A grinding concentration measuring device for soybean milk powder processing

Abstract

The invention discloses a grinding concentration measuring device for soybean milk powder processing, and relates to the technical field of soybean milk powder processing equipment. The system comprises the following components: a sampling unit, a spectral analysis unit, a rheological detection unit, a data processing and control unit, and a display and alarm unit. The invention adopts a dynamic adaptive flow algorithm through the sampling unit, and can adjust the sampling flow according to the real-time characteristics of the soybean milk sample to ensure the representativeness and accuracy of the sample. The spectral analysis unit and the rheological detection unit respectively use spectral detection technology and rheological characteristic measurement technology to deeply analyze the composition and rheological characteristics of the soybean milk sample. The data processing and control unit adopts a fusion intelligent algorithm to efficiently process and analyze the collected data, and finally calculates the accurate grinding concentration. This process not only improves the accuracy of the measurement, but also significantly improves the measurement efficiency, and provides strong technical support for the soybean milk powder processing process.

Description

Technical Field

[0001] The invention relates to the technical field of soybean milk powder processing equipment, in particular to a grinding concentration measuring device for soybean milk powder processing. Background Art

[0002] In the process of soy milk powder processing, accurate determination of grinding concentration is crucial to ensure product quality and production efficiency. Fluctuations in grinding concentration will directly affect the taste, nutritional value and stability of the final soy milk powder.

[0003] Although traditional technologies can measure the grinding concentration to a certain extent, they each have obvious shortcomings. Traditional technologies are often limited by the accuracy and stability of the spectrometer, making it difficult to accurately analyze the multiple components in soymilk samples. On the other hand, when measuring the rheological properties of soymilk samples, due to the limitations of sensor accuracy and data processing methods, the grinding concentration of soymilk is often unable to be accurately reflected.

[0004] In summary, traditional technology has many shortcomings in the determination of grinding concentration for soy milk powder processing, and cannot meet the needs of the modern soy milk powder processing industry for high precision, high efficiency and intelligence. Therefore, it is particularly important to develop a grinding concentration determination device for soy milk powder processing. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the prior art and to provide a pulping concentration measuring device for soy milk powder processing. The device can use the spectral detection technology and the rheological property measurement technology of the spectral analysis unit and the rheological detection unit respectively to conduct in-depth analysis of the composition and rheological properties of the soy milk sample. The data processing and control unit adopts a fusion intelligent algorithm to efficiently process and analyze the collected data, and finally calculate the accurate pulping concentration.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a grinding concentration measuring device for soybean milk powder processing, the system comprises the following components: a sampling unit, a spectral analysis unit, a rheological detection unit, a data processing and control unit, and a display and alarm unit;

[0007] The sampling unit is used to collect soybean milk samples from the grinding device, and has an adjustable sampling flow control device. The flow control device operates according to a dynamic adaptive flow algorithm, and the formula is:

[0008]

[0009] Among them, Q is the sampling flow rate mL / min, k1 is the basic coefficient of dynamic flow regulation, and adaptive correction is performed according to the real-time feedback data during the operation of the equipment. The correction logic optimizes the fitting of the relationship between the concentration measurement error and the flow rate after each sampling according to the least squares method.i is the estimated proportion of the i-th component in soy milk, which is obtained through the near-infrared spectroscopy rapid pre-analysis model of soybean raw materials. This model is constructed based on the near-infrared spectral data and component analysis data of a large number of soybean samples from different origins and varieties using the principal component regression algorithm. i is the flow impact factor corresponding to the i-th component, D t It is the dispersion index of the current soy milk, which is calculated by the laser particle size analysis data of the soy milk sample, reflecting the uniformity of the particle size distribution in the soy milk. The calculation formula is: where d j is the diameter of the jth particle (μm), is the average particle diameter, m is the total number of sampled particles;

[0010] The spectrum analysis unit is connected to the sampling unit, and uses spectrum detection technology to perform spectrum analysis on the collected soybean milk samples. It has a unique adaptive spectrum scanning module, and its scanning wavelength range and step length are determined by the intelligent spectrum planning algorithm. The formula is as follows:

[0011]

[0012] Among them, W start is the starting scanning wavelength (nm), W end is the end scanning wavelength (nm), Δλ is the scanning step length (nm), C k is the relative content of the kth chemical bond in soy milk, A k is the absorption sensitivity coefficient corresponding to the kth chemical bond, E r It is the energy efficiency index of the current device, which is obtained in real time through the energy monitoring module inside the device. This index is used to optimize the scanning range and step length to reduce energy consumption while ensuring the accuracy of spectral analysis;

[0013] The rheological detection unit: combines dynamic spectral imaging and spectral rheology to measure the rheological properties of the soy milk sample. The viscosity sensor of the rheological detection unit adopts a high-precision rotational viscometer, and its rotation speed is adjusted by the rheological feedback control algorithm. The formula is:

[0014]

