A data processing-based fermented quality prediction system for steamed bread

Abstract

The application discloses a kind of based on data processing's fermented quality prediction system of fermented bread, it is related to food intelligent manufacturing technical field, including, active module, fermentation acoustics-environment data are collected and preprocessed, and frequency domain feature conversion and Shannon entropy value calculation are carried out by acoustics analysis algorithm, obtain acoustic biological activity index time series;Modeling module, to acoustic biological activity index time series and the fermentation acoustics-environment data after pre-processing, dynamic viscosity and thermal conductivity are calculated simultaneously by inversion, obtain fermented dough dynamic physical property, and carry out three-dimensional dynamic modeling, obtain fermented bread fermentation simulation digital twin model;Prediction module, fermented state tensor and acoustic biological activity index time series are generated fermentation quality time series by using prediction algorithm fusion fermented bread fermentation;The application can accurately capture the biological activity change characteristics in fermentation process, improve the precision control of fermented quality for fermented bread processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of food intelligent manufacturing, and in particular to a fermented bread quality prediction system based on data processing. BACKGROUND

[0002] In the field of food processing intelligence, the quality monitoring technology of the fermentation process continues to evolve. Acoustic sensing technology collects bubble burst sound and microbial metabolic sound during the dough fermentation process, and combines with environmental temperature and humidity data to construct a multi-parameter evaluation system. This technology uses fast Fourier transform to extract the frequency domain features of the sound wave signal, analyzes the energy distribution of the biological activity frequency band, and identifies the dominant frequency components in different fermentation stages. At the same time, the randomness of the sound wave signal is quantified by calculating the Shannon entropy value, reflecting the dynamic changes of microbial metabolic activity.

[0003] The existing technology is mainly based on the static analysis of single physical field data (such as temperature, pH value), which is difficult to fully characterize the dynamic coupling relationship between the rheological properties and microbial metabolic activity in the fermentation process of fermented bread dough. In the processing of high-viscosity fermented bread dough, the existing method has obvious limitations. The static temperature monitoring cannot reflect the spatio-temporal evolution of the viscoelasticity of the dough, and the traditional acoustic signal processing method has insufficient feature extraction accuracy for the biological activity frequency band, resulting in a lag in the evaluation of microbial metabolic activity; the existing technology lacks in-depth analysis of the correlation mechanism between acoustic features and dough physical parameters, making it difficult to meet the time effectiveness and accuracy of the fermentation quality prediction model for high-standard industrial production needs. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a fermented bread quality prediction system based on data processing to solve the problem of inaccurate fermentation quality prediction caused by the inability to establish a nonlinear relationship between acoustic biological activity and dough rheological properties.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] The present application provides a fermented bread quality prediction system based on data processing, which comprises,

[0008] The activity module collects and preprocesses the fermentation acoustic-environmental data, and performs frequency domain feature conversion and Shannon entropy value calculation through acoustic analysis algorithm to obtain the time series sequence of acoustic biological activity index;

[0009] The modeling module is used for performing synchronous inversion calculation of dynamic viscosity and thermal conductivity on the time sequence of the acoustic biological activity index and the preprocessed fermentation acoustic-environment data, obtaining dynamic properties of the steamed bread dough, and performing three-dimensional dynamic modeling to obtain a steamed bread fermentation simulation digital twin model.

[0010] The simulation module is used for performing multi-physics field coupling calculation on the dynamic properties of the steamed bread dough by using the steamed bread fermentation simulation digital twin model, and outputting a steamed bread fermentation state tensor.

[0011] The prediction module is used for generating a fermentation quality time sequence by fusing the steamed bread fermentation state tensor and the time sequence of the acoustic biological activity index by using a prediction algorithm.

[0012] The compensation module is used for comparing the fermentation quality time sequence with a preset health threshold, obtaining a fermentation quality deviation, and generating a control instruction set through real-time compensation.

[0013] The evaluation module is used for performing spatiotemporal correlation analysis and abnormal marking on the control instruction set and the fermentation quality time sequence, generating a visual prediction curve atlas, and extracting key feature mapping to obtain a steamed bread fermentation quality report.

[0014] As a preferred scheme of the steamed bread fermentation quality prediction system based on data processing, the fermentation acoustic-environment data includes acoustic original signals and environment monitoring data.

[0015] The preprocessing includes signal denoising, timestamp alignment and data standardization.

[0016] As a preferred scheme of the steamed bread fermentation quality prediction system based on data processing, the frequency domain feature conversion and Shannon entropy value calculation are performed by using an acoustic analysis algorithm to obtain the time sequence of the acoustic biological activity index, and the specific steps are as follows.

[0017] Fast Fourier transform is performed on the preprocessed acoustic original signals to generate acoustic frequency domain feature data, biological activity frequency band relative energy data is extracted, the relative energy proportion is calculated, and the frequency band relative energy data is output.

[0018] Based on the frequency band relative energy data, the uncertainty measure of the acoustic original signals is calculated by using a Shannon entropy formula to generate a time sequence of acoustic Shannon entropy values.

[0019] The temperature drift compensation and humidity influence correction are performed on the time sequence of the acoustic Shannon entropy values by using the preprocessed environment monitoring data, the acoustic biological activity index data is output, and the time sequence of the acoustic biological activity index is generated by sorting according to the collection timestamps.

[0020] As a preferred embodiment of the data processing-based mochi bread fermentation quality prediction system of the present invention, the following steps are taken: The dynamic viscosity and thermal conductivity are simultaneously inverted and calculated from the time-series sequence of acoustic bioactivity indicators and the preprocessed fermentation acoustic-environment data to obtain the dynamic physical properties of the mochi dough.

[0021] Dynamic temperature and humidity compensation analysis is performed on the preprocessed environmental monitoring data to generate environmental temperature and humidity compensation parameters;

[0022] The time series of acoustic bioactivity indicators and environmental temperature and humidity compensation parameters are input into the rheology-thermal coupling inversion calculation architecture to perform dynamic viscosity calculation and thermal conductivity derivation, and output dynamic viscosity-thermal conductivity joint parameters.

[0023] By aligning the dynamic viscosity-thermal conductivity joint parameters over time and verifying data reliability, dynamic physical properties of mochi dough are generated.

