A food process parameter self-adaptive adjustment method and system

Abstract

This invention relates to the field of industrial control technology, specifically to a method and system for adaptive adjustment of food processing parameters. The adaptive adjustment system for food processing parameters includes: a cooling data acquisition module, a carcass heat disturbance analysis module, a predictive analysis module, and a cooling action stage replacement module. This invention constructs a time-windowed monitoring mechanism based on equivalent cooling output and temperature values ​​during carcass cooling. It generates a baseline equivalent cooling output sequence unaffected by carcass metabolism using a balanced equivalent cooling output function and a dynamic response coefficient function. Based on this, a carcass heat disturbance value sequence is obtained through inversion, thereby achieving a quantitative characterization of carcass metabolic exothermic behavior. Simultaneously, it performs short-term predictions of the carcass heat disturbance value sequence and temperature value sequence. Furthermore, under metabolic stability constraints such as upper metabolic limit, lower metabolic limit, and temperature threshold, it adaptively replaces and adjusts pre-set carcass cooling action stages.

Description

Technical Field

[0001] This invention relates to the field of industrial control technology, specifically to a method and system for adaptive adjustment of food processing parameters. Background Technology

[0002] In the slaughtering and processing of livestock and poultry, carcass cooling is a key technological step to ensure meat quality and food safety. Its purpose is to lower the carcass temperature to a safe range in a short period of time to inhibit microbial growth and delay quality deterioration. Existing carcass cooling processes typically control parameters such as ambient temperature, wind speed, and humidity through refrigeration equipment in a cooling chamber to achieve the cooling treatment of the carcass.

[0003] However, during actual cooling, carcasses continue to engage in cellular metabolic activities for some time after slaughter, consuming ATP and releasing metabolic heat. This metabolic heat release constitutes one of the main external heat loads that the cooling system needs to offset. Due to differences in breed, weight, physiological state, and pre-slaughter stress levels among different carcasses, the intensity and duration of their metabolic heat release exhibit significant uncertainty and individual variability, resulting in a significant temporal fluctuation in the carcass heat load during cooling.

[0004] In existing technologies, carcass cooling often employs pre-set staged cooling strategies or fixed process parameters. For example, high-intensity heat exchange stages and low-intensity heat exchange stages are set based on experience, and predetermined control parameters are kept constant during the cooling process, or simple feedback adjustments are made based solely on the cooling chamber temperature. These methods typically struggle to distinguish between changes in the cooling system's own output and load changes caused by carcass metabolic heat release. They also lack effective modeling and prediction capabilities for the time-varying trends of carcass metabolic heat, easily leading to a mismatch between cooling intensity and the actual metabolic state of the carcass. Summary of the Invention

[0005] This invention constructs a time-windowed monitoring mechanism based on equivalent cooling output and temperature values ​​during carcass cooling. It generates a baseline equivalent cooling output sequence unaffected by carcass metabolism using a balanced equivalent cooling output function and a dynamic response coefficient function. Based on this, a carcass heat perturbation value sequence is derived, thus achieving a quantitative characterization of carcass metabolic exothermic behavior. Simultaneously, a Transformer model is used to perform short-term predictions of the carcass heat perturbation and temperature value sequences. Furthermore, under metabolic stability constraints such as upper and lower metabolic limits and temperature thresholds, the pre-set carcass cooling action stages are adaptively replaced and adjusted. This allows the cooling system to suppress prolonged high metabolism in the carcass, reduce the risk of microbial growth, and avoid stunting due to excessively low temperatures while the carcass is still in the metabolic stage. A replacement fragment smoothing mechanism based on tag neighborhood consistency prevents frequent switching of cooling actions, achieving an adaptive adjustment effect of food processing parameters that balances carcass quality and safety with cooling stability.

[0006] This invention provides a method for adaptive adjustment of food processing parameters, comprising: During the carcass cooling process, the equivalent cooling output of the cooling system within a preset time window is collected at intervals and spliced ​​together in chronological order to form an equivalent cooling output sequence. The temperature values ​​of the cooling chamber within the preset time window are also collected at intervals and spliced ​​together in chronological order to form a temperature value sequence. A monitoring time point is set at each preset time window, and each monitoring time point corresponds to an equivalent cooling output sequence. The current monitoring time point is used to determine the cooling action stage. The overlap between the time range corresponding to the current monitoring time point and the preset time window corresponding to the current monitoring time point is recorded as the carcass disturbance analysis window. The baseline equivalent cooling output sequence within the carcass disturbance analysis window is determined according to the current monitoring time point. The time of the baseline equivalent cooling output sequence and the part of the equivalent cooling output sequence belonging to the carcass disturbance analysis window are aligned and subtracted item by item to obtain the carcass heating disturbance value sequence. The carcass heating disturbance value sequence includes the carcass heating disturbance value marked by the timestamp. The carcass heating perturbation value sequence and temperature value sequence are fed into the carcass heating perturbation value prediction network and temperature value prediction network respectively for processing, to obtain the predicted carcass heating perturbation value sequence and predicted temperature value sequence corresponding to the next preset time window respectively; Based on the predicted carcass heat disturbance value sequence and the predicted temperature value sequence, the cooling action phase of the next preset time window is replaced and adjusted by the constraint of metabolic stability.