[0015] Where N is the rotation speed of the rotational viscometer (rpm), k2 is the basic constant for speed regulation, and the initial value is obtained by neural network fitting the experimental data of rheological properties of soy milk samples in different concentration ranges under different temperature and pressure conditions. The gradient descent algorithm is then used to correct the deviation between the real-time detection data and the predicted data. target is the middle value of the target viscosity range (mPa·s), which is set according to the soymilk powder processing requirements.init is the estimated value of initial viscosity (mPa·s), which is calculated by using the empirical formula based on the initial moisture content and protein structure characteristic parameters of soybean raw materials. process is the time of the current soybean milk powder processing step (min), T ref is the reference processing time, R l is the relative concentration of the first rheologically influencing component in soy milk, F l is the viscosity contribution factor corresponding to the first rheological influencing component, the rheological detection unit further includes a temperature sensor and a pressure sensor, which are used to measure the temperature and pressure of the soymilk sample respectively, and the data collected by these sensors are transmitted to the data processing and control unit together with the viscosity data;

[0016] The data processing and control unit receives the data collected by the spectral analysis unit and the rheological detection unit, and processes and analyzes the data. The unit uses a fusion intelligent algorithm to model and analyze the data. The algorithm is as follows:

[0017] C final =α×C spec-new +β×C rh eo-new +γ×C int-new

[0018] Among them, C final is the final calculated pulp concentration (g / mL), C spec-new In order to improve the concentration value calculated from spectral data, a deep spectral feature extraction and conversion model is used. The model first performs wavelet transform on the original spectral data to extract spectral features at different scales, then uses an autoencoder neural network to compress and reconstruct the features, and finally inputs the reconstructed features into a support vector machine regression model to obtain the concentration value. The kernel function parameters in the model are optimized by a genetic algorithm to maximize the prediction accuracy of the model. rh eo-new In order to obtain the concentration value calculated based on the improved rheological data, a concentration calculation model based on a fuzzy logic reasoning system is constructed by performing principal component fusion analysis on the viscosity, temperature and pressure rheological data. The input of the fuzzy logic system is the fused principal component, and the output is the concentration value. α, β and γ are weight coefficients, which are determined by a multi-stage dynamic analysis of the influence of spectral and rheological data on the accuracy of concentration measurement at different process stages during the processing of soybean milk powder. In the early stage of refining, the α weight is relatively large. As the processing proceeds, the β weight gradually increases, and the γ weight is determined according to the coordinated change trend of spectral and rheological data. The data processing and control unit can also communicate with other equipment in the soybean milk powder processing process, and transmit the calculated refining concentration data to the PLC control system to realize intelligent regulation of the entire processing process.

[0019] The display and alarm unit is used to display the measured grinding concentration data and the operating status information of the equipment. The display and alarm unit adopts a touch display screen, and the alarm signal adopts an audible and visual alarm method. When the calculated grinding concentration exceeds the preset normal range, the microprocessor controls the alarm circuit to send out an audible and visual alarm signal. The preset normal range is set according to the quality standards and processing requirements of the soymilk powder product, and can be adjusted on the touch display screen.

[0020] Furthermore, the sampling unit also includes a sample pre-processing device, which includes a separation module based on the synergistic effect of centrifugal force and filtration, and its working principle is based on a centrifugal filtration algorithm:

[0021]

[0022] Among them, F sep is the separation force, m s is the mass of suspended solid particles in soy milk, obtained by preliminary mass analysis of soy milk samples, v is the centrifugal rotation speed, r is the centrifugal radius, k3 is the filtration synergy coefficient, P f It is the pressure drop of the filter medium, which is related to the pore size, thickness of the filter medium and the viscosity of the soy milk. The separation module can dynamically adjust the separation force according to the characteristics of the soy milk sample during the centrifugation process, effectively remove the impurity particles and bubbles in the soy milk, and improve the accuracy of subsequent spectral analysis and rheological detection. At the same time, the sample pretreatment device is also equipped with an electromagnetic stirrer, and its stirring speed is controlled by the microprocessor according to the viscosity and particle distribution of the soy milk sample. The stirring speed control algorithm is based on the flow field simulation and concentration uniformity analysis of the soy milk sample during the stirring process to ensure the uniformity of the sample during the pretreatment process.

[0023] Furthermore, the spectrometer of the spectral analysis unit adopts a photonic crystal fiber spectroscopic structure, and its spectroscopic principle is based on the band theory of photonic crystals and the coupled mode theory of light. The lattice structure parameters of the photonic crystal fiber are determined by simulating and optimizing the light propagation characteristics of different spectral ranges using the finite element method to achieve high-precision spectroscopic effect. The photodetector inside the spectrometer adopts a superconducting nanowire detector, and its detection sensitivity is much higher than that of traditional photodetectors. The operating temperature of the superconducting nanowire detector is maintained by a sophisticated low-temperature refrigeration system. The temperature control algorithm of the low-temperature refrigeration system is based on the proportional-integral-differential PID control principle and is optimized in combination with the superconducting characteristics of the detector to ensure that the detector operates in the best working state. After the optical signal is efficiently converted into an electrical signal, it is amplified by a low-noise amplifier and then transmitted to the data processing and control unit.