[0024] As a preferred embodiment of the data processing-based mochi bread fermentation quality prediction system of the present invention, the specific steps for obtaining the digital twin model of mochi fermentation simulation are as follows:

[0025] Three-dimensional spatial meshing was performed on the dynamic physical properties of mochi dough to generate a spatial distribution mesh set of physical property parameters. Spatial mapping and temporal binding were performed with the dynamic viscosity-thermal conductivity joint parameters to obtain multiphysics boundary condition data.

[0026] The multiphysics boundary condition data is coupled and solved to generate dynamic coupled multiphysics solution results. The time series is then integrated and the interactive interface is bound to generate a digital twin model for mochi fermentation simulation.

[0027] As a preferred embodiment of the data processing-based mochi bread fermentation quality prediction system of the present invention, the following steps are taken: A digital twin model for mochi fermentation simulation is used to perform multi-physics field coupling calculations on the dynamic properties of mochi dough, outputting a mochi fermentation state tensor.

[0028] The dynamic physical properties of the mochi dough are input into the digital twin model for mochi fermentation simulation. The dynamic physical property parameters of the mochi dough are loaded and the digital twin model for mochi fermentation simulation is initialized.

[0029] Using the initialized digital twin model of mochi fermentation simulation, coupled simulation calculations of fluid motion, gas diffusion and temperature distribution are performed on the dynamic physical properties of mochi dough to generate multiphysics dynamic simulation results.

[0030] The results of the multiphysics dynamic simulation are converted into a four-dimensional spatiotemporal tensor format, and the mochi fermentation state tensor is output.

[0031] As a preferred scheme of the data processing based fermented bread quality prediction system, the fermented bread quality time series is generated by adopting a prediction algorithm to fuse the fermented bread state tensor and the acoustic biological activity index time series, and the specific steps are as follows,

[0032] The time stamp synchronization and space grid mapping are performed on the fermented bread state tensor and the acoustic biological activity index time series to generate the time-space synchronous fusion data.

[0033] The cross-modal feature extraction and weight distribution are performed on the time-space synchronous fusion data, and the dynamic correlation between the physical field features and the biological activity features is strengthened to generate the fusion multi-modal feature matrix.

[0034] The fermented bread quality time series is generated by inputting the fusion multi-modal feature matrix into the time series prediction algorithm, analyzing the time series evolution law of the multi-modal features in time steps, and generating the fermented bread quality time series.

[0035] As a preferred scheme of the data processing based fermented bread quality prediction system, the physical field features include velocity field gradient, concentration field diffusion coefficient and temperature field heat flux.

[0036] The biological activity features include acoustic entropy value fluctuation and microbial metabolic intensity.

[0037] As a preferred scheme of the data processing based fermented bread quality prediction system, the fermented bread quality time series is compared with the preset health threshold to obtain the fermented bread quality deviation and generate the control instruction set through real-time compensation, and the specific steps are as follows,

[0038] The multi-dimensional feature analysis is performed on the historical fermented bread quality time series to extract the quality parameter distribution law under similar working conditions, generate the fermented bread quality feature clustering, and perform dynamic threshold scaling calculation to obtain the health threshold.

[0039] The fermented bread quality time series and the health threshold are compared and calculated point by point on the time axis to obtain the multi-dimensional quality deviation value, generate the fermented bread quality deviation, and perform dynamic compensation calculation and adjustment parameter mapping to generate the control instruction set.

[0040] As a preferred scheme of the data processing based fermented bread quality prediction system, the fermented bread quality prediction curve atlas is generated by performing time-space correlation analysis and abnormal marking on the control instruction set and the fermented bread quality time series, and the fermented bread quality quality report is obtained by extracting the key feature mapping, and the specific steps are as follows,

[0041] The time stamp synchronization matching and causal correlation modeling are performed on the control instruction set and the fermented bread quality time series to generate the instruction-quality joint analysis data.

[0042] The multi-dimensional dynamic rendering and interactive marking of the instruction-quality joint analysis data are carried out, visual prediction curve atlas is obtained, key feature analysis and multi-dimensional quality mapping are carried out, and a fermentation quality quality report of the steamed bread is obtained.

[0043] The present application has the beneficial effects that: by generating the time sequence of the acoustic biological activity index, the dynamic monitoring and quantitative characterization of the microbial metabolic activity in the fermentation process are realized, the change characteristics of the biological activity in the fermentation process can be accurately captured, the accurate control of the fermentation quality for the steamed bread processing is improved, and the qualified rate of the fermentation quality for the steamed bread processing is effectively improved; by the synchronous inversion calculation of the dynamic viscosity and the thermal conductivity, a synergistic analysis model of the dough rheological properties and the thermodynamic properties is established, the fermentation state of the dough is more accurately judged, and abnormal conditions are found in time; the stability of the product quality is significantly improved, and the quality problems caused by uneven fermentation are avoided, so that the production process is more controllable, technical support is provided for formulating more scientific process parameters, which is helpful to realize standardized production, ensures the fermentation quality of the steamed bread, and improves the production efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0045] Fig. 1 It is a schematic diagram of the steamed bread fermentation quality prediction system based on data processing.

[0046] Fig. 2 It is a flowchart for generating the time sequence of the acoustic biological activity index.

[0047] Fig. 3 It is a flowchart for generating the steamed bread fermentation state tensor.

[0048] Fig. 4 It is a flowchart for outputting the steamed bread fermentation quality quality report. DETAILED DESCRIPTION

[0049] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0050] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.

[0051] Second, the "one embodiment" or "an embodiment" described herein as including a particular implementation as part of the present application can include a particular feature, structure, or characteristic. However, each of the features, structures, or characteristics can be implemented in other embodiments of the application without being exclusive or selective to other embodiments.

[0052] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a data processing-based fermented bread quality prediction system, including the following steps:

[0053] An activity module collects and pre-processes fermentation acoustic-environmental data.

[0054] The fermentation acoustic-environmental data includes acoustic raw signals and environmental monitoring data.

[0055] The pre-processing includes signal denoising, timestamp alignment, and data standardization.

