according to claim 7, characterized in that, The steps of the multivariable feedback compensation control logic include: When the spatial weighted average moisture content is detected to exceed the target threshold set by the standard trajectory, and the absolute value of the global water loss rate is less than the preset minimum water loss slope, it is determined that the current environment has insufficient vapor pressure deficit. A ventilation acceleration command containing the fan speed parameter and an auxiliary dehumidification command containing the auxiliary heating parameter are generated, and the ventilation acceleration command and auxiliary dehumidification command are encapsulated into the global environment control command set. When the spatially weighted average moisture content falls into the preset critical range of the wilting endpoint threshold, and the mean value of the cell wall breakage rate distribution map exceeds the preset moderate standard, a dynamic temperature and humidity balance instruction containing parameters for reducing wind speed and maintaining constant temperature is generated and encapsulated into the global environmental control instruction set.

9. The method for detecting tea withering based on ultrasonic attenuation according to claim 7, characterized in that, The method for identifying the abnormal longitudinal displacement coordinate set includes: Extract the real-time local moisture content element value at the longitudinal displacement coordinate corresponding to each discrete spatial sampling point in the local moisture content distribution map. If the real-time local moisture content element value exceeds the spatial weighted average moisture content and the numerical deviation between the two exceeds the physiological deviation threshold, determine that the local tea accumulation unit indexed by the longitudinal displacement coordinate is a moisture evaporation obstruction area, and mark the corresponding longitudinal displacement coordinate and classify it into the abnormal longitudinal displacement coordinate set. Extract the real-time stacking thickness element value at the longitudinal displacement coordinate of each discrete spatial sampling point in the stacking thickness field matrix. If the real-time stacking thickness element value exceeds the geometric stacking threshold, determine that the local tea stacking unit indexed by the longitudinal displacement coordinate is the airflow penetration resistance zone, and mark the corresponding longitudinal displacement coordinate and classify it into the abnormal longitudinal displacement coordinate set. Extract the real-time cell wall breakage rate element value at the longitudinal displacement coordinate corresponding to each discrete spatial sampling point in the cell wall breakage rate distribution map. If the real-time cell wall breakage rate element value is less than the global cell wall breakage rate statistical benchmark, determine that the local tea accumulation unit indexed by the longitudinal displacement coordinate is a mechanical intervention compensation area, and mark the corresponding longitudinal displacement coordinate and classify it into the abnormal longitudinal displacement coordinate set. The calculation method for the global cell wall breakage rate statistical benchmark is as follows: perform spatial domain integration on the cell wall breakage rate distribution map to solve the mean value of the cell wall breakage rate of the entire tank, calculate the difference between the mean value of the cell wall breakage rate of the entire tank and the preset biochemical statistical deviation threshold, and determine the difference as the global cell wall breakage rate statistical benchmark.

10. A tea withering detection system based on ultrasonic attenuation, used to implement the tea withering detection method based on ultrasonic attenuation as described in any one of claims 1-9, characterized in that, The system includes a geometric field correction module, a sound attenuation feature extraction module, a parameter inversion module, and a closed-loop correction module. The geometric field correction module is used to synchronously acquire the original echo signal, the spatial position parameters of the sensor array, and the real-time ambient temperature to generate a discrete distance vector sequence. It uses a cubic spline interpolation operator to spatially integrate the discrete distance vector sequence to construct the packing thickness field matrix. It also performs numerical calculations on the packing thickness field matrix based on the power law function of compressibility rheology and outputs a porosity correction factor. The acoustic attenuation feature extraction module is used to collect the transmitted acoustic response signal, construct the transmission amplitude field and the sound wave propagation time field using the Hilbert transform operator and cross-correlation time delay estimation logic, construct the average sound velocity distribution data field based on the sound wave propagation time field, and perform density decoupling operation on the transmission amplitude field in combination with the porosity correction factor to generate the physiological attenuation coefficient and tissue damage index. The parameter inversion module constructs a composite feature vector based on the average sound velocity distribution data field, the physiological attenuation coefficient of the first characteristic frequency, the tissue structure damage index, and the environmental state parameter vector. It then uses a spatiotemporal gradient coupling feature extraction network to perform sliding convolution and nonlinear regression mapping on the composite feature vector to invert and solve the local water content distribution map and cell wall breakage rate distribution map. The closed-loop correction module: based on the local moisture content distribution map, cell wall breakage rate distribution map, calculated spatial weighted average moisture content and accumulation thickness field matrix, generates a global environmental control instruction set according to the multivariate feedback compensation control logic mapping, and calculates the target action parameter vector by combining the identified abnormal longitudinal displacement coordinate set, and generates a fixed-point targeted material rake instruction.

Images