[0024] Furthermore, the rotating shaft of the rotational viscometer of the rheological detection unit adopts a nano-composite coating material, which is a mixture of carbon nanotubes and polytetrafluoroethylene in a specific proportion. The mixing proportion is determined by experiments on the friction coefficient, corrosion resistance and biocompatibility of coating materials in different proportions in soy milk samples, so as to reduce the friction resistance between the rotating shaft and the soy milk sample, improve the measurement accuracy and extend the service life of the rotating shaft. The measurement range of the rotational viscometer is 0.5mPa·s-20000mPa·s, which can meet the viscosity measurement requirements of soy milk samples in a wider concentration range. The temperature sensor adopts a quantum dot fluorescence temperature sensor, whose fluorescence emission peak position moves linearly with temperature changes, and the temperature is determined by detecting the fluorescence emission peak position. The pressure sensor adopts a micro-electromechanical system pressure sensor, which is manufactured using a micro-machining process. These sensors are calibrated and installed in the rheological detection unit to ensure the accuracy of the collected data.

[0025] Furthermore, the fusion intelligent algorithm in the data processing and control unit adopts the data verification and update mechanism of the blockchain during operation. When the newly collected sample data is used for model update, the data is first encrypted and uploaded to the blockchain network. Multiple nodes in the network verify the data. Only the data that passes the verification is used for model update. For the update of the spectral concentration model, an incremental deep learning algorithm is adopted. Based on the original deep spectral feature extraction and conversion model, this algorithm adapts to the new data by dynamically adding new neural network layers or adjusting the parameters of existing layers. The addition of new layers or parameter adjustments are determined according to the distribution change characteristics of the data. For the update of the rheological concentration model, a fuzzy logic system optimization algorithm of reinforcement learning is adopted. This algorithm uses the reinforcement learning algorithm to adjust the rule base and membership function parameters of the fuzzy logic system according to the performance feedback of the model in actual applications to improve the adaptability and accuracy of the model. Through this data verification and update mechanism, the equipment can adapt to the concentration measurement requirements under different soybean raw materials and different refining process conditions, improve the versatility and accuracy of the equipment, and ensure the reliability and security of the data.

[0026] Furthermore, the communication between the data processing and control unit and the PLC control system adopts a communication architecture based on the integration of 5G and industrial Internet of Things. Under this architecture, data transmission adopts time division multiple access and frequency division multiple access hybrid multiplexing technology to meet the high real-time and large data volume transmission requirements. The data processing and control unit packages the refining concentration data and the equipment's operating status information according to a predetermined lightweight data protocol and sends them to the PLC control system. The PLC control system adjusts the motor speed and the number of grinding disc gaps of the refining equipment according to the received data to achieve precise control of the refining concentration. At the same time, the PLC control system can also feed back the parameters of other processing links to the data processing and control unit through the communication architecture. The data processing and control unit further optimizes the refining concentration measurement model according to these parameters to achieve coordinated optimization of the entire soy milk powder processing process. In addition, the communication architecture also has network security protection functions, and adopts an access control strategy based on identity authentication and encryption to prevent data from being illegally stolen or tampered with, thereby ensuring the security and stability of the soy milk powder processing process.

[0027] Furthermore, the touch screen of the display and alarm unit adopts a flexible organic light-emitting diode display screen, which is bendable, high-contrast and low-power. The operating interface of the touch screen is designed with artificial intelligence intelligent voice interaction and gesture recognition technology. The operator can conveniently set equipment parameters, query data, and export historical data through voice commands or simple gestures, thereby improving the convenience and intelligence of operation. The sound and light alarm device of the alarm signal includes a high-brightness color LED indicator light and a voice alarm with adjustable volume. When an alarm occurs, the LED indicator light displays different colors according to the alarm type, and the voice alarm prompts the alarm information with clear and loud voice, ensuring that the operator can detect abnormal conditions in time and take corresponding measures.

[0028] Furthermore, the internal circuit system of the device adopts a distributed intelligent control architecture, which is composed of multiple microcontroller units MCU. The MCUs are connected to each other through a high-speed serial bus. One master MCU is responsible for coordinating the data interaction and overall control logic of each functional module, such as receiving data from the spectral analysis unit and the rheological detection unit and distributing them to the corresponding data processing sub-modules, and transmitting the processing results to the display and alarm unit and the external PLC control system; multiple slave MCUs are specifically responsible for processing specific tasks such as spectral data acquisition and preliminary processing, and rheological data acquisition and preprocessing, so as to realize rapid parallel processing of data and improve the response speed and operation efficiency of the device. The device is also equipped with an intelligent calibration module, which is based on multi-sensor data fusion. The equipment uses a combination of technologies to automatically calibrate the wavelength accuracy of the spectral analysis unit and the sensor accuracy of the rheological detection unit on a regular basis. By comparing and analyzing the measurement data of the standard concentration sample with the preset standard value, the calibration coefficient of each sensor is calculated using the least squares algorithm, and the calibration parameters of the equipment are automatically updated to ensure the accuracy and stability of the equipment during long-term operation. In addition, the equipment has a self-learning optimization function. During operation, a large amount of actual measurement data is collected and compared with the corresponding manual detection or high-precision reference detection results. The concentration calculation model and various control algorithms in the data processing and control unit are continuously optimized using machine learning algorithms to adapt to different raw material characteristics, processing technology changes and equipment aging, thereby further improving the performance and reliability of the equipment.