[0056] Specifically, signal denoising: the acoustic raw signals are decomposed using a sym4 wavelet basis function for 4 layers. The reconstructed signal retains the energy of the exemplary 1500-45000 Hz frequency band; the environmental monitoring data is input into a median filter, with a window width of exemplary 5 consecutive data points to eliminate pulse-type interference;

[0057] Timestamp alignment: the acoustic raw signal collection time axis is used as a reference, and the environmental monitoring data original timestamp interval is exemplary 1 second. The environmental monitoring data is resampled using a linear interpolation algorithm and aligned to a time axis with an interval of, for example, 10 milliseconds;

[0058] Data standardization: the amplitude of the acoustic raw signal is dynamically normalized for, for example, 10-second segments; the temperature value of the environmental monitoring data is converted to the Celsius scale, and the humidity value is converted to a percentage. The abnormal value detection range is: temperature exemplary 0-60 degrees Celsius, humidity exemplary 20%-100%, and the abnormal value is replaced by the average value of the previous and subsequent 5 valid data points.

[0059] Frequency domain feature conversion and Shannon entropy value calculation are performed through acoustic analysis algorithm to obtain the time series sequence of acoustic biological activity index.

[0060] Fast Fourier transform is performed on the pre-processed acoustic raw signals to generate acoustic frequency domain feature data, extract biological activity frequency band relative energy data, calculate the relative energy proportion, and output the frequency band relative energy data.

[0061] Specifically, the pre-processed acoustic original signal is input into a fast Fourier transform processor to perform a discrete Fourier transform operation to obtain acoustic complex spectrum data, the acoustic original signal is processed by frame with a fixed length, each frame of acoustic original signal is weighted by applying a Hanning window function, the modulus value operation is performed on the transformed acoustic complex spectrum data to obtain the amplitude information of each frequency point of the acoustic spectrum data. In the example 1500-4500 Hz bioactivity frequency band range, the amplitude square value of each frequency point of the acoustic spectrum data is accumulated as the bioactivity frequency band energy representation;

[0062] At the same time, the cumulative amount of the amplitude square value of the acoustic spectrum data in the example 20-20000 Hz full frequency band is calculated as a reference total energy value, the relative energy proportion is obtained by the proportional relationship between the bioactivity frequency band energy representation and the reference total energy value, and the frequency band relative energy data containing the time stamp and the bioactivity frequency band energy distribution characteristics is generated. The bioactivity frequency band range boundary value is determined according to the acoustic original signal of the fermented dough, and the frequency band relative energy data is output;

[0063] It should be noted that the expression for performing the discrete Fourier transform operation to obtain the acoustic complex spectrum data is:

[0064] ;

[0065] Wherein, is the complex spectrum data of the i th frequency point, is the total number of sampling points of the discrete signal, is the index variable of the total number of sampling points of the discrete signal, is the amplitude of the i th sampling point of the time domain discrete signal, is the base of natural logarithm, is the imaginary unit, is the frequency point index. Based on the frequency band relative energy data, the uncertainty measure of the acoustic original signal is calculated by applying the Shannon entropy formula to generate the acoustic Shannon entropy value time sequence.

[0066] Based on the frequency band relative energy data, the uncertainty measure of the acoustic original signal is calculated by applying the Shannon entropy formula to generate the acoustic Shannon entropy value time sequence.

[0067] ​Specifically, the relative energy data of frequency bands is input into Shannon entropy. The relative energy data of the bioactive frequency bands corresponding to each timestamp is extracted as a probability distribution. This bioactive frequency band relative energy data is normalized to probability values, ensuring that the sum of the energy proportions of all bioactive frequency bands is 1. The Shannon entropy value at each time point is calculated using the Shannon entropy formula. The resulting Shannon entropy values ​​are arranged chronologically to generate an acoustic Shannon entropy time series. The temporal resolution of the acoustic Shannon entropy time series is consistent with that of the relative energy data of the frequency bands. The Shannon entropy calculation process is executed independently for each timestamp to ensure temporal continuity. The output format of the acoustic Shannon entropy time series includes two fields: timestamp and entropy value. The Shannon entropy calculation range is limited to the 1500-4500 Hz range defined by the relative energy data of the bioactive frequency bands, outputting the acoustic Shannon entropy time series.

[0068] It should be noted that the expression for calculating the Shannon entropy value at each time point using the Shannon entropy formula is as follows:

[0069] ;

[0070] in, It is the Shannon entropy value. It is a bioactive frequency band. It is a sequence index of the bioactive frequency band. yes The normalized energy percentage probability of each bioactive frequency band. It is the base of the logarithm.

[0071] Using preprocessed environmental monitoring data, temperature drift compensation and humidity correction are applied to the acoustic Shannon entropy time series, outputting acoustic biological activity index data, which are then sorted by collection timestamp to generate a time series of acoustic biological activity indexes.

[0072] Specifically, the preprocessed environmental monitoring data is precisely matched with the acoustic Shannon entropy time series to align the sampling time points. Temperature and humidity values ​​are read from the environmental monitoring data, and an example table for temperature drift compensation coefficients is created, where, for example, a 1-degree Celsius increase in temperature corresponds to a 0.02-bit compensation in acoustic Shannon entropy. A lookup table for humidity effect correction coefficients is also created, where, for example, a 1% increase in humidity corresponds to a 0.005-bit correction in acoustic Shannon entropy.

[0073] For each Shannon entropy data point in the acoustic Shannon entropy time series, compensation is calculated by querying the temperature compensation coefficient and humidity correction coefficient based on the corresponding timestamp's environmental monitoring temperature and humidity values. The compensated acoustic Shannon entropy values ​​are retained to four decimal places, generating corrected acoustic bioactivity index data. The corrected acoustic bioactivity index data are sorted according to the original acquisition timestamp to ensure the continuity of the time series, and the acoustic bioactivity index time series is output.

[0074] The modeling module performs synchronous inversion calculation of dynamic viscosity and thermal conductivity on the pre-processed fermentation acoustics-environment data and the time series of acoustic biological activity indicators, to obtain the dynamic properties of the sweet potato dough.

[0075] The pre-processed environmental monitoring data is analyzed for temperature and humidity dynamic compensation to generate environmental temperature and humidity compensation parameters.

