A process quality control system for a soy product production process

Abstract

The application discloses a kind of soy products production process material measuring index process quality control system, it is related to soy products processing quality control technical field, the present application includes: state calibration module, before feeding, empty tank state is passed into first steam pulse, according to tank temperature rise, to calibrate equipment heat transfer efficiency coefficient, after feeding, cold state stationary period is passed into second steam pulse, according to pulp temperature rise, to identify the equivalent total heat capacity of pulp and calculate theoretical minimum heat demand value;Determination module is used to shield pulp sensor during slurry boiling start period, only collect steam condensate flow, when the accumulated mass of condensate reaches the lower limit of condensate mass corresponding to the theoretical minimum heat demand value, and the second derivative of condensate instantaneous flow appears the preset inflection point characteristics caused by protein irreversible endothermic mutation, output start period termination signal, and record the time difference between the time when the cumulative standard is met and the inflection point time, the present application overcomes the defect that the matching accuracy continuously attenuates due to the gradual aging of the existing database matching method.

Description

Technical Field

[0001] This invention relates to the field of quality control technology for soybean product processing, and in particular to a process quality control system for physical indicators in soybean product production. Background Technology

[0002] In the production of soy products, the protein thermal denaturation stage of the boiling process directly determines the gel strength and water retention of the finished product. The protein, moisture and fat content of different batches of soybean raw materials varies significantly, and the heat exchange efficiency of the equipment continues to decline with the operating cycle. The control system must have adaptive capabilities to ensure that the product quality is stable and consistent between batches.

[0003] Existing technologies mainly employ two strategies. One is full-sensor closed-loop PID control: multi-modal sensors such as viscosity, temperature, and turbidity are activated throughout the process, and feedback adjustment is made based on the deviation between real-time measured values ​​and setpoints. However, during the feeding and start-up phase, the suspension of soybean residue, the inclusion of air bubbles, and the effect of probe wall adhesion in the slurry cause severe fluctuations in sensor signals and extremely low signal-to-noise ratios. False deviations can easily lead to PID integral saturation and repeated valve overshoot, and in severe cases, trigger safety interlocks to cause erroneous shutdowns. Even with digital filtering and delay smoothing, it is impossible to physically distinguish between bubble noise and real colloidal phase transition signals. More importantly, protein thermal denaturation is irreversible. In the process, PID control only focuses on whether the current instantaneous value meets the standard and cannot identify the phase transition path that the value has undergone. That is, the same viscosity reading may come from the orderly unfolding of protein or from the cooling and drop after overheating and aggregation. The corresponding gel microstructures are completely different, but the sensor cannot distinguish them. Secondly, there is the raw material pre-inspection and database matching: by pre-detecting indicators such as soybean protein and moisture, the preset process parameters are retrieved from the database. However, such databases are highly dependent on the equipment calibration status. Gradual aging such as heat exchanger scaling and agitator shaft wear will cause the actual effect of the original parameters to drift continuously, resulting in the matching accuracy decreasing over time.

[0004] Existing technologies based on real-time sensor feedback or static database matching cannot achieve reliable adaptive control of the soybean product cooking process under the dual constraints of sensor physical distortion during startup and the unobservable irreversible phase transition path of proteins. Therefore, this invention proposes a process quality control system for physical indicators in soybean product production. Summary of the Invention

[0005] The purpose of this invention is to provide a process quality control system for physical indicators in the production of soybean products, which solves the problem that existing technologies based on real-time sensor feedback or static database matching cannot achieve reliable adaptive control of the soybean product boiling process under the dual constraints of sensor physical distortion during the start-up period and the unobservable irreversible phase transition path of proteins.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: This invention relates to a process quality control system for physical indicators in the production of soybean products, comprising: The state calibration module introduces a first steam pulse into the empty tank before feeding to calibrate the heat transfer efficiency coefficient of the equipment based on the temperature rise of the tank. After feeding, during the cold resting period, a second steam pulse is introduced to identify the equivalent total heat capacity of the slurry based on the temperature rise of the slurry and calculate the theoretical minimum heat requirement. The judgment module is used to shield the slurry sensor during the start-up period of slurry cooking and only collect the steam condensate flow rate. When the cumulative mass of condensate reaches the lower limit of condensate mass corresponding to the theoretical minimum heat requirement, and the second derivative of the instantaneous flow rate of condensate shows the preset inflection point characteristic caused by irreversible endothermic mutation of protein, the start-up period termination signal is output and the time difference between the cumulative target time and the inflection point time is recorded. The phase transition trajectory generation module obtains the full-time curve of the instantaneous flow rate of condensate during the start-up period of the soy boiling process and extracts the morphological fingerprint parameter set from it. The parameter set includes at least the initial endothermic decay slope, the ratio of flow rate change before and after the endothermic abrupt change, and the recovery characteristic time constant. The parameter set and time difference are input into the trajectory mapping model to generate the protein ordered denaturation target trajectory. The trajectory includes the temperature and viscosity phase plane heating curve, the inflection point interval of protein secondary structure transformation, and the corresponding expected viscoelastic response characteristics. The phase change process control module restarts the slurry sensor after the start-up period ends, applies variable frequency and amplitude stirring excitation disturbance in the inflection zone of the target trajectory, and uses the viscoelastic response and the comparison result with the corresponding expected characteristics to gate the trajectory advancement. It also compares the real-time temperature and viscosity phase plane trajectory with the target trajectory shape. If the deviation exceeds the preset corridor, the power is increased by lag; if it is ahead, the attenuation amount is pre-calculated based on the heat transfer efficiency coefficient and attenuation is applied before reaching the upper limit of the preset corridor. The model self-correction module uses the batch-to-batch evolution trend of morphological fingerprint parameter groups to perform targeted corrections on the trajectory mapping model and the slurry equivalent total heat capacity identification model.