[0029] Compared with the prior art, the soymilk powder processing grinding concentration measuring device has the following beneficial effects:

[0030] 1. The equipment adopts a dynamic adaptive flow algorithm through the sampling unit, which can adjust the sampling flow according to the real-time characteristics of the soy milk sample to ensure the representativeness and accuracy of the sample. The spectral analysis unit and the rheological detection unit use spectral detection technology and rheological property measurement technology respectively to conduct in-depth analysis of the composition and rheological properties of the soy milk sample. The data processing and control unit uses a fusion intelligent algorithm to efficiently process and analyze the collected data, and finally calculate the precise grinding concentration. This process not only improves the accuracy of the measurement, but also significantly improves the measurement efficiency, providing strong technical support for the soy milk powder processing process.

[0031] 2. The equipment has achieved seamless docking with the PLC control system, and realized real-time transmission and sharing of data through the communication architecture based on the integration of 5G and industrial Internet of Things. The data processing and control unit transmits the calculated refining concentration data and the equipment's operating status information to the PLC control system. The PLC control system accurately adjusts the motor speed and grinding disc gap parameters of the refining equipment based on these data, thereby realizing intelligent regulation of the refining concentration. At the same time, the PLC control system can also feed back the parameters of other processing links to the data processing and control unit. The data processing and control unit further optimizes the refining concentration measurement model based on these parameters, thereby realizing the coordinated optimization of the entire soymilk powder processing process. This process not only improves the automation level of the soymilk powder processing process, but also significantly improves the product quality and production efficiency.

[0032] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 This is a flow chart of the grinding concentration measurement equipment used in soy milk powder processing. DETAILED DESCRIPTION

[0035] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0036] Embodiment 1

[0037] This example describes the purchase of soybean raw materials from three different origins, namely, Northeast China, the United States and Brazil, in a large soybean milk powder production plant, and the determination of the grinding concentration of the soybean raw materials from different origins.

[0038] The sampling unit starts working. For the soybean raw materials from Northeast China, the near-infrared spectrum rapid pre-analysis model is used to estimate the proportion of each component in the soy milk. The model is built based on a large number of soybean sample data from different origins. For example, the estimated protein proportion is 36%, the carbohydrate proportion is 30%, and the dispersion index is calculated. Through laser particle size analysis of the soy milk sample, assuming that the total number of sampled particles is 5000, the diameter of each particle is measured, and the average particle diameter is calculated to be 15μm, and then the dispersion index is obtained. According to the dynamic adaptive flow algorithm, Among them, Q is the sampling flow rate mL / min, k1 is the basic coefficient of dynamic flow regulation, and adaptive correction is performed according to the real-time feedback data during the operation of the equipment. The correction logic optimizes the fitting of the relationship between the concentration measurement error and the flow rate after each sampling according to the least squares method. i is the estimated proportion of the i-th component in soy milk, which is obtained through the near-infrared spectroscopy rapid pre-analysis model of soybean raw materials. This model is constructed based on the near-infrared spectral data and component analysis data of a large number of soybean samples from different origins and varieties using the principal component regression algorithm. i is the flow impact factor corresponding to the i-th component, D t It is the dispersion index of the current soy milk, which is calculated by the laser particle size analysis data of the soy milk sample, reflecting the uniformity of the particle size distribution in the soy milk. The calculation formula is: where d j is the diameter of the jth particle (μm), is the average particle diameter, m is the total number of sampled particles, and the initial sampling flow rate is calculated to be 50 mL / min. During the sampling process, the dynamic flow adjustment basic coefficient is optimized by least squares fitting according to real-time feedback data, such as concentration measurement error, and the sampling flow rate is adjusted in real time.

[0039] The collected soy milk samples enter the spectrum analysis unit, and the scanning wavelength range and step length are determined according to the intelligent spectrum planning algorithm. The starting scanning wavelength and scanning step length are determined by analyzing the relative content of various chemical bonds in the soy milk and the current energy efficiency index of the equipment.