[0007] Preferably, based on the predicted carcass heat perturbation value sequence and the predicted temperature value sequence, the cooling action phase of the next preset time window is replaced and adjusted by the constraint of metabolic stability, specifically including the following steps: Obtain the balanced equivalent cooling output function and dynamic response coefficient function corresponding to the cooling action stage. The balanced equivalent cooling output function uses the temperature value as the independent variable and the balanced equivalent cooling output as the dependent variable. The dynamic response coefficient function uses the temperature value as the independent variable and the dynamic response coefficient as the dependent variable. For each timestamp within the body perturbation analysis window, the following steps are performed: obtain the temperature value of the timestamp in the temperature value sequence, and substitute the obtained temperature value into the equilibrium equivalent cooling output function and dynamic response coefficient function corresponding to the cooling action stage to obtain the corresponding equilibrium equivalent cooling output and dynamic response coefficient. The output obtained by subtracting the equilibrium equivalent cooling output from the baseline equivalent cooling output of the previous timestamp is recorded as the steady-state deviation. Then, the product of the steady-state deviation and the dynamic response coefficient is calculated and summed with the baseline equivalent cooling output of the previous timestamp to obtain the baseline equivalent cooling output corresponding to the current timestamp.

[0008] Preferably, the balanced equivalent cooling output function and the dynamic response coefficient function are constructed in the following manner: Collect several historical datasets corresponding to the cooling action phase. The historical datasets are labeled with temperature values ​​and include several equivalent cooling outputs arranged in chronological order. For each historical dataset, find consecutive timestamps where the standard deviation of the equivalent cooling output is less than the fluctuation threshold, and use the average of all equivalent cooling outputs corresponding to the consecutive timestamps as the balanced equivalent cooling fitting output. Combine the temperature values ​​corresponding to the historical datasets with the balanced equivalent cooling fitting output to form balanced equivalent cooling fitting points. Then, perform function fitting on the balanced equivalent cooling fitting points corresponding to all historical datasets for the cooling action stage to obtain the balanced equivalent cooling output function. For each timestamp in the historical dataset, the temperature value corresponding to the timestamp is substituted into the balanced equivalent cooling output function to obtain the balanced equivalent cooling output corresponding to the timestamp. The difference between the balanced equivalent cooling output and the equivalent cooling output corresponding to the timestamp is recorded as the steady-state transformation quantity. If the steady-state transformation quantity is higher than the change threshold, the difference between the equivalent cooling output corresponding to the next timestamp and the equivalent cooling output corresponding to the current timestamp is calculated, and the ratio between the obtained difference and the steady-state transformation quantity is recorded as the local dynamic response coefficient. The average value of all local dynamic response coefficients in the historical dataset is recorded as the dynamic response fitting coefficient. The temperature value corresponding to the historical dataset and the balanced dynamic response fitting coefficient form the dynamic response fitting point. Then, the dynamic response fitting points corresponding to all historical datasets corresponding to the cooling action stage are fitted with a function to obtain the dynamic response coefficient function.

[0009] Preferably, based on the predicted carcass heat perturbation value sequence and the predicted temperature value sequence, the cooling action phase of the next preset time window is replaced and adjusted by the constraint of metabolic stability, specifically including the following: Iterate through all timestamps in the next preset time window. For each timestamp, if the predicted carcass heat perturbation value corresponding to the timestamp is higher than the upper metabolic threshold, mark the timestamp as a high metabolism label. If the predicted carcass heat perturbation value corresponding to the timestamp is not higher than the upper metabolic threshold, determine whether the predicted carcass heat perturbation value corresponding to the timestamp is higher than the lower metabolic threshold. Then, judge the predicted temperature value corresponding to the timestamp. If the predicted temperature value corresponding to the timestamp is lower than the temperature threshold, mark the timestamp as a low temperature label. If the predicted temperature value corresponding to the timestamp is not lower than the temperature threshold, mark the timestamp as a maintenance label. If the predicted carcass heat perturbation value corresponding to the timestamp is not higher than the lower metabolic threshold, mark the timestamp as a heat exchange label. High metabolism label, low temperature label, and heat exchange label are all label values. Iterate through all timestamps of the next preset time window. For each timestamp, select the label with the highest frequency in the neighborhood window corresponding to the timestamp and record it as the target label. If the label corresponding to the timestamp is the same as the target label, no operation is performed. If the label corresponding to the timestamp is different from the target label, replace the label corresponding to the timestamp with the target label. The neighborhood window is a combination of timestamps that are in the same fixed time range as the timestamp. The time range defined by the timestamps corresponding to consecutive identical labels is recorded as the replacement fragment. The replacement fragment is marked by the label. If the label corresponding to the replacement fragment is a high-metabolism label, the time range of the replacement fragment is set as a high-gradient heat exchange action. If the label corresponding to the replacement fragment is a low-temperature label, the time range of the replacement fragment is set as a low-gradient heat exchange action. If the label corresponding to the replacement fragment is a heat exchange label, the time range of the replacement fragment is set as a short-term heat exchange inhibition action.