[0007] Furthermore, the model self-correction module acquires the morphological fingerprint parameter set after the start-up period of each batch of soy milk, and compares the initial endothermic decay slope, flow rate change ratio and recovery feature time constant with a dynamic baseline constructed from the statistical features of the corresponding parameters of a preset number of historical batches. The dynamic baseline is the moving average or exponentially weighted average of the corresponding parameters within a preset number of historical batches. When the initial heat absorption attenuation slope drifts beyond the threshold in the same direction relative to the number of consecutive preset batches of its dynamic baseline, the mapping coefficients related to the heat transfer rate in the trajectory mapping model are corrected, and the correction results are synchronously written back to the state calibration module for the first steam pulse calibration of the next batch. When the flow rate change exceeds the threshold in the same direction as the number of consecutive preset batches, the mapping coefficients related to the heat capacity of the slurry in the trajectory mapping model are corrected. When the number of consecutive preset batches of the recovery characteristic time constant exceeds the threshold in the same direction and the initial heat absorption attenuation slope and flow rate change ratio do not show a trend of drift, the identification model of the equivalent total heat capacity of the slurry in the state calibration module is corrected. The correction of the mapping coefficients is carried out in a gradual manner, with the correction amount in a single batch not exceeding a preset proportion of the current coefficient value. The correction of the identification model is carried out by updating the model parameters.

[0008] Furthermore, the trajectory mapping model generates the protein's ordered denaturation target trajectory according to the following mapping rules: The initial heat absorption decay slope is used to map the upper limit of the heating rate in the first segment of the target trajectory. The larger the decay slope, the higher the upper limit of the heating rate and the shorter the duration of the first segment. The width and curvature of the trajectory inflection point interval are mapped by the ratio of flow change. The larger the ratio of change, the wider the inflection point interval and the smaller the curvature, which corresponds to a smoother process of protein secondary structure transformation. The duration of the stable period in the latter part of the target trajectory is mapped by the recovery feature time constant. The larger the time constant, the longer the stable period. The time difference is used to map the overall offset of each inflection point interval on the time axis. The larger the time difference, the more the overall inflection point interval shifts backward. The expected viscoelastic response characteristics are determined jointly based on the protein secondary structure transition type corresponding to the current inflection point interval and the flow rate change ratio. The larger the flow rate change ratio, the longer the expected stress relaxation time and the lower the crossover frequency of the expected storage modulus and loss modulus.

[0009] Furthermore, the frequency range and amplitude range of the variable frequency and amplitude stirring excitation disturbance are determined by the protein secondary structure transition type corresponding to the current inflection point interval. The interval from ordered structure to random coil transition adopts a first frequency range and a first amplitude range, while the interval from random coil to gel state transition adopts a second frequency range lower than the first frequency range and a second amplitude range higher than the first amplitude range. The viscoelastic response characteristic parameters include at least one of stress relaxation time and the cross frequency of storage modulus and loss modulus, wherein the stress relaxation time is the time required for the slurry shear stress to decay to a preset proportion of the initial value after the stirring excitation disturbance is removed. The comparison involves comparing the deviation between the collected viscoelastic response characteristic parameters and the expected viscoelastic response characteristic parameters corresponding to the inflection point interval with a preset allowable deviation. When the deviation is within the allowable range, it is determined that the protein structure transformation has been completed and the trajectory is allowed to enter the next stage. When the deviation exceeds the allowable range, the current trajectory stage is maintained and stirring excitation is continuously applied until the deviation falls back to the allowable range.

[0010] Furthermore, the advance suppression power attenuation amount is jointly determined by the heat transfer efficiency coefficient, the current heating power, the advance offset between the real-time phase plane trajectory and the target trajectory, and the advance offset rate as inputs. The attenuation amount is directly proportional to the product of the advance offset and the advance offset rate, inversely proportional to the heat transfer efficiency coefficient, and multiplied by a preset safety factor. This ensures that the trajectory advancement amount generated by the expected heating rate corresponding to the attenuated heating power within a preset time window is not greater than the current advance offset, so that the real-time trajectory stops further approaching the preset corridor upper limit after the attenuation takes effect. The preset time window is determined based on the equivalent total heat capacity of the slurry and the current heating power; the larger the equivalent total heat capacity, the longer the time window.

[0011] Furthermore, the preset inflection point feature is that the second derivative of the instantaneous flow rate of condensate crosses a preset negative threshold from positive within a preset detection time window, and the duration for which the first derivative remains negative after the crossing exceeds a preset minimum duration. The preset detection time window starts at the moment when the accumulated mass of condensate reaches the lower mass limit, and the window length is determined based on the theoretical minimum heat requirement. The preset negative threshold and the preset minimum duration are dynamically determined based on the equivalent total heat capacity of the slurry and the theoretical minimum heat requirement. The larger the equivalent total heat capacity, the smaller the absolute value of the preset negative threshold and the longer the preset minimum duration, so that the inflection point feature has adaptive detection sensitivity for slurries with different protein contents and different filling volumes.

[0012] Furthermore, the heat transfer efficiency coefficient is obtained by normalizing the ratio of the input heat of the first steam pulse to the temperature rise of the tank after normalizing it with the known heat capacity of the tank, which characterizes the effectiveness of the current batch of equipment in transferring steam heat to the medium inside the tank; the equivalent total heat capacity of the slurry is obtained by correcting the ratio of the input heat of the second steam pulse to the temperature rise of the slurry with the heat transfer efficiency coefficient; the theoretical minimum heat requirement is the product of the equivalent total heat capacity of the slurry and the difference between the target endpoint temperature and the initial cold temperature; the first steam pulse and the second steam pulse have different settings in terms of pulse duration and steam pressure, with the duration of the first steam pulse being shorter than that of the second steam pulse and the steam pressure being higher than that of the second steam pulse.