[0040] Among them, W start is the starting scanning wavelength (nm), W end is the end scanning wavelength (nm), Δλ is the scanning step length (nm), C k is the relative content of the kth chemical bond in soy milk, A k is the absorption sensitivity coefficient corresponding to the kth chemical bond, E rIt is the energy efficiency index of the current equipment, which is obtained in real time through the energy monitoring module inside the equipment. This index is used to optimize the scanning range and step length to reduce energy consumption while ensuring the accuracy of spectral analysis. A spectrometer with a photonic crystal fiber spectrometer structure is used for spectral analysis. Its lattice structure parameters have been optimized by the finite element method. The internal superconducting nanowire detector works under a precision low-temperature refrigeration system to convert optical signals into electrical signals, which are amplified by a low-noise amplifier and then transmitted.

[0041] When the soy milk sample is in the rheological detection unit, the rotating shaft of the rotational viscometer starts to rotate. The initial rotation speed is calculated based on the initial moisture content of the soybean raw material, the initial viscosity estimated value calculated by the protein structure characteristic parameters, the middle value of the target viscosity range, the current processing step time, the reference processing time, and the relative concentration and viscosity contribution factor of the rheologically influencing components in the soy milk. Assuming that the initial rotation speed is 30 rpm, the initial rotation speed is determined by the formula Where N is the rotation speed of the rotational viscometer (rpm), k2 is the basic constant for speed regulation, and the initial value is obtained by neural network fitting the experimental data of rheological properties of soy milk samples in different concentration ranges under different temperature and pressure conditions. The gradient descent algorithm is then used to correct the deviation between the real-time detection data and the predicted data. target is the middle value of the target viscosity range (mPa·s), which is set according to the soymilk powder processing requirements. init is the estimated value of initial viscosity (mPa·s), which is calculated by using the empirical formula based on the initial moisture content and protein structure characteristic parameters of soybean raw materials. process is the time of the current soybean milk powder processing step (min), T ref is the reference processing time, R l is the relative concentration of the first rheologically influencing component in soy milk, F l The viscosity contribution factor corresponding to the lth rheological influencing component. The temperature sensor and the pressure sensor measure the temperature and pressure of the soymilk sample in real time and transmit them together with the viscosity data.

[0042] After receiving the spectral and rheological data, the data processing and control unit uses a fusion intelligent algorithm to model and analyze. final =α×C spec-new +β×C rheo-new +γ×C int-new , where C final is the final calculated pulp concentration (g / mL), C spec-new To improve the concentration value calculated from spectral data, C rheo-newis the concentration value calculated based on the improved rheological data, α, β, and γ are weight coefficients, which are determined through a multi-stage dynamic analysis of the influence of spectral and rheological data on the accuracy of concentration measurement at different process stages during the processing of soymilk powder. In the early stage of refining, the weight of α is relatively large. As the processing proceeds, the weight of β gradually increases, and the weight of γ is determined according to the coordinated change trend of spectral and rheological data. For spectral data, it is processed by deep spectral feature extraction and conversion model, including wavelet transform, autoencoder neural network compression reconstruction, and finally the support vector machine regression model is used to calculate the concentration value. The kernel function parameters are optimized by genetic algorithm. For rheological data, a fuzzy logic reasoning system is constructed through principal component fusion analysis to calculate the concentration value. The weight coefficient is dynamically determined according to the influence of spectral and rheological data on the accuracy of concentration measurement at different process stages during the processing. The weight α of spectral data is relatively large in the early stage of refining. As the processing proceeds, the weight β of rheological data gradually increases. The final refining concentration is calculated to be 0.08 g / mL, and the data is transmitted to the PLC control system.

[0043] The touch screen shows that the measured refining concentration is 0.08g / mL and the equipment operating status information. The operator can query data and export historical data through voice commands or gestures. The preset normal refining concentration range is 0.06-0.1g / mL. If the concentration exceeds this range, the sound and light alarm device is activated, the color LED indicator turns red, and the voice alarm prompts the abnormal concentration.

[0044] Embodiment 2

[0045] This embodiment describes that in a soybean milk powder processing workshop, the pulping equipment is in continuous production, the processing process includes soaking, pulping, separation, and pulping steps, and the pulping concentration is measured at different stages of the processing.

[0046] After the soaking process is completed, the sampling unit collects soy milk samples from the grinding equipment. Among them, Q is the sampling flow rate mL / min, k1 is the basic coefficient of dynamic flow regulation, and adaptive correction is performed according to the real-time feedback data during the operation of the equipment. The correction logic optimizes the fitting of the relationship between the concentration measurement error and the flow rate after each sampling according to the least squares method. i is the estimated proportion of the i-th component in soy milk, which is obtained through the near-infrared spectroscopy rapid pre-analysis model of soybean raw materials. This model is constructed based on the near-infrared spectral data and component analysis data of a large number of soybean samples from different origins and varieties using the principal component regression algorithm. i is the flow impact factor corresponding to the i-th component, D t It is the dispersion index of the current soy milk, which is calculated by the laser particle size analysis data of the soy milk sample, reflecting the uniformity of the particle size distribution in the soy milk. The calculation formula is: where d j is the diameter of the jth particle (μm), is the average particle diameter, m is the total number of sampled particles. At this time, the proportion of soy milk components is estimated according to the characteristics of soybean raw materials and the state after soaking, such as the protein proportion is 38%. The dispersion index is calculated and the sampling flow rate is determined to be 45mL / min. As the processing reaches the middle stage of the refining process, the raw material characteristics and the state of soy milk change. When the sample is collected again, the sampling flow rate is dynamically adjusted. The dynamic flow adjustment basic coefficient is corrected through the real-time feedback of the concentration measurement error, and the new sampling flow rate is adjusted to 55mL / min.