[0010] Preferably, the carcass heat perturbation prediction network and the temperature prediction network are trained in the following ways: Several training samples of carcass heat perturbation values ​​are obtained. The training samples of carcass heat perturbation values ​​include the sequence of carcass heat perturbation values. All the training samples of carcass heat perturbation values ​​are combined into a training set of carcass heat perturbation values. The carcass heat perturbation value prediction network is trained using the training set of carcass heat perturbation values. During the training period, the training samples of carcass heat perturbation values ​​are used as input, and the training samples of carcass heat perturbation values ​​in the next preset time window corresponding to the training samples of carcass heat perturbation values ​​are used as the target output. Several temperature value training samples are obtained, including temperature value sequences. All temperature value training samples are combined into a temperature value training set. The temperature value prediction network is trained using the temperature value training set. During training, the temperature value training samples are used as input, and the temperature value training samples corresponding to the next preset time window are used as the target output.

[0011] Preferably, both the carcass heating perturbation value prediction network and the temperature value prediction network adopt the Transformer model.

[0012] The present invention also provides a food processing parameter adaptive adjustment system, comprising: The cooling data acquisition module is used to collect the equivalent cooling output of the cooling system within a preset time window at intervals during the body cooling process, and to splice the data into an equivalent cooling output sequence in chronological order. It also collects the temperature values ​​of the cooling chamber within the preset time window at intervals and splices them into a temperature value sequence in chronological order. A monitoring time point is set for each preset time window, and each monitoring time point corresponds to an equivalent cooling output sequence. The carcass heating disturbance analysis module is used to obtain the cooling action stage at the current monitoring time point. The overlap between the time range corresponding to the cooling action stage at the current monitoring time point and the preset time window corresponding to the current monitoring time point is recorded as the carcass disturbance analysis window. The module determines the benchmark equivalent cooling output sequence within the carcass disturbance analysis window based on the cooling action stage at the current monitoring time point. Then, the module aligns the time of the benchmark equivalent cooling output sequence and the part of the equivalent cooling output sequence belonging to the carcass disturbance analysis window and performs subtraction item by item to obtain the carcass heating disturbance value sequence. The carcass heating disturbance value sequence includes the carcass heating disturbance value marked by the timestamp. The predictive analysis module is used to send the carcass heat perturbation value sequence and temperature value sequence into the carcass heat perturbation value prediction network and temperature value prediction network respectively for processing, and obtain the predicted carcass heat perturbation value sequence and predicted temperature value sequence corresponding to the next preset time window respectively; The cooling action phase replacement module is used to replace and adjust the cooling action phase in the next preset time window based on the predicted carcass heat disturbance value sequence and the predicted temperature value sequence, and by constraining metabolic stability.

[0013] The present invention has the following advantages: This invention constructs a time-windowed monitoring mechanism based on equivalent cooling output and temperature values ​​during carcass cooling. It generates a baseline equivalent cooling output sequence unaffected by carcass metabolism using a balanced equivalent cooling output function and a dynamic response coefficient function. Based on this, a carcass heat perturbation value sequence is derived, thus achieving a quantitative characterization of carcass metabolic exothermic behavior. Simultaneously, a Transformer model is used to perform short-term predictions of the carcass heat perturbation and temperature value sequences. Furthermore, under metabolic stability constraints such as upper and lower metabolic limits and temperature thresholds, the pre-set carcass cooling action stages are adaptively replaced and adjusted. This allows the cooling system to suppress prolonged high metabolism in the carcass, reduce the risk of microbial growth, and avoid stunting due to excessively low temperatures while the carcass is still in the metabolic stage. A replacement fragment smoothing mechanism based on tag neighborhood consistency prevents frequent switching of cooling actions, achieving an adaptive adjustment effect of food processing parameters that balances carcass quality and safety with cooling stability. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the adaptive adjustment system for food processing parameters used in an embodiment of the present invention. Detailed Implementation