[0013] Furthermore, the phase transition trajectory generation module also extracts the inflection point region time domain width parameter from the full-time curve of the instantaneous flow rate of the condensate. The inflection point region time domain width parameter is the duration from when the second derivative of the instantaneous flow rate of the condensate deviates from the normal decay trajectory to when it crosses the preset negative threshold and recovers to the preset proportion before crossing. It characterizes the degree of spatial non-uniformity of the irreversible endothermic mutation of the protein advancing layer by layer from the wall to the center in this batch of slurry.

[0014] Furthermore, the trajectory mapping model adaptively scales the preset corridor width of each inflection point interval according to the time-domain width parameter of the inflection point region; the larger the time-domain width parameter, the wider the corridor width.

[0015] Furthermore, the phase change process control module performs the morphological comparison using the scaled corridor width.

[0016] The present invention has the following beneficial effects: This invention employs a dual-pulse separation calibration mechanism: a first steam pulse before feeding calibrates the equipment's heat transfer efficiency coefficient in an empty tank state, and a second steam pulse during the cold settling period after feeding identifies the equivalent total heat capacity of the slurry. This decouples equipment aging and raw material differences at the physical level. During the slurry start-up period, it shields the slurry-side distortion sensor, using only the cumulative mass of steam condensate for heat proxy measurement and the second derivative of the instantaneous flow rate of condensate to detect the inflection point of irreversible endothermic protein reaction. This fundamentally bypasses the false signal interference during the period of physical distortion of the sensor.

[0017] This invention uses the morphological fingerprint parameter set of the full-time curve of condensate during the start-up period, combined with the time difference between condensate reaching the standard and the inflection point, to generate a unique protein ordered denaturation target trajectory for that batch through a trajectory mapping model. In the trajectory inflection point interval, variable frequency and amplitude stirring excitation is used as an active structural probe, and the trajectory stage is gated by comparing the results of viscoelastic response characteristics. Asymmetric correction is implemented in phase plane trajectory tracking, that is, lag can be caught, and advance is pre-calculated with the calibrated heat transfer efficiency coefficient for attenuation, and attenuation is applied before reaching the limit. This mechanism prevents irreversible excessive denaturation of proteins, so that the protein gel formation process is completed in an orderly manner along the target trajectory, ensuring the stability and consistency of gel strength and water retention between batches of products.

[0018] This invention utilizes the morphological fingerprint parameter set acquired after the start-up period of each batch, compares each parameter with the dynamic baseline of historical batches, identifies the equipment aging component by the trend drift of the initial endothermic decay slope, identifies the raw material fluctuation component by the trend drift of the flow rate change ratio before and after the endothermic mutation, and identifies the mismatch between heat transfer and heat capacity coupling by the independent drift of the recovery characteristic time constant. Differentiated progressive correction or model update is performed on the three types of components, and the correction results are written back to the state calibration module, so as to complete the predictive directional correction before the quality deviation occurs, overcoming the defect of existing database matching methods that cause the matching accuracy to continuously decline due to the progressive aging of equipment. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 The present invention provides a system block diagram of the process quality control system for physical indicators in the production process of soybean products. Detailed Implementation

[0021] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0022] Example 1 See Figure 1 This embodiment provides a process quality control system for physical indicators in the production of soybean products. During the start-up phase of the boiling process, when the slurry sensor is physically distorted due to soybean residue suspension, air bubbles, and probe wall adhesion, measurable physical quantities from the steam side replace unreliable sensor signals from the slurry side. This achieves full adaptive control of the irreversible phase transition process of protein thermal denaturation, ensuring consistent gel strength and water retention between batches. The system includes: The state calibration module introduces a first steam pulse into the empty tank before feeding to calibrate the heat transfer efficiency coefficient of the equipment based on the temperature rise of the tank. After feeding, during the cold resting period, a second steam pulse is introduced to identify the equivalent total heat capacity of the slurry based on the temperature rise of the slurry and calculate the theoretical minimum heat requirement. The heat transfer efficiency coefficient is obtained by normalizing the ratio of the input heat of the first steam pulse to the temperature rise of the tank after normalizing it to the known heat capacity of the tank, and it characterizes the effectiveness of the current batch of equipment in transferring steam heat to the medium inside the tank. The equivalent total heat capacity of the slurry is obtained by correcting the ratio of the input heat of the second steam pulse to the temperature rise of the slurry after normalizing it to the heat transfer efficiency coefficient. The theoretical minimum heat requirement is the product of the equivalent total heat capacity of the slurry and the difference between the target endpoint temperature and the initial cold temperature. The first steam pulse and the second steam pulse have different settings in terms of pulse duration and steam pressure. The duration of the first steam pulse is shorter than that of the second steam pulse, and the steam pressure is higher than that of the second steam pulse. That is, the first pulse acts on the metal wall of the empty tank, and the thermal response is fast. A short high-pressure pulse can obtain a sufficient temperature rise signal. The second pulse acts on the slurry containing water, and the thermal response is slow. A longer pulse duration is required to ensure that the slurry is fully heat-absorbing and generates a identifiable overall temperature rise.

[0023] Through the status calibration module, the heat transfer capacity of the cooking pot in the current batch and the heat capacity of the raw materials are quantified and decoupled from each other, providing the judgment module with a heat calculation benchmark independent of the sensor signal.

[0024] During the initial boiling phase, the soybean residue inside the slurry is suspended and contains a large amount of air bubbles. The viscosity and temperature sensors are affected by the probe wall effect, resulting in severe signal jitter and an extremely low signal-to-noise ratio. If such signals are relied upon for feedback control, false deviations can easily lead to integral saturation and repeated valve overshoot. In this embodiment, all sensors on the slurry side are actively shielded during this stage. Instead, the condensate flow rate on the steam side is used as a proxy heat measurement signal. That is, the latent heat released by steam condenses into water. The instantaneous flow rate and cumulative mass of the condensate directly reflect the real-time heat absorption rate and total heat absorption of the slurry, and are not affected by physical interference inside the slurry.