[0076] Specifically, the temperature value (in Celsius) and humidity value (in percentage) data columns of the pre-processed environmental monitoring data are extracted, and a temperature-entropy compensation relationship table is exemplarily established to define the compensation coefficients corresponding to different temperature intervals, for example, the compensation coefficient for the temperature interval of 25-30 degrees Celsius is 0.015 bits / degree Celsius. A humidity-entropy correction relationship table is exemplarily established to define the correction coefficients corresponding to different humidity intervals, for example, the correction coefficient for the humidity interval of 70-80% is 0.008 bits / percentage;

[0077] The temperature value data column is processed by exemplarily 5-point moving average filtering, and the humidity value data column is processed by exemplarily 3-point median filtering. The temperature and humidity data after median filtering are time-stamped matched and aligned with the acoustic Shannon entropy value time series. The temperature compensation amount is obtained by querying the temperature-entropy compensation relationship table, and the humidity correction amount is obtained by querying the humidity-entropy correction relationship table. The temperature compensation amount and humidity correction amount are combined to generate the environmental temperature and humidity compensation parameters.

[0078] The rheological-thermal coupling inversion calculation architecture is inputted with the time series of acoustic biological activity indicators and the environmental temperature and humidity compensation parameters, to perform dynamic viscosity calculation and thermal conductivity characteristic derivation, and output the dynamic viscosity-thermal conductivity joint parameters.

[0079] Specifically, the time series of acoustic biological activity indicators and the environmental temperature and humidity compensation parameters are inputted into the rheological-thermal coupling inversion calculation architecture for time-stamped accurate matching and alignment. The Shannon entropy value in the time series of acoustic biological activity indicators and the temperature compensation amount and humidity correction amount in the environmental temperature and humidity compensation parameters are read;

[0080] For example, the initial viscosity decreases by 1.5 mPa·s for each increase of 0.1 bit in the Shannon entropy value. The initial viscosity estimate is dynamically adjusted in combination with the temperature compensation amount in the ambient temperature and humidity compensation parameters, for example, the viscosity correction coefficient is 1.2 mPa·s for each increase of 0.015 bit in the temperature compensation amount. The initial thermal conductivity increases by 0.03 W / (m·K) for each decrease of 0.1 bit in the example entropy value. The initial thermal conductivity estimate is dynamically adjusted in combination with the humidity correction amount in the ambient temperature and humidity compensation parameters, for example, the thermal conductivity correction coefficient is 0.05 W / (m·K) for each increase of 0.008 bit in the humidity correction amount. The adjusted viscosity value and thermal conductivity value are bound by the timestamp to generate a dynamic viscosity-thermal conductivity joint parameter.

[0081] The dynamic viscosity-thermal conductivity joint parameter is time-dimensionally aligned and data reliability verified to generate the dynamic physical properties of the sticky dough.

[0082] Specifically, the dynamic viscosity-thermal conductivity joint parameter is checked for timestamp continuity, and missing dynamic viscosity-thermal conductivity joint parameter points are identified and marked. Linear interpolation method is used to fill in the missing dynamic viscosity-thermal conductivity joint parameter to ensure the integrity of the time series. The sliding window standard deviation analysis is performed on the viscosity value, and the viscosity value exceeding the example deviation range is removed. The moving average consistency analysis is performed on the thermal conductivity value, and the thermal conductivity value with a difference exceeding the example allowed value is marked. The cross-correlation coefficient is calculated for the viscosity value and the thermal conductivity value, and the cross-correlation coefficient time period below the cross-correlation coefficient threshold is reprocessed. The viscosity value and the thermal conductivity value that pass the verification are integrated according to the timestamp to form the dynamic physical properties of the sticky dough;

[0083] It should be noted that the expression for calculating the cross-correlation coefficient of the viscosity value and the thermal conductivity value is:

[0084] ;

[0085] Wherein, is the cross-correlation coefficient of the viscosity value and the thermal conductivity value at a time delay , is the expectation operation, is the time delay parameter, is the viscosity measurement value at time , is the mean value of the viscosity value, is the mean value of the thermal conductivity value, is the standard deviation of the viscosity value, is the standard deviation of the thermal conductivity value.

[0086] It should be noted that the setting process of the correlation number threshold is: based on the correlation coefficient of viscosity and thermal conductivity value, taking the exemplary 5th percentile of the correlation coefficient of viscosity and thermal conductivity value under normal working condition as the initial threshold lower limit; the abnormal recognition rate of the initial threshold is verified through actual fermentation test, and the final correlation number threshold is determined when the recognition rate reaches more than 92%, which is exemplary 0.75.

[0087] A simulation digital twin model of fermented mian-shao is obtained.

[0088] The dynamic physical properties of mian-shao dough are divided into a three-dimensional space grid, a physical property parameter space distribution grid set is generated, and the dynamic viscosity-thermal conductivity combined parameters are spatially mapped and time-sequentially bound to obtain multi-physical field boundary condition data.

[0089] Specifically, the dynamic physical properties of mian-shao dough are divided into a three-dimensional space grid by octree structure, generating voxel units with an exemplary size of 1mm³. The center points of each voxel unit are defined as octree grid space nodes, and the number of octree grid space nodes is automatically determined according to the volume of the dough. The viscosity and thermal conductivity values in the dynamic physical properties of mian-shao dough are read, and a trilinear interpolation operation is performed to map the dynamic physical properties of mian-shao dough to the octree grid space nodes. The interpolation weight is calculated based on the exemplary Euclidean distance of the octree grid space node mian-shao dough dynamic physical properties, and the distance constraint is set to exemplary 2mm;

[0090] The correspondence between the time stamp and the octree grid space node is established, and each octree grid space node records the dynamic viscosity-thermal conductivity combined parameters of the current time stamp. The spatially mapped octree grid space node data is arranged in time sequence to form a mian-shao dough dynamic physical property space distribution grid. The dynamic viscosity-thermal conductivity combined parameters are associated with the spatial grid by means of the octree grid space node index to obtain multi-physical field boundary condition data.

[0091] The multi-physical field boundary condition data is coupled and calculated to generate multi-physical field dynamic coupling calculation results, which are time-sequentially integrated and bound to an interactive interface to generate a simulation digital twin model of fermented mian-shao.

[0092] Specifically, the multi-physical field boundary condition data is input into a coupling calculation engine to perform fluid-thermal conduction coupling calculation. The finite volume method is used to discretize the control equation, and the conservation laws of fluid motion and heat conduction are discretized to provide mathematical basis for numerical solution of Navier-Stokes equation and energy equation. The spatial discretization is based on the topology structure of octree grid space nodes, and the time step is set to exemplary 0.1 seconds.