[0047] In the sample analysis after the soaking process, according to the relative content changes of chemical bonds of soy milk components and the energy efficiency of the equipment, the intelligent spectrum planning algorithm is used. Among them, W start is the starting scanning wavelength (nm), W end is the end scanning wavelength (nm), Δλ is the scanning step length (nm), C k is the relative content of the kth chemical bond in soy milk, A k is the absorption sensitivity coefficient corresponding to the kth chemical bond, E r It is the energy efficiency index of the current equipment, which is obtained in real time through the energy monitoring module inside the equipment. This index is used to optimize the scanning range and step length to reduce energy consumption while ensuring the accuracy of spectral analysis. The new scanning wavelength range is determined to be 420-2400nm, and the scanning step length is 3nm. Spectral analysis is performed. When it comes to the pulping process, the soy milk temperature rises and the chemical bond state further changes. The spectrometer readjusts the scanning parameters according to the algorithm, the starting scanning wavelength is adjusted to 450nm, the ending scanning wavelength is adjusted to 2300nm, and the scanning step length is adjusted to 2.5nm, and the spectral data continues to be analyzed.

[0048] After the soaking process, the initial viscosity of the soy milk is low. Based on the estimated initial viscosity, the middle value of the target viscosity range, the current process time, and the reference processing time parameters, the rotational viscometer speed is calculated to be 25 rpm. The rheological properties of the soy milk are measured and the soy milk enters the boiling process. The temperature rise causes the viscosity to change. The viscosity sensor monitors the viscosity change in real time. The rotational viscometer speed is adjusted according to the real-time data and algorithm, assuming it is adjusted to 40 rpm. At the same time, the temperature sensor and pressure sensor continue to collect data and transmit it.

[0049] The data processing and control unit receives the spectral and rheological data of each process stage. After the soaking process, the pulping concentration is calculated according to the fusion intelligent algorithm. At this time, the spectral data weight α is relatively large, and the calculated concentration is 0.07g / mL. As the processing proceeds to the pulping process, the rheological data weight β increases, and the pulping concentration is recalculated to be 0.09g / mL. The data is transmitted to the PLC control system to adjust the parameters of the pulping process, such as heating temperature and pulping time.

[0050] The touch screen displays the pulping concentration data and equipment operating status in real time at each process stage. The concentration after the soaking process is 0.07g / mL, and the concentration in the pulping process is 0.09g / mL. If the concentration exceeds the preset normal range in a certain process, the sound and light alarm device will be activated, the LED indicator light will turn yellow, and the voice alarm will prompt the abnormal concentration. The operator can adjust the processing parameters in time.