[0015] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0016] Example 1: A method for adaptive adjustment of food processing parameters, comprising: During the carcass cooling process, the equivalent cooling output of the cooling system within a preset time window is collected at intervals and then pieced together in chronological order to form an equivalent cooling output sequence. The length of the preset time window is set by the operator. Temperature values ​​of the cooling chamber within the preset time window are also collected at intervals and then pieced together in chronological order to form a temperature value sequence. It should be noted that during the carcass cooling process, each carcass is cooled in its corresponding cooling chamber, and each cooling chamber is equipped with a cooling system. The cooling system uses a variable frequency compressor for cooling operation. The equivalent cooling output refers to the cooling... The output power of the compressor in the system and the equivalent cooling output represent the equivalent cooling capacity output by the cooling system to counteract external heat load disturbances. Since carcass metabolic heat release is the main external heat load during carcass cooling, the equivalent cooling output also includes information on carcass cooling metabolic heat release. Here, carcass cooling refers to the forced cooling process of the whole carcass after slaughter and before dissection. During cooling, the cells inside the carcass will still consume ATP and release heat metabolically. A monitoring time point is set at each preset time window, and each monitoring time point corresponds to an equivalent cooling output sequence. The cooling action stage at the current monitoring time point is determined. The overlap between the time range corresponding to the current monitoring time point's cooling action stage and the preset time window corresponding to the current monitoring time point is recorded as the carcass disturbance analysis window. A baseline equivalent cooling output sequence within the carcass disturbance analysis window is determined based on the current monitoring time point's cooling action stage. Then, the time portions belonging to the carcass disturbance analysis window in the baseline equivalent cooling output sequence and the equivalent cooling output sequence corresponding to the monitoring time point are aligned and subtracted item by item to obtain the carcass heat generation disturbance value sequence. The carcass heat generation disturbance value sequence includes carcass heat generation disturbance values ​​marked with timestamps. It should be noted that the cooling action stage refers to the cooling mode set by the operator for the cooling system, including low-gradient heat exchange action, high-gradient heat exchange action, and short-term heat exchange suppression action. Each cooling action stage has control parameters set by the operator; for example, the wind speed in low-gradient heat exchange action is set to 0.3– The air velocity is set to 0.8 m / s, relative humidity is controlled at 80–90%, and compressor power is set to 30–60% of rated power for slow heat exchange. In the high-gradient heat exchange action, the air velocity is set to 1.5–3.0 m / s, relative humidity is controlled at 65–80%, and compressor power is set to 70–100% of rated power for rapid heat removal. In the short-term heat exchange suppression action, the air velocity is set to less than 0.2 m / s, relative humidity is not controlled, and compressor power is less than 30% to buffer the temperature difference between the inside and outside of the carcass. Each cooling action stage has a corresponding duration, which is sent to the cooling system as an instruction for carcass cooling. Before carcass cooling, a carcass cooling strategy is pre-set. The carcass cooling strategy is a combination of cooling action stages, generally set to 2 hours for high-gradient heat exchange and 4 hours for high-gradient heat exchange. The time corresponding to each cooling action stage is much longer than the preset time window. Generally, the carcass disturbance analysis window is the preset time window. The carcass heating perturbation value sequence and temperature value sequence are fed into the carcass heating perturbation value prediction network and temperature value prediction network, respectively, for processing. The predicted carcass heating perturbation value sequence and predicted temperature value sequence corresponding to the next preset time window are obtained, respectively. It should be noted that the predicted carcass heating perturbation value sequence includes the predicted carcass heating perturbation value corresponding to each timestamp within the next preset time window. Under the action of the cooling system, the carcass heating perturbation value and temperature value will show a temporal change pattern in a short period of time, which can realize the temporal prediction within a short period of time. The predicted temperature value sequence includes the predicted temperature value corresponding to each timestamp within the next preset time window. Both the carcass heating perturbation value prediction network and the temperature value prediction network adopt the Transformer model. Based on the predicted carcass heat disturbance value sequence and the predicted temperature value sequence, the cooling action stage of the next preset time window is replaced and adjusted by the constraint of metabolic stability, so that the carcass heat disturbance value cannot be higher than the set threshold for a long time, thus reducing the growth of microorganisms. Furthermore, while the carcass is still in the process of metabolism, the temperature in the cooling chamber is not lower than the set threshold, thus avoiding the phenomenon of cold stiffness. This completes the adaptive adjustment of process parameters during the carcass cooling process. Based on the predicted carcass heat perturbation value sequence and the predicted temperature value sequence, the cooling action phase of the next preset time window is replaced and adjusted by the constraint of metabolic stability. The specific steps include the following: Obtain the balanced equivalent cooling output function and dynamic response coefficient function corresponding to the cooling action stage. The balanced equivalent cooling output function uses the temperature value as the independent variable and the balanced equivalent cooling output as the dependent variable. The balanced equivalent cooling output is used to reflect the steady-state cooling output capability under a fixed temperature and no heat load. The dynamic response coefficient function uses the temperature value as the independent variable and the dynamic response coefficient as the dependent variable. The dynamic response coefficient is used to reflect the rate at which the system changes towards steady-state cooling output capability. For each timepoint within the carcass disturbance analysis window, the following steps are performed: The temperature value of the timepoint in the temperature value sequence is obtained, and this temperature value is substituted into the equilibrium equivalent cooling output function and dynamic response coefficient function corresponding to the cooling action stage to obtain the corresponding equilibrium equivalent cooling output and dynamic response coefficient. The output obtained by subtracting the equilibrium equivalent cooling output from the baseline equivalent cooling output of the previous timepoint is recorded as the steady-state deviation. The product of the steady-state deviation and the dynamic response coefficient is then calculated and summed with the baseline equivalent cooling output of the previous timepoint to obtain the baseline equivalent cooling output corresponding to the current timepoint. Furthermore, the calculation of the baseline equivalent cooling output can span across cooling action stages, tracing back to the equivalent cooling output before the carcass was placed in the cooling chamber. The equivalent cooling output function and the dynamic response coefficient function are balanced and constructed as follows: Several historical datasets corresponding to the cooling action phase are collected. The historical datasets are marked by temperature values ​​and include several equivalent cooling outputs arranged in chronological order. It should be noted that when collecting the historical datasets, the cooling action phase and the temperature value in the fixed cooling chamber are set. Without applying heat load, the equivalent cooling outputs of the cooling system are collected to form the historical datasets. For each historical dataset, find consecutive timestamps where the standard deviation of the equivalent cooling output is less than the fluctuation threshold. Use the average of all equivalent cooling outputs corresponding to these consecutive timestamps as the balanced equivalent cooling fitting output. Combine the temperature values ​​corresponding to the historical datasets with the balanced equivalent cooling fitting output to form balanced equivalent cooling fitting points. Then, perform function fitting on the balanced equivalent cooling fitting points corresponding to all historical datasets for each cooling action stage to obtain the balanced equivalent cooling output function. Polynomial fitting can be used for fitting. For each timestamp in the historical dataset, the temperature value corresponding to the timestamp is substituted into the balanced equivalent cooling output function to obtain the balanced equivalent cooling output corresponding to the timestamp. The difference between the balanced equivalent cooling output and the equivalent cooling output corresponding to the timestamp is recorded as the steady-state transformation quantity. If the steady-state transformation quantity is higher than the change threshold (set by the operator), the difference between the equivalent cooling output corresponding to the next timestamp and the equivalent cooling output corresponding to the current timestamp is calculated. The ratio between the obtained difference and the steady-state transformation quantity is recorded as the local dynamic response coefficient. The average value of all local dynamic response coefficients in the historical dataset is recorded as the dynamic response fitting coefficient. The temperature value corresponding to the historical dataset and the balanced dynamic response fitting coefficient form the dynamic response fitting point. Then, the dynamic response fitting points corresponding to all historical datasets corresponding to the cooling action stage are fitted with a function to obtain the dynamic response coefficient function. Based on the predicted carcass heat perturbation value sequence and the predicted temperature value sequence, the cooling action phase in the next preset time window is replaced and adjusted by the constraint of metabolic stability, specifically including the following: Iterate through all timestamps in the next preset time window. For each timestamp, if the predicted carcass heat disturbance value corresponding to the timestamp is higher than the upper metabolic threshold (set by the operator), the timestamp is marked as high-metabolic. If the predicted carcass heat disturbance value is consistently higher than the upper metabolic threshold, the carcass will experience prolonged high temperatures after metabolism, promoting microbial growth and damaging the quality of the cooled carcass. If the predicted carcass heat disturbance value corresponding to the timestamp is not higher than the upper metabolic threshold, determine if the predicted carcass heat disturbance value is higher than the lower metabolic threshold (set by the operator). The lower metabolic threshold reflects whether there is still ATP available for metabolism within the carcass. If the predicted carcass heat disturbance value is higher than the lower metabolic threshold, it indicates that metabolism is still occurring within the carcass. The time stamp is marked as an "over-low temperature" tag if the predicted temperature value corresponding to the time stamp is lower than the temperature threshold (set by the operator, typically 15 degrees Celsius). This indicates that the carcass is undergoing metabolism while the temperature is too low. In this case, the muscle fibers inside the carcass will contract, leading to stiffness and affecting the quality of the cooled carcass. If the predicted temperature value corresponding to the time stamp is not lower than the temperature threshold, the time stamp is marked as a "maintain" tag. If the predicted carcass heat disturbance value corresponding to the time stamp is not higher than the lower metabolic limit threshold, it indicates that the carcass may no longer be undergoing metabolism at that time stamp, or that the heat exchange between the carcass and the outside may be inadequate, resulting in the heat disturbance value not affecting the equivalent cooling output. In this case, the time stamp is marked as a "heat exchange" tag. The high metabolism tag, over-low temperature tag, and heat exchange tag are all label values. Iterate through all timestamps in the next preset time window. For each timestamp, select the label with the highest frequency in the neighborhood window corresponding to the timestamp and record it as the target label. If the label corresponding to the timestamp is the same as the target label, no operation is performed. If the label corresponding to the timestamp is different from the target label, it means that the label corresponding to the timestamp does not conform to the distribution of the surrounding area. By replacing the label corresponding to the timestamp with the target label, the label corresponding to all timestamps can be evenly distributed, avoiding the appearance of one or two discrete labels in a group of consecutive identical labels. The neighborhood window is a combination of timestamps within a fixed time range. Specifically, assuming that the fixed time range includes 2C timestamps, the neighborhood window corresponding to the timestamp is a combination of the C timestamps before and the C timestamps after the currently selected timestamp. If the timestamp is located at the edge of the preset time window, making it impossible to select all C timestamps before or after, then select more timestamps from the other side to make up the difference. The time range defined by the timestamps corresponding to consecutive identical labels is recorded as the replacement fragment, and the replacement fragment is marked by the label. If the label corresponding to the replacement fragment is a high-metabolism label, the time range of the replacement fragment is set as a high-gradient heat transfer action. If the label corresponding to the replacement fragment is a low-temperature label, the time range of the replacement fragment is set as a low-gradient heat transfer action. If the label corresponding to the replacement fragment is a heat exchange label, the time range of the replacement fragment is set as a short-term heat exchange inhibition action. It should be noted that the replacement operation refers to the replacement of the pre-set carcass cooling strategy. If the label corresponding to the replacement fragment is a high-metabolism label, and the replacement fragment is already a high-gradient heat transfer action in the pre-set carcass cooling strategy, then no replacement operation actually occurs. The carcass heat perturbation prediction network and the temperature prediction network were trained in the following ways: Several training samples of carcass heat perturbation values ​​are obtained. These training samples include carcass heat perturbation value sequences. Specifically, the operators conduct experiments with carcass heat load in the cooling chamber and actually collect equivalent cooling output sequences and construct corresponding carcass heat perturbation value sequences. All training samples are combined into a carcass heat perturbation value training set. The carcass heat perturbation value prediction network is trained using this training set. During training, the training samples are used as input, and the next preset time window's training samples are used as the target output. A loss value is constructed based on the difference between the predicted output of the carcass heat perturbation value prediction network and the corresponding target output. The construction method can use MSE (Mean Sequence of Change). Inverse gradient descent is performed based on the loss value to adjust the parameters within the carcass heat perturbation value prediction network until the accuracy of the network reaches the expected level. Several temperature value training samples are obtained, including temperature value sequences. Specifically, the temperature value sequences are collected by operators during experiments with body heat load in the cooling chamber. All temperature value training samples are combined into a temperature value training set. The temperature value prediction network is trained using the temperature value training set. During training, the temperature value training samples are used as input, and the temperature value training samples of the next preset time window corresponding to the temperature value training samples are used as the target output. A loss value is constructed based on the difference between the predicted output of the temperature value prediction network and the corresponding target output. The construction method can be MSE. Based on the loss value, inverse gradient descent is performed to adjust the parameters in the temperature value prediction network until the accuracy of the temperature value prediction network reaches the expected level. This application constructs a time-windowed monitoring mechanism based on equivalent cooling output and temperature values ​​during carcass cooling. It generates a baseline equivalent cooling output sequence unaffected by carcass metabolism using a balanced equivalent cooling output function and a dynamic response coefficient function. Based on this, a carcass heat perturbation value sequence is obtained, thus achieving a quantitative characterization of carcass metabolic exothermic behavior. Simultaneously, a Transformer model is used to perform short-term predictions of the carcass heat perturbation and temperature value sequences. Furthermore, under metabolic stability constraints such as upper and lower metabolic limits and temperature thresholds, the application adaptively replaces and adjusts pre-set carcass cooling action stages. This allows the cooling system to suppress long-term high metabolism in the carcass, reduce the risk of microbial growth, and avoid stunting due to excessively low temperatures while the carcass is still in the metabolic stage. A replacement fragment smoothing mechanism based on label neighborhood consistency prevents frequent switching of cooling actions, achieving an adaptive adjustment effect of food processing parameters that balances carcass quality and safety with cooling stability.