[0093] The multi-physics field state parameters of each octree grid space grid node are obtained by bringing the viscosity value and the thermal conductivity value into the Navier-Stokes equation and the energy equation and solving at each time step. The coupling calculation iteration convergence condition is set to be less than 1e-5. After the physical field is solved, the multi-physics field dynamic coupling calculation result is obtained, and the fluid motion field data, i.e. the velocity field data, the pressure field data and the viscosity distribution data, the heat conduction field data, i.e. the temperature field data, the heat flux density data and the thermal conductivity distribution data, and the coupling effect data, i.e. the natural convection velocity data, the viscous heat generation distribution data and the thermal expansion pressure data, are extracted from the multi-physics field dynamic coupling calculation result. The four-dimensional data set is integrated according to the time stamp. The input channel and the control command output channel of the environment monitoring sensor, i.e. the temperature sensor, the humidity sensor and the acoustic sensor, are bound, and the real-time data interaction interface refresh frequency matches the example 10Hz sampling rate. The multi-physics field dynamic coupling calculation result and the interface configuration information are merged to generate a fermented fermented simulation digital twin model containing complete space-time evolution data.

[0094] The simulation module uses the fermented fermented simulation digital twin model to perform multi-physics field coupling calculation on the dynamic properties of the fermented fermented simulation digital twin model, and outputs the fermented fermented state tensor.

[0095] The dynamic properties of the fermented fermented simulation digital twin model are input into the fermented fermented simulation digital twin model, and the dynamic properties of the fermented fermented simulation digital twin model are loaded and initialized.

[0096] Specifically, the dynamic properties of the fermented fermented simulation digital twin model are input into the fermented fermented simulation digital twin model, the viscosity value column and the thermal conductivity value column in the dynamic properties of the fermented fermented simulation digital twin model are read, the viscosity value column is mapped to the fluid mechanics parameter loading port of the fermented fermented simulation digital twin model, and the thermal conductivity value column is mapped to the thermal conduction parameter loading port.

[0097] The dynamic properties of the fermented fermented simulation digital twin model are input into the fermented fermented simulation digital twin model, and the dynamic properties of the fermented fermented simulation digital twin model are loaded and initialized.

[0098] The dynamic properties of the steamed bun dough are coupled and simulated by fluid motion, gas diffusion and temperature distribution through the initialized steamed bun fermentation simulation digital twin model to generate multi-physical field dynamic simulation results.

[0099] Specifically, the initialized steamed bun fermentation simulation digital twin model starts the coupled simulation calculation process and reads the dynamic physical property parameters of the steamed bun dough at the current time step, i.e. the viscosity value column and the thermal conductivity value column. In the example 0.1 second time step, the three-dimensional flow velocity data of each steamed bun fermentation simulation digital twin model space grid node is obtained by solving the Navier-Stokes equation, the gas concentration data is obtained by solving the Fick diffusion equation, and the temperature distribution data is obtained by solving the Fourier heat conduction equation;

[0100] The dynamic physical property parameters of the steamed bun dough are called in real time during the calculation process, the viscosity value column is used for the calculation of the viscous force term of the fluid motion equation, and the thermal conductivity value column is used for the calculation of the thermal resistance term of the heat conduction equation. A bidirectional coupling relationship is established between each physical field, the fluid motion affects the gas diffusion rate, the gas diffusion changes the dough density distribution, and the temperature gradient drives natural convection. Each time step is iteratively calculated until the residual is less than the example 1e-5 convergence value, and the three-dimensional flow velocity data, gas concentration data and temperature distribution data of each steamed bun fermentation simulation digital twin model space grid node at the current time are output to obtain the multi-physical field dynamic simulation results;

[0101] It should be noted that the expression for obtaining the three-dimensional flow velocity data of each steamed bun fermentation simulation digital twin model space grid node by solving the Navier-Stokes equation is:

[0102] ;

[0103] Wherein, is the three-dimensional flow velocity data at the spatial position and the time , is the first principal axis direction, is the second principal axis direction, is the third principal axis direction, is the spatial position of the steamed bun fermentation simulation digital twin model space grid node, is the simulation time variable;

[0104] The expression for obtaining the gas concentration data by solving the Fick diffusion equation is:

[0105] ;

[0106] Wherein, is the three-dimensional flow velocity data at the spatial position and the time gas concentration data at the spatial position and the time , gas concentration data at the spatial position and the time , gas diffusion coefficient exemplary value 1.0x10⁻ 9 ~ 5.0x10⁻ 9 m² / s, Laplacian operator, diffusion ability of the gas in the steamed bread dough,

[0107] auxiliary variable in the time integral operation;

[0108] The expression of the temperature distribution data obtained by solving the Fourier heat conduction equation is:

[0109] wherein, temperature distribution data at the spatial position and the time , temperature distribution data at the spatial position and the time , heat conduction coefficient exemplary value 0.2~0.5 W / (m·K), density of the steamed bread dough, specific heat capacity of the steamed bread dough.

[0110] The multi-physics dynamic simulation results are converted into a four-dimensional space-time tensor format, and the steamed bread fermentation state tensor is output.

[0111] Specifically, the spatial grid nodes, timestamp data, three-dimensional flow velocity data, gas concentration data and temperature distribution data in the multi-physics dynamic simulation results are extracted; a four-dimensional tensor structure is established: the first dimension is defined as the time index axis, arranged in the order of simulation time steps, and the time step is maintained at an example of 0.1 seconds; the second dimension is defined as the spatial X-axis index, corresponding to the discrete coordinate serial number of the octree spatial grid node in the X direction; the third dimension is defined as the spatial Y-axis index, corresponding to the discrete coordinate serial number of the octree spatial grid node in the Y direction; and the fourth dimension is defined as the spatial Z-axis index, corresponding to the discrete coordinate serial number of the octree spatial grid node in the Z direction;

[0112] The multi-physics data of the spatial grid node position in each multi-physics dynamic simulation result is filled into the tensor in a fixed channel order: channels 1 to 3 store the X / Y / Z components of the three-dimensional flow rate data, channel 4 stores the gas concentration data, and channel 5 stores the temperature distribution data. The integrity verification is performed twice: the first time, the dimension of the physical field data is checked to match the number of spatial grid nodes in the multi-physics dynamic simulation result, and the spatial grid node missing rate in the multi-physics dynamic simulation result is set to an example 0.001%; the second time, the time step continuity is checked, and the continuous missing time step number threshold is set to an example 3 steps. If the missing time step number threshold is exceeded, linear interpolation is performed to complete the output of the fermented state tensor of the fermented soybean;

[0113] It should be noted that the continuous missing time step number threshold setting process: scanning the time sequence, recording the length of the continuous blank time interval, and when the continuous blank step number is an example ≥ 3, the continuous missing time step number threshold is obtained, and linearly interpolated data is inserted between the start and end points of the missing interval.