[0025] The judgment module is used to shield the slurry sensor during the start-up period of slurry cooking and only collect the steam condensate flow rate. When the cumulative mass of condensate reaches the lower limit of condensate mass corresponding to the theoretical minimum heat requirement, and the second derivative of the instantaneous flow rate of condensate shows the preset inflection point characteristic caused by irreversible endothermic mutation of protein, the start-up period termination signal is output and the time difference between the cumulative target time and the inflection point time is recorded. The preset inflection point is characterized by the second derivative of the instantaneous condensate flow rate crossing a preset negative threshold from positive within a preset detection time window, and the duration for which the first derivative remains negative after crossing exceeds a preset minimum duration. The preset detection time window starts at the moment when the accumulated condensate mass reaches the lower mass limit, and the window length is determined based on the theoretical minimum heat requirement. The physical essence of this inflection point is that protein molecules undergo an irreversible conformational change after reaching the denaturation temperature. This process absorbs a large amount of latent heat without a significant temperature rise, resulting in a characteristic acceleration, sudden increase, and then decline in the instantaneous heat absorption rate, which is reflected in the condensate flow rate. The inflection point is defined as a single event in which the first derivative first decreases and then increases, and the second derivative abruptly turns from positive to negative. A dual-condition joint determination is made: the second derivative crossing the negative threshold and the first derivative remaining negative for more than a minimum duration ensure that the inflection point is indeed caused by irreversible endothermic protein activity, rather than a transient spurious signal caused by accidental fluctuations in steam pressure or bubble interference. The preset negative threshold and preset minimum duration are dynamically determined based on the slurry's equivalent total heat capacity and theoretical minimum heat requirement. The larger the equivalent total heat capacity, the smaller the absolute value of the preset negative threshold and the longer the preset minimum duration, making the inflection point characteristic adaptively sensitive to slurries with different protein contents and different canning volumes.

[0026] In this embodiment, the preset negative threshold and the preset shortest duration are quantitatively determined as follows: Let the preset negative threshold be... , Taking a negative value, the dimension is consistent with the second derivative of the instantaneous flow rate of condensate, and the equivalent total heat capacity of the slurry is... (Unit: kJ / ℃), reference heat capacity is , Take the arithmetic mean of the equivalent total heat capacity of a preset number of historical batches, in kJ / ℃, with a baseline negative threshold of [value missing]. , If the value is negative, then:

[0027] In the above formula, the exponent p is taken as 0.3~0.7, preferably 0.5.

[0028] Let the preset minimum duration be...

[0029] The exponent q ranges from 0.5 to 1.0, with 0.7 being the preferred value.

[0030] Baseline negative threshold Compared with the shortest duration of the benchmark The initial calibration procedure is as follows: During the initial commissioning phase of the system or after a major equipment overhaul, take a standard protein solution with a known protein mass fraction, i.e., a soybean protein aqueous dispersion with a protein mass fraction of 3.5% to 4.5%. Under standard tank filling conditions, perform the standard boiling start-up process no less than N times, where N≥5. Each time, record the peak value and duration of the second derivative of the instantaneous flow rate of condensate crossing the negative value. The peak value is the minimum value of the second derivative. Calculate the arithmetic mean of N measurements, and determine it according to the following relationship:

[0031]

[0032] in, Take a value of 0.6 to 0.9. Take a value of 1.1 to 1.5; The duration of the journey.

[0033] When a major equipment overhaul, heat exchanger replacement, or boiler replacement occurs, the above calibration procedure should be repeated to update the calibration status. and .

[0034] The two key quantities output by the determination module are the start-up termination signal, which is used to trigger the subsequent phase change process control module to take over the system, and the time difference between the cumulative target time and the inflection point time, which records the delay between the theoretical heat arrival and the actual occurrence of phase change. This time difference directly reflects the magnitude of the thermal inertia of the slurry-equipment coupling system of this batch.

[0035] The phase transition trajectory generation module obtains the full-time curve of the instantaneous flow rate of condensate during the start-up period of the soy boiling process and extracts the morphological fingerprint parameter set from it. The parameter set includes at least the initial endothermic decay slope, the ratio of flow rate change before and after the endothermic abrupt change, and the recovery characteristic time constant. The parameter set and time difference are input into the trajectory mapping model to generate the protein ordered denaturation target trajectory. The trajectory includes the temperature and viscosity phase plane heating curve, the inflection point interval of protein secondary structure transformation, and the corresponding expected viscoelastic response characteristics. Among them, the initial endothermic decay slope reflects the thermal diffusion rate of the cold slurry in the early stage of heating. The faster the decay, the faster the heat conduction inside the slurry. The ratio of the flow rate change before and after the endothermic mutation reflects the jump in the endothermic intensity of the protein denaturation. The larger the ratio, the more latent heat is consumed in the denaturation of this batch of proteins. The recovery characteristic time constant reflects the speed at which the slurry returns to the normal endothermic mode after the denaturation endothermic peak. The larger the time constant, the stronger the thermal inertia of the gelation stage.

[0036] In this embodiment, the trajectory mapping model generates the protein ordered denaturation target trajectory according to the following mapping rules: The initial heat absorption decay slope is used to map the upper limit of the heating rate in the first segment of the target trajectory. The larger the decay slope, the higher the upper limit of the heating rate and the shorter the duration of the first segment. The width and curvature of the trajectory inflection point interval are mapped by the ratio of flow change. The larger the ratio of change, the wider the inflection point interval and the smaller the curvature, which corresponds to a smoother process of protein secondary structure transformation. The duration of the stable period in the latter part of the target trajectory is mapped by the recovery feature time constant. The larger the time constant, the longer the stable period. The time difference is used to map the overall offset of each inflection point interval on the time axis. The larger the time difference, the more the overall inflection point interval shifts backward. The expected viscoelastic response characteristics are determined jointly based on the protein secondary structure transition type corresponding to the current inflection point interval and the flow rate change ratio. The larger the flow rate change ratio, the longer the expected stress relaxation time and the lower the crossover frequency of the expected storage modulus and loss modulus.