[0017] Example 2: An adaptive adjustment system for food processing parameters, such as... Figure 1 As shown, it includes: The cooling data acquisition module is used to collect the equivalent cooling output of the cooling system within a preset time window at intervals during the carcass cooling process. These data are then concatenated into a sequence of equivalent cooling output values ​​in chronological order. The length of the preset time window is set by the operator. The module also collects the temperature values ​​of the cooling chambers within the preset time window at intervals and concatenates these values ​​into a sequence of temperature values ​​in chronological order. It should be noted that during the carcass cooling process, each carcass is cooled in its corresponding cooling chamber, and each cooling chamber is equipped with a cooling system. The cooling system uses a variable frequency compressor for cooling operation. The equivalent cooling output... The equivalent cooling output refers to the output power of the compressor in the cooling system. The equivalent cooling output represents the equivalent cooling capacity output by the cooling system to counteract external heat load disturbances. Since carcass metabolic heat release is the main external heat load during carcass cooling, the equivalent cooling output also includes information on carcass cooling metabolic heat release. Here, carcass cooling refers to the forced cooling process of the whole carcass after slaughter and before butchering. During cooling, the cells inside the carcass will still consume ATP and release heat through metabolism. A monitoring time point is set at each preset time window, and each monitoring time point corresponds to an equivalent cooling output sequence. The carcass heating disturbance analysis module is used to obtain the cooling action stage at the current monitoring time point. The overlap between the time range corresponding to the cooling action stage at the current monitoring time point and the preset time window corresponding to the current monitoring time point is recorded as the carcass disturbance analysis window. Based on the cooling action stage at the current monitoring time point, a baseline equivalent cooling output sequence within the carcass disturbance analysis window is determined. Then, the time portions belonging to the carcass disturbance analysis window in the baseline equivalent cooling output sequence and the equivalent cooling output sequence corresponding to the monitoring time point are aligned and subtracted item by item to obtain the carcass heating disturbance value sequence. The carcass heating disturbance value sequence includes carcass heating disturbance values ​​marked with timestamps. It should be noted that the cooling action stage refers to the cooling mode set by the operator for the cooling system, including low-gradient heat exchange action, high-gradient heat exchange action, and short-term heat exchange suppression action. Each cooling action stage has control parameters set by the operator, such as wind speed in low-gradient heat exchange action. The air velocity is set to 0.3–0.8 m / s, relative humidity is controlled at 80–90%, and compressor power is set to 30–60% of rated power for slow heat exchange. In the high-gradient heat exchange action, the air velocity is set to 1.5–3.0 m / s, relative humidity is controlled at 65–80%, and compressor power is set to 70–100% of rated power for rapid heat removal. In the short-term heat exchange suppression action, the air velocity is set to less than 0.2 m / s, relative humidity is not controlled, and compressor power is less than 30% to buffer the temperature difference between the inside and outside of the carcass. Each cooling action stage has a corresponding duration, which is sent to the cooling system as an instruction for carcass cooling. Before carcass cooling, a carcass cooling strategy is pre-set. The carcass cooling strategy is a combination of cooling action stages, generally set to 2 hours for high-gradient heat exchange and 4 hours for high-gradient heat exchange. The time corresponding to each cooling action stage is much longer than the preset time window. Generally, the carcass disturbance analysis window is the preset time window. The predictive analysis module is used to feed the carcass heat perturbation value sequence and temperature value sequence into the carcass heat perturbation value prediction network and temperature value prediction network, respectively, for processing, to obtain the predicted carcass heat perturbation value sequence and predicted temperature value sequence corresponding to the next preset time window. It should be noted that the predicted carcass heat perturbation value sequence includes the predicted carcass heat perturbation value corresponding to each timestamp within the next preset time window. Under the action of the cooling system, the carcass heat perturbation value and temperature value will show a temporal change pattern in a short period of time, which can realize the temporal prediction within a short period of time. The predicted temperature value sequence includes the predicted temperature value corresponding to each timestamp within the next preset time window. Both the carcass heat perturbation value prediction network and the temperature value prediction network adopt the Transformer model. The cooling action phase replacement module is used to replace and adjust the cooling action phase in the next preset time window based on the predicted carcass heat disturbance value sequence and the predicted temperature value sequence, through the constraint of metabolic stability. This ensures that the carcass heat disturbance value cannot be higher than the set threshold for a long time, thereby reducing the growth of microorganisms. It also ensures that the temperature in the cooling chamber does not fall below the set threshold while the carcass is still in the process of metabolism, thus avoiding the phenomenon of stunted growth.