[0114] The prediction module adopts a prediction algorithm to fuse the fermented state tensor of the fermented soybean and the time sequence of the acoustic biological activity index to generate a fermentation quality time sequence.

[0115] The time stamp synchronization and spatial grid mapping of the fermented state tensor of the fermented soybean and the time sequence of the acoustic biological activity index are performed to generate spatiotemporally synchronized fusion data.

[0116] Specifically, the time step sequence of the fermented state tensor of the fermented soybean is extracted, a time window is created based on the time stamp of the fermented state tensor of the fermented soybean, and all active data points in the acoustic biological activity index time sequence that fall within the window are captured; when there are active data points in the window, the arithmetic mean of the acoustic biological activity index is obtained, and when there are no active data points, an invalid value is marked;

[0117] The three-dimensional installation coordinates of the acoustic sensor in the fermentation tank are obtained; the Euclidean distance between the acoustic sensor coordinates and all octree grid spatial nodes of the fermented state tensor of the fermented soybean is calculated; the octree grid spatial node with the smallest Euclidean distance is found; when the minimum distance is ≤ an example 2 mm, the acoustic biological activity index is bound to the octree grid spatial node; when the minimum distance is > an example 2 mm, the second closest octree grid spatial node is searched within a radius of an example 5 mm and bound; when no node can be matched, a new octree grid spatial node is activated at the sensor coordinates;

[0118] The processed acoustic biological activity index is added to the fermented state tensor of the fermented soybean, the bound octree grid spatial node position is written with an actual value, and the unbound octree grid spatial node is filled with an example -1 marker to obtain spatiotemporally synchronized fusion data;

[0119] It should be noted that the expression for calculating the Euclidean distance between the acoustic sensor coordinates and all octree grid space nodes of the fermentation state tensor of the fermented soy is:

[0120]

[0121] wherein, is the Euclidean distance between the acoustic sensor coordinates and all octree grid space nodes of the fermentation state tensor of the fermented soy, is the three-dimensional space coordinates of the acoustic sensor, is the three-dimensional space coordinates of all octree grid space nodes of the fermentation state tensor of the fermented soy.

[0122] Cross-modal feature extraction and weight allocation are performed on the spatiotemporal synchronous fusion data, and the dynamic correlation between the physical field features and the biological activity features is strengthened to generate a fusion multi-modal feature matrix.

[0123] Specifically, the physical field features and the biological activity features are separated from the spatiotemporal synchronous fusion data, the physical field features include three-dimensional flow rate data, gas concentration data, and temperature distribution data, and the biological activity features are acoustic biological activity index data. Standardization processing (mean 0, variance 1) is performed on the physical field features, and the original physical field features are retained for subsequent correlation analysis. Short-time Fourier transform is performed on the biological activity features to extract energy spectrum features in an exemplary 0-1000Hz frequency band, and biological activity frequency domain features are obtained;

[0124] The dynamic correlation weight of the physical field features and the biological activity features is calculated: within an exemplary 5-second time window width, the absolute value of the Pearson correlation coefficient of the physical field features (three-dimensional flow rate, gas concentration, temperature) and the biological activity frequency domain features is directly calculated, and if it exceeds an exemplary 0.7, it is marked as a strongly correlated feature pair. The strongly correlated feature pair is assigned a weight value, and the physical field feature and the biological activity feature weight are normalized to an exemplary 0-1 range;

[0125] The weighted physical field features and biological activity frequency domain features are arranged in time step order to construct an initial feature matrix, the initial feature matrix row corresponds to the time step, and the column contains the physical field feature projection, the biological activity frequency spectrum, and the dynamic weight three parts of data, and a fusion multi-modal feature matrix is obtained.

[0126] The fusion multi-modal feature matrix is input into a time series prediction algorithm to analyze the time sequence evolution law of the multi-modal features in time step expansion, and a fermentation quality time series is generated.

[0127] ​​​​​Specifically, the fusion multi-modal feature matrix is input into a time series prediction algorithm, the fusion multi-modal feature matrix is read at an exemplary 0.1 second step, the fusion multi-modal feature matrix at each time step is unfolded in an exemplary 30-step (3 seconds) window, time series evolution indicators of physical field features (three-dimensional flow velocity linear regression slope, gas concentration moving average change rate, and temperature range and standard deviation) and biological activity frequency domain features (main period and amplitude extracted by autocorrelation analysis of acoustic biological activity energy spectrum) in the window are obtained, the time series evolution indicators are arranged in time step order to generate a fermentation quality time series containing physical field trend, biological activity period, and weight stability, missing values are handled by forward and backward filling (an exemplary maximum interval of 5 steps), and a fermentation quality time series with an exemplary 0.1 second time resolution is output.

[0128] The physical field features include velocity field gradient, concentration field diffusion coefficient, and temperature field heat flux.

[0129] The biological activity features include acoustic entropy value fluctuation and microbial metabolic intensity.

[0130] The compensation module compares the fermentation quality time series with a preset health threshold to obtain a fermentation quality deviation and generates a control instruction set through real-time compensation.

[0131] The historical fermentation quality time series is analyzed for multi-dimensional features, the quality parameter distribution law under similar working conditions is extracted, fermentation quality feature clusters are generated, and dynamic threshold scaling calculation is performed to obtain a health threshold.

[0132] Specifically, historical fermentation quality time series data is read, and time series evolution indicators are extracted; after standardization, K-means clustering is performed based on Euclidean distance, an exemplary 3-5 classes are obtained, iteration is performed 300 times, the quality parameter distribution of each cluster is obtained; the Mahalanobis distance is calculated based on the cluster center, an exemplary 95% confidence interval is set as the health threshold range, and when the Mahalanobis distance exceeds an exemplary 3.0, it is marked as abnormal; the cluster center and the health threshold range are updated every 24 hours, the cluster quality (contour coefficient > 0.6) and the health threshold range effectiveness (abnormal detection accuracy > 90%) are verified, and an exemplary 3.0 health threshold is output.