[0051] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A grinding concentration measuring device for soybean milk powder processing, characterized in that: The system includes a sampling unit, a spectral analysis unit, a rheological detection unit, a data processing and control unit, and a display and alarm unit: The sampling unit is used to collect soybean milk samples from the grinding device, and has an adjustable sampling flow control device. The flow control device operates according to a dynamic adaptive flow algorithm, and the formula is: Among them, Q is the sampling flow rate mL / min, k1 is the basic coefficient of dynamic flow regulation, and adaptive correction is performed according to the real-time feedback data during the operation of the equipment. The correction logic optimizes the fitting of the relationship between the concentration measurement error and the flow rate after each sampling according to the least squares method. i is the estimated proportion of the i-th component in soy milk, which is obtained through the near-infrared spectroscopy rapid pre-analysis model of soybean raw materials. This model is constructed based on the near-infrared spectral data and component analysis data of a large number of soybean samples from different origins and varieties using the principal component regression algorithm. i is the flow impact factor corresponding to the i-th component, D t It is the dispersion index of the current soy milk, which is calculated by the laser particle size analysis data of the soy milk sample, reflecting the uniformity of the particle size distribution in the soy milk. The calculation formula is: where d j is the diameter of the jth particle (μm), is the average particle diameter, m is the total number of sampled particles; The spectrum analysis unit is connected to the sampling unit, and uses spectrum detection technology to perform spectrum analysis on the collected soybean milk samples. It has a unique adaptive spectrum scanning module, and its scanning wavelength range and step length are determined by the intelligent spectrum planning algorithm. The formula is as follows: Among them, W start is the starting scanning wavelength (nm), W end is the end scanning wavelength (nm), Δλ is the scanning step length (nm), C k is the relative content of the kth chemical bond in soy milk, A k is the absorption sensitivity coefficient corresponding to the kth chemical bond, E r It is the energy efficiency index of the current device, which is obtained in real time through the energy monitoring module inside the device. This index is used to optimize the scanning range and step length to reduce energy consumption while ensuring the accuracy of spectral analysis; The rheological detection unit: combines dynamic spectral imaging and spectral rheology to measure the rheological properties of the soy milk sample. The viscosity sensor of the rheological detection unit adopts a high-precision rotational viscometer, and its rotation speed is adjusted by the rheological feedback control algorithm. The formula is: Where N is the rotation speed of the rotational viscometer (rpm), k2 is the basic constant for speed regulation, and the initial value is obtained by neural network fitting the experimental data of rheological properties of soy milk samples in different concentration ranges under different temperature and pressure conditions. The gradient descent algorithm is then used to correct the deviation between the real-time detection data and the predicted data. target is the middle value of the target viscosity range (mPa·s), which is set according to the soymilk powder processing requirements. init is the estimated value of initial viscosity (mPa·s), which is calculated by using the empirical formula based on the initial moisture content and protein structure characteristic parameters of soybean raw materials. process is the time of the current soybean milk powder processing step (min), T ref is the reference processing time, R l is the relative concentration of the first rheologically influencing component in soy milk, F l is the viscosity contribution factor corresponding to the first rheological influencing component, the rheological detection unit further includes a temperature sensor and a pressure sensor, which are used to measure the temperature and pressure of the soymilk sample respectively, and the data collected by these sensors are transmitted to the data processing and control unit together with the viscosity data; The data processing and control unit receives the data collected by the spectral analysis unit and the rheological detection unit, and processes and analyzes the data. The unit uses a fusion intelligent algorithm to model and analyze the data. The algorithm is as follows: C final =α×C spec-new +β×C rheo-new +γ×C int-new Among them, C final is the final calculated pulp concentration (g / mL), C spec-new In order to improve the concentration value calculated from spectral data, a deep spectral feature extraction and conversion model is used. The model first performs wavelet transform on the original spectral data to extract spectral features at different scales, then uses an autoencoder neural network to compress and reconstruct the features, and finally inputs the reconstructed features into a support vector machine regression model to obtain the concentration value. The kernel function parameters in the model are optimized by a genetic algorithm to maximize the prediction accuracy of the model. rheo-new In order to obtain the concentration value calculated based on the improved rheological data, a concentration calculation model based on a fuzzy logic reasoning system is constructed by performing principal component fusion analysis on the viscosity, temperature and pressure rheological data. The input of the fuzzy logic system is the fused principal component, and the output is the concentration value. α, β and γ are weight coefficients, which are determined by a multi-stage dynamic analysis of the influence of spectral and rheological data on the accuracy of concentration measurement at different process stages during the processing of soybean milk powder. In the early stage of refining, the α weight is relatively large. As the processing proceeds, the β weight gradually increases, and the γ weight is determined according to the coordinated change trend of spectral and rheological data. The data processing and control unit can also communicate with other equipment in the soybean milk powder processing process, and transmit the calculated refining concentration data to the PLC control system to realize intelligent regulation of the entire processing process. The display and alarm unit is used to display the measured grinding concentration data and the operating status information of the equipment. The display and alarm unit adopts a touch display screen, and the alarm signal adopts an audible and visual alarm method. When the calculated grinding concentration exceeds the preset normal range, the microprocessor controls the alarm circuit to send out an audible and visual alarm signal. The preset normal range is set according to the quality standards and processing requirements of the soymilk powder product, and can be adjusted on the touch display screen.

2. The grinding concentration measuring device for soybean milk powder processing according to claim 1, characterized in that: The sampling unit also includes a sample pre-processing device, which includes a separation module based on the synergistic effect of centrifugal force and filtration, and its working principle is based on the centrifugal filtration algorithm: Among them, F sep is the separation force, m s is the mass of suspended solid particles in soy milk, obtained by preliminary mass analysis of soy milk samples, v is the centrifugal rotation speed, r is the centrifugal radius, k3 is the filtration synergy coefficient, P f In order to filter the pressure drop of the medium, the separation module can dynamically adjust the separation force according to the characteristics of the soy milk sample during the centrifugation process, so as to effectively remove the impurity particles and bubbles in the soy milk. At the same time, the sample pretreatment device is also equipped with an electromagnetic stirrer, and its stirring speed is controlled by a microprocessor according to the viscosity and particle distribution of the soy milk sample. The stirring speed control algorithm is based on the flow field simulation and concentration uniformity analysis of the soy milk sample during the stirring process.

3. The grinding concentration measuring device for soybean milk powder processing according to claim 1, characterized in that: The spectrometer of the spectral analysis unit adopts a photonic crystal fiber spectroscopic structure, and its spectroscopic principle is based on the band theory of photonic crystals and the coupled mode theory of light. The lattice structure parameters of the photonic crystal fiber are determined by simulating and optimizing the light propagation characteristics of different spectral ranges using the finite element method. The photodetector inside the spectrometer adopts a superconducting nanowire detector, and its detection sensitivity is much higher than that of traditional photodetectors. The operating temperature of the superconducting nanowire detector is maintained by a set of precise low-temperature refrigeration system. The temperature control algorithm of the low-temperature refrigeration system is based on the proportional-integral-differential PID control principle and is optimized in combination with the superconducting characteristics of the detector. After the optical signal is efficiently converted into an electrical signal, it is amplified by a low-noise amplifier and then transmitted to the data processing and control unit.