[0037] In this embodiment, the quantitative mapping relationship of the trajectory mapping model is as follows: Let the initial heat absorption decay slope be s, and the upper limit of the heating rate in the first segment of the target trajectory be... The two satisfy a piecewise linear mapping relationship: when hour, ; when hour,

[0038] when hour, ; in, Take 0.02~0.08 kg·s -2 , Take 0.10~0.20 kg·s -2 , Use a flow rate of 0.8~2.0℃ / min. The velocity is 2.5–4.5 ℃ / min, and the duration of the first segment of the target trajectory is... It is determined by the following formula:

[0039] in, This is the target temperature rise from the initial cold temperature to the protein denaturation initiation temperature.

[0040] Let the ratio of the flow rate change before and after the endothermic abrupt change be . The width of the inflection point interval is The width of the benchmark inflection point interval is Width adjustment coefficient is ,satisfy:

[0041] in, The lower limit is 0.8 times. The upper limit is 2.5 times. .

[0042] Inflection point interval curvature Specifically, it is represented by the following formula.

[0043] in, As the reference curvature, This is the curvature adjustment coefficient, ranging from 0.2 to 0.8. Let the recovery feature time constant be . The duration of the later stable period of the target trajectory is The scaling factor is , Determined through linear fitting of historical batch data, with initial default values ​​ranging from 0.8 to 1.5, the following is observed:

[0044] Let the time difference between the cumulative achievement time and the inflection point time be denoted as . The overall offset of each inflection point interval on the time axis is lag coefficient This manifests as:

[0045] Let the expected stress relaxation time be... The reference stress relaxation time is The expected crossover frequency between the energy storage modulus and the loss modulus is: The reference cross frequency is Then it will manifest as:

[0046]

[0047] in, Take a value of 0.2 to 1.0. Take a value of 0.2 to 1.0.

[0048] In the above mapping rules, a large initial endothermic decay slope means that the slurry conducts heat quickly, so a higher heating rate can be used without generating an excessive temperature gradient; a large flow rate change ratio means that the protein denaturation endothermic intensity is high and the phase transition process itself lasts a long time, so the inflection point interval needs a wider process window and a smoother transition curvature; a large recovery time constant means that the denatured slurry has a large thermal inertia, requiring a longer stabilization period to wait for the gel network to fully form; a large time difference means that the delay from heating to actual denaturation is long, so the entire trajectory needs to be shifted backward on the time axis.

[0049] The phase change process control module restarts the slurry sensor after the start-up period ends, applies variable frequency and amplitude stirring excitation disturbance in the inflection zone of the target trajectory, and uses the viscoelastic response and the comparison result with the corresponding expected characteristics to gate the trajectory advancement. It also compares the real-time temperature and viscosity phase plane trajectory with the target trajectory shape. If the deviation exceeds the preset corridor, the power is increased by lag; if it is ahead, the attenuation amount is pre-calculated based on the heat transfer efficiency coefficient and attenuation is applied before reaching the upper limit of the preset corridor. The stirring excitation perturbation is not conventional stirring for the purpose of equalizing temperature, but rather serves as an active structural probe: within the critical range of protein secondary structure transition, by applying mechanical perturbations of known frequency and amplitude and measuring the viscoelastic response of the slurry, it is determined whether the protein conformation has truly completed the expected structural transition.

[0050] The frequency range and amplitude range of the variable frequency and amplitude stirring excitation disturbance are determined by the protein secondary structure transition type corresponding to the current inflection point interval. The interval from ordered structure to random coil adopts a first frequency range and a first amplitude range, and the interval from random coil to gel state adopts a second frequency range lower than the first frequency range and a second amplitude range higher than the first amplitude range. The viscoelastic response characteristic parameters include at least one of stress relaxation time and the cross frequency of storage modulus and loss modulus, wherein the stress relaxation time is the time required for the slurry shear stress to decay to a preset proportion of the initial value after the stirring excitation disturbance is removed. The physical basis for this differentiated setting is as follows: the transformation from ordered structure to random coil involves the unfolding of α-helices and natural β-sheets. This process is mainly characterized by the breaking of intramolecular hydrogen bonds. The response to external mechanical excitation is such that a significant modulus change can be detected by high-frequency, small-amplitude perturbations. Therefore, a higher frequency and a smaller amplitude are used. The transformation from random coil to gel state involves intermolecular cross-linking and network formation. The overall viscoelasticity of the slurry changes significantly. A lower frequency and a larger amplitude perturbation are required to effectively excite and measure the response.

[0051] The comparison involves comparing the deviation between the collected viscoelastic response characteristic parameters and the expected viscoelastic response characteristic parameters corresponding to the inflection point interval with a preset allowable deviation. When the deviation is within the allowable range, it is determined that the protein structure transformation has been completed and the trajectory is allowed to enter the next stage. When the deviation exceeds the allowable range, the current trajectory stage is maintained and stirring excitation is continuously applied until the deviation falls back to the allowable range.

[0052] In phase plane trajectory morphology comparison, the system detects morphological deviations between the actual trajectory formed by real-time acquired viscosity and temperature data on the temperature and viscosity phase planes and the target trajectory; when the deviation exceeds the preset corridor range, the correction strategy exhibits directional asymmetry. The advance suppression power attenuation amount is jointly determined by the heat transfer efficiency coefficient, the current heating power, the advance offset between the real-time phase plane trajectory and the target trajectory, and the advance offset rate as inputs. The attenuation amount is directly proportional to the product of the advance offset and the advance offset rate, inversely proportional to the heat transfer efficiency coefficient, and multiplied by a preset safety factor. This ensures that the trajectory advancement amount generated by the expected heating rate corresponding to the attenuated heating power within a preset time window is not greater than the current advance offset, so that the real-time trajectory stops further approaching the preset corridor upper limit after the attenuation takes effect. The preset time window is determined based on the equivalent total heat capacity of the slurry and the current heating power. The larger the equivalent total heat capacity, the longer the time window.