[0133] It should be noted that the expression for calculating the Mahalanobis distance based on the cluster center is:

[0134] ;

[0135] wherein, is the Mahalanobis distance, is the sample feature vector, is the cluster center mean vector, is the cluster covariance matrix, is the transpose operator.

[0136] The fermentation quality time series is compared point by point with the health threshold along the time axis to calculate the multi-dimensional quality deviation value, generate the fermentation quality deviation, and perform dynamic compensation calculation and parameter mapping to generate the control instruction set.

[0137] Specifically, the fermentation quality time series data is read, and the time step is kept at an example of 0.1 seconds. For the physical field characteristics and biological activity frequency domain characteristics of each time step, the item-by-item comparison with the health threshold is performed: the relative deviation percentage of the physical field characteristics and the biological activity frequency domain characteristics from the health threshold is calculated, and the deviation percentage exceeding the example of 5% is marked as an abnormal dimension, and the quality deviation is calculated for the marked abnormal dimension;

[0138] According to the positive / negative sign and size of the quality deviation, the compensation amount is calculated within the example of 0.1 second time window, the compensation amount calculation adopts the proportional-integral algorithm, the proportional coefficient is set to the example of 0.8, and the integral time constant is set to the example of 2 seconds. The upper limit of the compensation amount is set to the example of ±20% adjustment range. The compensation amount is mapped to the fermentation control parameters (temperature set value, stirring speed, aeration rate), and the compensated fermentation control parameters are converted into device control instructions according to the time step. The output frequency of the device control instructions matches the example of 0.1 second step of the fermentation quality time series. Before instruction execution, the device control instruction safety range (example of temperature 20-40℃, stirring speed 50-300rpm, aeration rate 0.1-1.0L / min) is checked, and when the range is exceeded, it is truncated to the boundary value and an alarm is triggered. The output control instruction set;

[0139] It should be noted that the expression for calculating the quality deviation for the marked abnormal dimension is:

[0140] ;

[0141] Wherein, is the quality deviation, is the physical field characteristics and biological activity frequency domain characteristics of the current time step, is the health threshold.

[0142] The evaluation module performs spatio-temporal correlation analysis and abnormal marking on the control instruction set and the fermentation quality time series, generates a visual prediction curve atlas, and extracts key features to map to obtain a fermented quality report of the steamed bread.

[0143] The control instruction set and the fermentation quality time series are time-stamped and matched for causal correlation modeling to generate instruction-quality joint analysis data.

[0144] Specifically, the control instruction set and the fermentation quality time sequence are read and aligned with timestamps at an exemplary 0.1-second time resolution, and an exemplary 0.5-second delay compensation calibration is applied to the control instructions. A correlation dataset of fermentation quality indicators is established, including timestamps, control instruction parameters (temperature setpoint, stirring speed, aeration rate), and physical field characteristics (three-dimensional flow rate linear regression slope, gas concentration moving average change rate, and temperature range and standard deviation), and biological activity frequency domain characteristics (main period and amplitude extracted by autocorrelation analysis of acoustic biological activity energy spectrum). The time delay cross-correlation relationship between the control instruction set and the fermentation quality indicators is analyzed within an exemplary 5-second window, and strong correlation pairs with a peak value exceeding an exemplary 0.8 are extracted, and the time delay parameters and correlation coefficients of the strong correlation pairs are recorded. The consistency of the causal relationship direction is verified (exemplary 90% compliance rate), and abnormal correlations with residuals exceeding an exemplary 3 times the standard deviation are removed to obtain instruction-quality joint analysis data.

[0145] The instruction-quality joint analysis data is subjected to multi-dimensional dynamic rendering and interactive labeling to obtain visualized prediction curve maps, and key feature analysis and multi-dimensional quality mapping are performed to obtain a fermented dough bread fermentation quality report.

[0146] Specifically, based on the control instruction parameters and fermentation quality indicators in the instruction-quality joint analysis data, the control instruction curve is dynamically rendered as an exemplary red solid line, the quality indicator curve is dynamically rendered as an exemplary blue dashed line, and the correlation marker points are dynamically rendered as exemplary orange circular points with a diameter of 8 pixels at an exemplary 50-millisecond refresh rate. An exemplary 10-second interval is defined by interactive operation, exemplary instruction change characteristics temperature step > 2℃ and quality response characteristics concentration change rate > 0.5% / s are identified, and quality scores (0-100 points) are calculated based on exemplary weights temperature stability 40%, biological activity synchronicity 30%; characteristic data, score results, and abnormal event correlation strength <0.6 or score <60 points are integrated to generate a fermented dough bread fermentation quality report.

[0147] In summary, the present application achieves dynamic monitoring and quantitative characterization of microbial metabolic activity during fermentation by generating acoustic biological activity indicator time series, accurately captures biological activity change characteristics during fermentation, improves the precision control of fermented dough bread processing fermentation quality, and effectively improves the fermented dough bread processing fermentation quality pass rate; a collaborative analysis model of dough rheological properties and thermodynamic properties is established by dynamic viscosity and thermal conductivity synchronous inversion calculation, which more accurately judges the fermentation state of the dough and timely discovers abnormal conditions; significantly improves the stability of product quality, avoids quality problems caused by uneven fermentation, makes the production process more controllable, provides technical support for formulating more scientific process parameters, helps to realize standardized production, ensures the fermentation quality of fermented dough bread, and improves the production efficiency.