[0018] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for adaptive adjustment of food process parameters, characterized in that, include: During the carcass cooling process, the equivalent cooling output of the cooling system within a preset time window is collected at intervals and spliced ​​together in chronological order to form an equivalent cooling output sequence. The temperature values ​​of the cooling chamber within the preset time window are also collected at intervals and spliced ​​together in chronological order to form a temperature value sequence. A monitoring time point is set at each preset time window, and each monitoring time point corresponds to an equivalent cooling output sequence. The current monitoring time point is used to determine the cooling action stage. The overlap between the time range corresponding to the current monitoring time point and the preset time window corresponding to the current monitoring time point is recorded as the carcass disturbance analysis window. The baseline equivalent cooling output sequence within the carcass disturbance analysis window is determined according to the current monitoring time point. The time of the baseline equivalent cooling output sequence and the part of the equivalent cooling output sequence belonging to the carcass disturbance analysis window are aligned and subtracted item by item to obtain the carcass heating disturbance value sequence. The carcass heating disturbance value sequence includes the carcass heating disturbance value marked by the timestamp. The carcass heating perturbation value sequence and temperature value sequence are fed into the carcass heating perturbation value prediction network and temperature value prediction network respectively for processing, to obtain the predicted carcass heating perturbation value sequence and predicted temperature value sequence corresponding to the next preset time window respectively; Based on the predicted carcass heat disturbance value sequence and the predicted temperature value sequence, the cooling action phase of the next preset time window is replaced and adjusted by the constraint of metabolic stability.