[0053] The model self-correction module uses the batch-to-batch evolution trend of morphological fingerprint parameter groups to perform targeted corrections on the trajectory mapping model and the slurry equivalent total heat capacity identification model.

[0054] It should be further explained that after the start-up period of each batch of boiling soy milk is completed, the model self-correction module obtains the morphological fingerprint parameter group and compares the initial endothermic decay slope, flow rate change ratio and recovery feature time constant with the dynamic baseline constructed by the statistical features of the corresponding parameters of a preset number of historical batches. The dynamic baseline is the moving average or exponentially weighted average of the corresponding parameters within a preset number of historical batches. When the initial heat absorption attenuation slope drifts beyond the threshold in the same direction relative to the number of consecutive preset batches of its dynamic baseline, the mapping coefficients related to the heat transfer rate in the trajectory mapping model are corrected, and the correction results are synchronously written back to the state calibration module for the first steam pulse calibration of the next batch. When the flow rate change exceeds the threshold in the same direction as the number of consecutive preset batches, the mapping coefficients related to the heat capacity of the slurry in the trajectory mapping model are corrected. When the number of consecutive preset batches of the recovery characteristic time constant exceeds the threshold in the same direction and the initial heat absorption attenuation slope and flow rate change ratio do not show a trend of drift, the identification model of the equivalent total heat capacity of the slurry in the state calibration module is corrected. The correction of the mapping coefficients is carried out in a gradual manner, with the correction amount in a single batch not exceeding a preset proportion of the current coefficient value. The correction of the identification model is carried out by updating the model parameters.

[0055] In this embodiment, the change in the initial heat absorption attenuation slope is mainly due to the continuous decline in heat transfer efficiency caused by heat exchanger fouling, which belongs to the equipment aging component. The heat transfer rate mapping coefficient in the trajectory mapping model should be corrected and the correction result should be written back to the state calibration module so that the heat transfer efficiency calibration benchmark of subsequent batches can be updated synchronously with the equipment state. The change in the flow rate change ratio before and after the heat absorption mutation is mainly due to the difference in protein content and composition caused by the change of soybean origin or variety, which belongs to the raw material fluctuation component. The heat capacity mapping coefficient in the trajectory mapping model should be corrected. When neither of the above two parameters shows a trend of drift but the characteristic time constant drifts alone, it indicates that the coupling relationship between the equipment heat transfer characteristics and the raw material heat capacity has changed systematically. That is, there is a mismatch between the existing heat capacity identification model and the actual physical system, and the identification model itself needs to be updated. The mapping coefficients are adjusted incrementally and a maximum correction ratio is set for each batch to prevent abnormal data from polluting the model. The identification model is corrected by updating all model parameters. The correction is completed at the end of the startup period and before trajectory generation, without waiting for quality inspection after discharge. This allows the trajectory generation of this batch to benefit from the latest correction results, achieving predictive maintenance rather than reactive correction.

[0056] The five modules described above form a closed loop through the following data flow: the state calibration module's dual-pulse calibration produces the heat transfer efficiency coefficient and equivalent total heat capacity; the judgment module uses these calibration values ​​as a benchmark to collect the full-time curve of condensate during the startup period and detect the inflection point of irreversible phase change, producing startup and termination signals and time differences; the phase change trajectory generation module receives the full-time curve of condensate, calibration parameters, and time differences, extracts the morphological fingerprint parameter set, and generates a batch-specific target trajectory through the trajectory mapping model; the phase change process control module receives the target trajectory, performs structural probe-type gating with stirring excitation disturbance, and performs trajectory tracking under irreversible constraints using an asymmetric strategy; the model self-correction module receives the morphological fingerprint parameter set for each batch, and corrects the trajectory mapping model and identification model through cross-batch trend analysis, writing the correction results back to the state calibration module for use in the next batch; the entire closed loop enables the system to automatically adapt to three types of long-term drift factors—equipment aging, raw material batch differences, and model mismatch—without human intervention.

[0057] Example 2 In the above embodiments, the position of the target trajectory inflection point interval is offset by the time difference between the time when the condensate accumulation reaches the target and the inflection point. This implicitly assumes that the thermal hysteresis is uniform throughout the slurry, and protein denaturation occurs synchronously at the same time. However, as steam heat is transferred from the tank wall to the center of the slurry, the irreversible endothermic mutation of the protein is not completed synchronously throughout the entire slurry. Instead, it progresses layer by layer along the radial direction from the wall to the center, forming a spatially non-uniform denaturation wavefront. If the denaturation wavefront progresses rapidly and the spatial non-uniformity is low, the slurry denatures almost synchronously. In this case, the preset corridor of the target trajectory should be narrowed to accommodate higher control precision requirements. If the progress is slow and the spatial non-uniformity is high, the denaturation process of each layer of the slurry differs significantly. In this case, the preset corridor should be widened to avoid misinterpreting normal spatial distribution differences as trajectories. The above embodiment, which uses a single time difference to uniformly determine the preset corridor, cannot adaptively adjust the corridor width according to the actual spatial non-uniformity within each batch of slurry. This results in the preset corridor being too wide, weakening the precision control of batches with good synchronization, and too narrow, causing miscorrection of non-uniform batches. To solve this problem, in this embodiment, the phase change trajectory generation module also extracts the inflection point region time domain width parameter from the full-time curve of the instantaneous flow rate of condensate. The inflection point region time domain width parameter is the duration from the second derivative of the instantaneous flow rate of condensate deviating from the normal decay trajectory to crossing the preset negative threshold and recovering to the preset proportion before crossing. It characterizes the spatial non-uniformity of the irreversible endothermic mutation of the protein advancing layer by layer from the wall to the center within the batch of slurry.