[0148] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced, without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A data processing-based system for predicting the fermentation quality of mochi bread, characterized in that: include, The active module collects fermentation acoustic-environment data for preprocessing, and performs frequency domain feature transformation and Shannon entropy calculation through acoustic analysis algorithms to obtain the time series of acoustic biological activity indicators. The modeling module performs simultaneous inversion calculations of dynamic viscosity and thermal conductivity on the time series of acoustic bioactivity indicators and preprocessed fermentation acoustic-environment data to obtain the dynamic physical properties of mochi dough. Then, it performs three-dimensional dynamic modeling to obtain a digital twin model for mochi fermentation simulation. The specific steps are as follows: Dynamic temperature and humidity compensation analysis is performed on the preprocessed environmental monitoring data to generate environmental temperature and humidity compensation parameters; The time series of acoustic bioactivity indicators and environmental temperature and humidity compensation parameters are input into the rheology-thermal coupling inversion calculation architecture to perform dynamic viscosity calculation and thermal conductivity derivation, and output dynamic viscosity-thermal conductivity joint parameters. Time dimension alignment and data credibility verification were performed on the joint parameters of dynamic viscosity and thermal conductivity to generate dynamic physical properties of mochi dough. Three-dimensional spatial meshing was performed on the dynamic physical properties of mochi dough to generate a spatial distribution mesh set of physical property parameters. Spatial mapping and temporal binding were performed with the dynamic viscosity-thermal conductivity joint parameters to obtain multiphysics boundary condition data. Coupled solution is performed on multiphysics boundary condition data to generate dynamic coupled solution results of multiphysics, and time series integration and interactive interface binding are performed to generate a digital twin model of mochi fermentation simulation. The simulation module uses a digital twin model of mochi fermentation simulation to perform multi-physics field coupling calculations on the dynamic physical properties of mochi dough and outputs the mochi fermentation state tensor. The prediction module uses a prediction algorithm to fuse the time series of mochi fermentation state tensor and acoustic biological activity index to generate a fermentation quality time series. The compensation module compares the fermentation quality time series with the preset health threshold, obtains the fermentation quality deviation, and generates a control instruction set through real-time compensation. The evaluation module performs spatiotemporal correlation analysis and anomaly labeling on the control instruction set and fermentation quality time series, generates a visual prediction curve, and extracts key feature mappings to obtain a quality report on the fermentation of mochi bread.

2. The data processing-based mochi bread fermentation quality prediction system as described in claim 1, characterized in that: The fermentation acoustic-environmental data includes raw acoustic signals and environmental monitoring data; The preprocessing includes signal denoising, timestamp alignment, and data normalization.

3. The data processing-based mochi bread fermentation quality prediction system as described in claim 2, characterized in that: The process involves frequency domain feature transformation and Shannon entropy calculation using acoustic analysis algorithms to obtain a time-series sequence of acoustic biological activity indicators. The specific steps are as follows: A fast Fourier transform is performed on the preprocessed acoustic raw signal to generate acoustic frequency domain feature data, and relative energy data of the bioactive frequency band is extracted, the relative energy ratio is calculated, and the relative energy data of the frequency band is output. Based on the relative energy data of the frequency band, the uncertainty measure of the original acoustic signal is calculated by applying the Shannon entropy formula, and the time series of acoustic Shannon entropy values ​​is generated. Using preprocessed environmental monitoring data, temperature drift compensation and humidity correction are applied to the acoustic Shannon entropy time series, outputting acoustic biological activity index data, which are then sorted by collection timestamp to generate a time series of acoustic biological activity indexes.

4. The data processing-based mochi bread fermentation quality prediction system as described in claim 1, characterized in that: The method utilizes a digital twin model to simulate mochi fermentation, performing multiphysics coupling calculations on the dynamic properties of mochi dough and outputting a mochi fermentation state tensor. The specific steps are as follows: The dynamic physical properties of mochi dough are input into the digital twin model of mochi fermentation simulation, the dynamic physical property parameters of mochi dough are loaded, and the digital twin model of mochi fermentation simulation is initialized. Using the initialized digital twin model of mochi fermentation simulation, coupled simulation calculations of fluid motion, gas diffusion and temperature distribution are performed on the dynamic physical properties of mochi dough to generate multiphysics dynamic simulation results. The results of multiphysics dynamic simulation are converted into a four-dimensional spatiotemporal tensor format, and the mochi fermentation state tensor is output.

5. The data processing-based mochi bread fermentation quality prediction system as described in claim 4, characterized in that: The method employs a prediction algorithm to fuse the time series sequences of mochi fermentation state tensor and acoustic biological activity indicators to generate a fermentation quality time series. The specific steps are as follows: Time-stamp synchronization and spatial grid mapping were performed on the time series of mochi fermentation state tensor and acoustic bioactivity index to generate spatiotemporally synchronized fused data. Cross-modal feature extraction and weight allocation are performed on spatiotemporally synchronized fused data, and the dynamic correlation between physical field features and biological activity features is enhanced to generate a fused multimodal feature matrix. The multimodal feature matrix is ​​input into the time series prediction algorithm, and the temporal evolution law of the multimodal features is analyzed by time step expansion to generate the fermentation quality time series.

6. The data processing-based mochi bread fermentation quality prediction system as described in claim 5, characterized in that: The physical field characteristics include velocity field gradient, concentration field diffusion coefficient, and temperature field heat flux; The bioactivity characteristics include acoustic entropy fluctuations and microbial metabolic intensity.

7. The data processing-based mochi bread fermentation quality prediction system as described in claim 5, characterized in that: The fermentation quality time series is compared with a preset health threshold to obtain the fermentation quality deviation, and a control instruction set is generated through real-time compensation. The specific steps are as follows: Multidimensional feature analysis was performed on the historical fermentation quality time series to extract the distribution pattern of quality parameters under similar working conditions, fermentation quality feature clusters were generated, and dynamic threshold scaling calculations were performed to obtain the health threshold. The fermentation quality time series is compared with the health threshold point by point along the time axis to calculate the multi-dimensional quality deviation value, generate the fermentation quality deviation amount, and perform dynamic compensation calculation and adjustment parameter mapping to generate a control instruction set.

8. The data processing-based mochi bread fermentation quality prediction system as described in claim 7, characterized in that: The specific steps for performing spatiotemporal correlation analysis and anomaly labeling on the control command set and fermentation quality time series, generating a visualized prediction curve, and extracting key feature maps to obtain a mochi bread fermentation quality report are as follows. Time stamp synchronization matching and causal relationship modeling were performed on the control command set and fermentation quality time series to generate command-quality joint analysis data; Multi-dimensional dynamic rendering and interactive labeling are performed on the instruction-quality joint analysis data to obtain a visualized prediction curve map. Key feature analysis and multi-dimensional quality mapping are then performed to obtain a quality report on the fermentation of mochi bread.

Images