2. The adaptive adjustment method for food processing parameters according to claim 1, characterized in that, Based on the predicted carcass heat perturbation value sequence and the predicted temperature value sequence, the cooling action phase of the next preset time window is replaced and adjusted by the constraint of metabolic stability. The specific steps include the following: Obtain the balanced equivalent cooling output function and dynamic response coefficient function corresponding to the cooling action stage. The balanced equivalent cooling output function uses the temperature value as the independent variable and the balanced equivalent cooling output as the dependent variable. The dynamic response coefficient function uses the temperature value as the independent variable and the dynamic response coefficient as the dependent variable. For each timestamp within the body perturbation analysis window, the following steps are performed: obtain the temperature value of the timestamp in the temperature value sequence, and substitute the obtained temperature value into the equilibrium equivalent cooling output function and dynamic response coefficient function corresponding to the cooling action stage to obtain the corresponding equilibrium equivalent cooling output and dynamic response coefficient. The output obtained by subtracting the equilibrium equivalent cooling output from the baseline equivalent cooling output of the previous timestamp is recorded as the steady-state deviation. Then, the product of the steady-state deviation and the dynamic response coefficient is calculated and summed with the baseline equivalent cooling output of the previous timestamp to obtain the baseline equivalent cooling output corresponding to the current timestamp.

3. The adaptive adjustment method for food processing parameters according to claim 2, characterized in that, The equivalent cooling output function and the dynamic response coefficient function are balanced and constructed as follows: Collect several historical datasets corresponding to the cooling action phase. The historical datasets are labeled with temperature values ​​and include several equivalent cooling outputs arranged in chronological order. For each historical dataset, find consecutive timestamps where the standard deviation of the equivalent cooling output is less than the fluctuation threshold, and use the average of all equivalent cooling outputs corresponding to the consecutive timestamps as the balanced equivalent cooling fitting output. Combine the temperature values ​​corresponding to the historical datasets with the balanced equivalent cooling fitting output to form balanced equivalent cooling fitting points. Then, perform function fitting on the balanced equivalent cooling fitting points corresponding to all historical datasets for the cooling action stage to obtain the balanced equivalent cooling output function. For each timestamp in the historical dataset, the temperature value corresponding to the timestamp is substituted into the balanced equivalent cooling output function to obtain the balanced equivalent cooling output corresponding to the timestamp. The difference between the balanced equivalent cooling output and the equivalent cooling output corresponding to the timestamp is recorded as the steady-state transformation quantity. If the steady-state transformation quantity is higher than the change threshold, the difference between the equivalent cooling output corresponding to the next timestamp and the equivalent cooling output corresponding to the current timestamp is calculated, and the ratio between the obtained difference and the steady-state transformation quantity is recorded as the local dynamic response coefficient. The average value of all local dynamic response coefficients in the historical dataset is recorded as the dynamic response fitting coefficient. The temperature value corresponding to the historical dataset and the balanced dynamic response fitting coefficient form the dynamic response fitting point. Then, the dynamic response fitting points corresponding to all historical datasets corresponding to the cooling action stage are fitted with a function to obtain the dynamic response coefficient function.