[0058] If the internal temperature gradient of the slurry is small and the heat transfer is uniform, the proteins near the wall and in the central region reach the denaturation temperature almost simultaneously. The second derivative of condensate will exhibit a short-duration, sharp-peaked single-pass event, resulting in a small time-domain width in the inflection point region. If the internal temperature gradient of the slurry is large and the heat transfer is uneven, the proteins near the wall denature first, followed by the proteins in the central region. The denaturation wavefront advances layer by layer from the outside to the inside, and the passage of the second derivative of condensate will be widened on the time axis, resulting in a large time-domain width in the inflection point region. Therefore, the time-domain width of the inflection point region directly quantifies the degree of non-uniformity of protein denaturation in the spatial dimension within this batch of slurry.

[0059] The trajectory mapping model adaptively scales the preset corridor width of each inflection point interval according to the time-domain width parameter of the inflection point region; the larger the time-domain width parameter, the wider the corridor width.

[0060] When the time domain width of the inflection point region is small, the slurry undergoes near-synchronous deformation, and the background noise caused by spatial non-uniformity in the temperature and viscosity data collected by the sensors is extremely low. At this time, the corridor width should be narrowed to achieve high-precision tracking of the trajectory and avoid wasting control precision under good synchronization conditions. When the time domain width of the inflection point region is large, the slurry undergoes layer-by-layer deformation with significant internal differences. The average temperature and viscosity data collected by the sensors will inevitably contain inherent fluctuations caused by spatial non-uniformity. At this time, the corridor width should be widened so that the control system will not misjudge this unavoidable physical fluctuation as a heating deviation and make incorrect power interventions.

[0061] The phase transition process control module performs the morphological comparison using the scaled corridor width.

[0062] Therefore, based on the overall offset trajectory position through time difference in Embodiment 1, this embodiment further adaptively adjusts the corridor width of trajectory tracking by adjusting the time domain width of the inflection point region. This allows the same control system to match appropriate control accuracy under different spatial non-uniformity conditions. When non-uniform, the tolerance is relaxed to avoid false corrections, and when uniform, the tolerance is tightened to fully utilize control accuracy. The two parameters adaptively adjust the target trajectory from the time and space dimensions, respectively, together constituting a complete characterization of the thermodynamic properties of this batch of slurry.

[0063] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A process quality control system for physical indicators in the production of soybean products, characterized in that, include: The state calibration module introduces a first steam pulse into the empty tank before feeding to calibrate the heat transfer efficiency coefficient of the equipment based on the temperature rise of the tank. After feeding, during the cold resting period, a second steam pulse is introduced to identify the equivalent total heat capacity of the slurry based on the temperature rise of the slurry and calculate the theoretical minimum heat requirement. The judgment module is used to shield the slurry sensor during the start-up period of slurry cooking and only collect the steam condensate flow rate. When the cumulative mass of condensate reaches the lower limit of condensate mass corresponding to the theoretical minimum heat requirement, and the second derivative of the instantaneous flow rate of condensate shows the preset inflection point characteristic caused by irreversible endothermic mutation of protein, the start-up period termination signal is output and the time difference between the cumulative target time and the inflection point time is recorded. The phase transition trajectory generation module obtains the full-time curve of the instantaneous flow rate of condensate during the start-up period of the soy boiling process and extracts the morphological fingerprint parameter set from it. The parameter set includes at least the initial endothermic decay slope, the ratio of flow rate change before and after the endothermic abrupt change, and the recovery characteristic time constant. The parameter set and time difference are input into the trajectory mapping model to generate the protein ordered denaturation target trajectory. The trajectory includes the temperature and viscosity phase plane heating curve, the inflection point interval of protein secondary structure transformation, and the corresponding expected viscoelastic response characteristics. The phase change process control module restarts the slurry sensor after the start-up period ends, applies variable frequency and amplitude stirring excitation disturbance in the inflection zone of the target trajectory, and uses the viscoelastic response and the comparison result with the corresponding expected characteristics to gate the trajectory advancement. It also compares the real-time temperature and viscosity phase plane trajectory with the target trajectory shape. If the deviation exceeds the preset corridor, the power is increased by lag; if it is ahead, the attenuation amount is pre-calculated based on the heat transfer efficiency coefficient and attenuation is applied before reaching the upper limit of the preset corridor. The model self-correction module uses the batch-to-batch evolution trend of morphological fingerprint parameter groups to perform targeted corrections on the trajectory mapping model and the slurry equivalent total heat capacity identification model.

2. The process quality control system for physical indicators in the production of soybean products according to claim 1, characterized in that, The model self-correction module acquires the morphological fingerprint parameter set after the start-up period of each batch of soy milk. It compares the initial endothermic decay slope, the ratio of flow rate change before and after endothermic mutation, and the recovery characteristic time constant with the dynamic baseline constructed by the statistical characteristics of the corresponding parameters of a preset number of historical batches. The dynamic baseline is the moving average or exponentially weighted average of the corresponding parameters within a preset number of historical batches. When the initial heat absorption attenuation slope drifts beyond the threshold in the same direction relative to the number of consecutive preset batches of its dynamic baseline, the mapping coefficients related to the heat transfer rate in the trajectory mapping model are corrected, and the correction results are synchronously written back to the state calibration module for the first steam pulse calibration of the next batch. When the change in flow rate before and after the heat absorption mutation exceeds the threshold in the same direction as the number of consecutive preset batches, the mapping coefficients related to the heat capacity of the slurry in the trajectory mapping model are corrected. When the number of consecutive preset batches of the recovery characteristic time constant exceeds the threshold in the same direction and the initial heat absorption attenuation slope and flow rate change ratio do not show a trend of drift, the identification model of the equivalent total heat capacity of the slurry in the state calibration module is corrected. The correction of the mapping coefficients is carried out in a gradual manner, with the correction amount in a single batch not exceeding a preset proportion of the current coefficient value. The correction of the identification model is carried out by updating the model parameters.