4. The grinding concentration measuring device for soybean milk powder processing according to claim 1, characterized in that: The rotating shaft of the rotational viscometer of the rheological detection unit adopts a nano-composite coating material, which is a mixture of carbon nanotubes and polytetrafluoroethylene in a specific ratio. The mixing ratio is determined by experiments on the friction coefficient, corrosion resistance and biocompatibility of coating materials with different ratios in soy milk samples. The temperature sensor adopts a quantum dot fluorescent temperature sensor, whose fluorescence emission peak position moves linearly with temperature changes, and the temperature is determined by detecting the fluorescence emission peak position. The pressure sensor adopts a micro-electromechanical system pressure sensor, which is manufactured using a micro-machining process. These sensors are all calibrated and installed in the rheological detection unit.

5. The grinding concentration measuring device for soybean milk powder processing according to claim 1, characterized in that: During operation, the fusion intelligent algorithm in the data processing and control unit adopts the data verification and update mechanism of the blockchain. When the newly collected sample data is used for model update, the data is first encrypted and uploaded to the blockchain network. Multiple nodes in the network verify the data. Only the data that passes the verification is used for model update. For the update of the spectral concentration model, an incremental deep learning algorithm is adopted. Based on the original deep spectral feature extraction and conversion model, this algorithm adapts to the new data by dynamically adding new neural network layers or adjusting the parameters of existing layers. The addition of new layers or the adjustment of parameters are determined according to the distribution change characteristics of the data. For the update of the rheological concentration model, a fuzzy logic system optimization algorithm of reinforcement learning is adopted. This algorithm uses the reinforcement learning algorithm to adjust the rule base and membership function parameters of the fuzzy logic system according to the performance feedback of the model in actual applications.

6. The grinding concentration measuring device for soybean milk powder processing according to claim 1, characterized in that: The communication between the data processing and control unit and the PLC control system adopts a communication architecture based on the integration of 5G and industrial Internet of Things. Under this architecture, data transmission adopts time division multiple access and frequency division multiple access hybrid multiplexing technology. The data processing and control unit packages the refining concentration data and the operating status information of the equipment according to a predetermined lightweight data protocol and sends them to the PLC control system. The PLC control system adjusts the motor speed and the number of grinding disc gaps of the refining equipment according to the received data. At the same time, the PLC control system can also feed back the parameters of other processing links to the data processing and control unit through the communication architecture. The data processing and control unit further optimizes the refining concentration measurement model according to these parameters to achieve coordinated optimization of the entire soymilk powder processing process. In addition, the communication architecture also has network security protection functions, and adopts an access control strategy based on identity authentication and encryption to prevent data from being illegally stolen or tampered with.

7. The grinding concentration measuring device for soybean milk powder processing according to claim 1, characterized in that: The touch display screen of the display and alarm unit adopts a flexible organic light-emitting diode display screen, which is bendable, high-contrast and low-power. The operating interface of the touch display screen is designed with artificial intelligence intelligent voice interaction and gesture recognition technology. The operator can conveniently set equipment parameters, query data, and export historical data through voice commands or simple gestures. The sound and light alarm device of the alarm signal includes a high-brightness color LED indicator light and a voice alarm with adjustable volume. When an alarm occurs, the LED indicator light displays different colors according to the alarm type, and the voice alarm prompts the alarm information with clear and loud voice.

8. The grinding concentration measuring device for soybean milk powder processing according to claim 1, characterized in that: The internal circuit system of the device adopts a distributed intelligent control architecture, which is composed of multiple microcontroller units MCU. The MCUs are connected to each other through a high-speed serial bus. One master MCU is responsible for coordinating the data interaction and overall control logic of each functional module, and multiple slave MCUs are responsible for processing spectral data acquisition and preliminary processing, rheological data acquisition and preprocessing specific tasks respectively. The device is also equipped with an intelligent calibration module. Based on multi-sensor data fusion technology, the module regularly automatically calibrates the wavelength accuracy of the spectral analysis unit and the sensor accuracy of the rheological detection unit. By comparing and analyzing the measurement data of the standard concentration sample with the preset standard value, the calibration coefficient of each sensor is calculated using the least squares algorithm, and the calibration parameters of the device are automatically updated. In addition, the device has a self-learning optimization function. During operation, by collecting a large amount of actual measurement data and comparing and learning with the corresponding manual detection or high-precision reference detection results, the concentration calculation model and various control algorithms in the data processing and control unit are continuously optimized using a machine learning algorithm to adapt to different raw material characteristics, processing technology changes and equipment aging.

Images