4. The adaptive adjustment method for food processing parameters according to claim 3, characterized in that, Based on the predicted carcass heat perturbation value sequence and the predicted temperature value sequence, the cooling action phase in the next preset time window is replaced and adjusted by the constraint of metabolic stability, specifically including the following: Iterate through all timestamps in the next preset time window. For each timestamp, if the predicted carcass heat perturbation value corresponding to the timestamp is higher than the upper metabolic threshold, mark the timestamp as a high metabolism label. If the predicted carcass heat perturbation value corresponding to the timestamp is not higher than the upper metabolic threshold, determine whether the predicted carcass heat perturbation value corresponding to the timestamp is higher than the lower metabolic threshold. Then, judge the predicted temperature value corresponding to the timestamp. If the predicted temperature value corresponding to the timestamp is lower than the temperature threshold, mark the timestamp as a low temperature label. If the predicted temperature value corresponding to the timestamp is not lower than the temperature threshold, mark the timestamp as a maintenance label. If the predicted carcass heat perturbation value corresponding to the timestamp is not higher than the lower metabolic threshold, mark the timestamp as a heat exchange label. High metabolism label, low temperature label, and heat exchange label are all label values. Iterate through all timestamps of the next preset time window. For each timestamp, select the label with the highest frequency in the neighborhood window corresponding to the timestamp and record it as the target label. If the label corresponding to the timestamp is the same as the target label, no operation is performed. If the label corresponding to the timestamp is different from the target label, replace the label corresponding to the timestamp with the target label. The neighborhood window is a combination of timestamps that are in the same fixed time range as the timestamp. The time range defined by the timestamps corresponding to consecutive identical labels is recorded as the replacement fragment. The replacement fragment is marked by the label. If the label corresponding to the replacement fragment is a high-metabolism label, the time range of the replacement fragment is set as a high-gradient heat exchange action. If the label corresponding to the replacement fragment is a low-temperature label, the time range of the replacement fragment is set as a low-gradient heat exchange action. If the label corresponding to the replacement fragment is a heat exchange label, the time range of the replacement fragment is set as a short-term heat exchange inhibition action.

5. The adaptive adjustment method for food processing parameters according to claim 4, characterized in that, The carcass heat perturbation prediction network and the temperature prediction network were trained in the following ways: Several training samples of carcass heat perturbation values ​​are obtained. The training samples of carcass heat perturbation values ​​include the sequence of carcass heat perturbation values. All the training samples of carcass heat perturbation values ​​are combined into a training set of carcass heat perturbation values. The carcass heat perturbation value prediction network is trained using the training set of carcass heat perturbation values. During the training period, the training samples of carcass heat perturbation values ​​are used as input, and the training samples of carcass heat perturbation values ​​in the next preset time window corresponding to the training samples of carcass heat perturbation values ​​are used as the target output. Several temperature value training samples are obtained, including temperature value sequences. All temperature value training samples are combined into a temperature value training set. The temperature value prediction network is trained using the temperature value training set. During training, the temperature value training samples are used as input, and the temperature value training samples corresponding to the next preset time window are used as the target output.

6. The adaptive adjustment method for food processing parameters according to claim 5, characterized in that, Both the carcass heating perturbation prediction network and the temperature prediction network use the Transformer model.

7. A food processing parameter adaptive adjustment system, characterized in that, The system employs an adaptive adjustment method for food processing parameters as described in any one of claims 1-6, comprising: The cooling data acquisition module is used to collect the equivalent cooling output of the cooling system within a preset time window at intervals during the body cooling process, and to splice the data into an equivalent cooling output sequence in chronological order. It also collects the temperature values ​​of the cooling chamber within the preset time window at intervals and splices them into a temperature value sequence in chronological order. A monitoring time point is set for each preset time window, and each monitoring time point corresponds to an equivalent cooling output sequence. The carcass heating disturbance analysis module is used to obtain the cooling action stage at the current monitoring time point. The overlap between the time range corresponding to the cooling action stage at the current monitoring time point and the preset time window corresponding to the current monitoring time point is recorded as the carcass disturbance analysis window. The module determines the benchmark equivalent cooling output sequence within the carcass disturbance analysis window based on the cooling action stage at the current monitoring time point. Then, the module aligns the time of the benchmark equivalent cooling output sequence and the part of the equivalent cooling output sequence belonging to the carcass disturbance analysis window and performs subtraction item by item to obtain the carcass heating disturbance value sequence. The carcass heating disturbance value sequence includes the carcass heating disturbance value marked by the timestamp. The predictive analysis module is used to send the carcass heat perturbation value sequence and temperature value sequence into the carcass heat perturbation value prediction network and temperature value prediction network respectively for processing, and obtain the predicted carcass heat perturbation value sequence and predicted temperature value sequence corresponding to the next preset time window respectively; The cooling action phase replacement module is used to replace and adjust the cooling action phase in the next preset time window based on the predicted carcass heat disturbance value sequence and the predicted temperature value sequence, and by constraining metabolic stability.

Images