3. The process quality control system for physical indicators in the production of soybean products according to claim 1, characterized in that, The trajectory mapping model generates the protein's ordered denaturation target trajectory according to the following mapping rules: The initial endothermic decay slope is used to map the upper limit of the heating rate in the first segment of the target trajectory. The larger the decay slope, the higher the upper limit of the heating rate and the shorter the duration of the first segment. The width and curvature of the trajectory inflection point interval are mapped by the ratio of flow change. The larger the ratio of change, the wider the inflection point interval and the smaller the curvature, which corresponds to a smoother process of protein secondary structure transformation. The duration of the stable period in the latter part of the target trajectory is mapped by the recovery feature time constant. The larger the time constant, the longer the stable period. The time difference is used to map the overall offset of each inflection point interval on the time axis. The larger the time difference, the more the overall inflection point interval shifts backward. The expected viscoelastic response characteristics are determined jointly based on the protein secondary structure transition type corresponding to the current inflection point interval and the flow rate change ratio. The larger the flow rate change ratio, the longer the expected stress relaxation time and the lower the crossover frequency of the expected storage modulus and loss modulus.

4. The process quality control system for physical indicators in the production of soybean products according to claim 1, characterized in that, The frequency range and amplitude range of the variable frequency and amplitude stirring excitation disturbance are determined by the protein secondary structure transition type corresponding to the current inflection point interval. The interval from ordered structure to random coil transition adopts a first frequency range and a first amplitude range, while the interval from random coil to gel state transition adopts a second frequency range lower than the first frequency range and a second amplitude range higher than the first amplitude range. The viscoelastic response characteristic parameters include at least one of stress relaxation time and the cross frequency of storage modulus and loss modulus, wherein the stress relaxation time is the time required for the slurry shear stress to decay to a preset proportion of the initial value after the stirring excitation disturbance is removed. The comparison involves comparing the deviation between the collected viscoelastic response characteristic parameters and the expected viscoelastic response characteristic parameters corresponding to the inflection point interval with a preset allowable deviation. When the deviation is within the allowable range, it is determined that the protein structure transformation has been completed and the trajectory is allowed to enter the next stage. When the deviation exceeds the allowable range, the current trajectory stage is maintained and stirring excitation is continuously applied until the deviation falls back to the allowable range.

5. The process quality control system for physical indicators in the production of soybean products according to claim 1, characterized in that, The power attenuation of advance suppression is jointly determined by the heat transfer efficiency coefficient, the current heating power, the advance offset between the real-time phase plane trajectory and the target trajectory, and the advance offset rate. The attenuation is directly proportional to the product of the advance offset and the advance offset rate, inversely proportional to the heat transfer efficiency coefficient, and multiplied by a preset safety factor. This ensures that the trajectory advancement amount generated by the expected heating rate corresponding to the attenuated heating power within the preset time window is not greater than the current advance offset, so that the real-time trajectory stops further approaching the preset corridor upper limit after the attenuation takes effect. The preset time window is determined based on the equivalent total heat capacity of the slurry and the current heating power; the larger the equivalent total heat capacity, the longer the time window.

6. The process quality control system for physical indicators in the production of soybean products according to claim 1, characterized in that, The preset inflection point feature is that the second derivative of the instantaneous condensate flow rate crosses a preset negative threshold from positive within a preset detection time window, and the duration for which the first derivative remains negative after the crossing exceeds a preset minimum duration. The preset detection time window starts at the moment when the accumulated condensate mass reaches the lower mass limit, and the window length is determined based on the theoretical minimum heat requirement. The preset negative threshold and the preset minimum duration are dynamically determined based on the equivalent total heat capacity of the slurry and the theoretical minimum heat requirement. The larger the equivalent total heat capacity, the smaller the absolute value of the preset negative threshold and the longer the preset minimum duration, so that the inflection point feature has adaptive detection sensitivity for slurries with different protein contents and different filling volumes.

7. The process quality control system for physical indicators in the production of soybean products according to claim 1, characterized in that, The heat transfer efficiency coefficient is obtained by normalizing the ratio of the input heat of the first steam pulse to the temperature rise of the tank after normalizing it with the known heat capacity of the tank, and it characterizes the effectiveness of the current batch of equipment in transferring steam heat to the medium inside the tank; the equivalent total heat capacity of the slurry is obtained by correcting the ratio of the input heat of the second steam pulse to the temperature rise of the slurry with the heat transfer efficiency coefficient; the theoretical minimum heat requirement is the product of the equivalent total heat capacity of the slurry and the difference between the target endpoint temperature and the initial cold temperature; the first steam pulse and the second steam pulse have different settings in terms of pulse duration and steam pressure, with the duration of the first steam pulse being shorter than that of the second steam pulse and the steam pressure being higher than that of the second steam pulse.

8. The process quality control system for physical indicators in the production of soybean products according to claim 1, characterized in that, The phase transition trajectory generation module also extracts the inflection point region time domain width parameter from the full-time curve of the instantaneous flow rate of the condensate. The inflection point region time domain width parameter is the duration from when the second derivative of the instantaneous flow rate of the condensate deviates from the normal decay trajectory to when it crosses the preset negative threshold and recovers to the preset proportion before crossing. It characterizes the degree of spatial non-uniformity of the irreversible endothermic mutation of the protein advancing layer by layer from the wall to the center in this batch of slurry.

9. The process quality control system for physical indicators in the production of soybean products according to claim 8, characterized in that, The trajectory mapping model adaptively scales the preset corridor width of each inflection point interval according to the time-domain width parameter of the inflection point region; the larger the time-domain width parameter, the wider the corridor width.

10. The process quality control system for physical indicators in the production of soybean products according to claim 9, characterized in that, The phase transition process control module performs the morphological comparison using the scaled corridor width.

Images