A food quality management and control collaborative contract method based on quality stage identification

Abstract

This invention discloses a collaborative contract method for food quality control based on quality stage identification. It acquires multi-source data on fermentation and distribution, aligns, repairs, and normalizes the data over time; constructs comprehensive quality observations and their reliability; updates stage beliefs based on prior continuity and observational support, outputting the stage and confidence level; triggers information sharing based on the confidence level and adaptively adjusts information accuracy; generates quality status estimates and links them to input and pricing; and calculates parameters for benefit sharing, cost sharing, and rewards / penalties to achieve consistent contract execution.

Description

Technical Field

[0001] This invention relates to the fields of supply chain collaborative management and food quality control technology, specifically to a method for identifying quality stages and calculating collaborative contracts for fermented foods during their sales cycle, which enables consistent execution of revenue distribution, cost sharing, and reward / penalty linkage between producers and retailers under observation-limited conditions. Background Technology

[0002] During the production, fermentation, distribution, and sales of fermented foods, the quality exhibits phased changes over time, commonly characterized by transitions between enhancement, stabilization, and decline phases. Existing collaborative contracts often rely on fixed quality assumptions or static allocation rules, making it difficult to incorporate quality phase changes into the dynamic updates of contract parameters. This leads to inconsistencies in decision-making and allocation at phase boundaries. Furthermore, the diverse sources of quality observations and varying sampling frequencies result in gaps, noise, and fluctuations in reliability, easily causing quality judgment biases and further amplifying information asymmetry. Current practices typically treat information sharing as an independent management action, lacking a triggering mechanism linked to phased uncertainties. This leads to insufficient or excessive sharing, making it difficult to maintain consistency in collaborative inputs, pricing decisions, and settlement rules, ultimately impacting contract enforceability and quality control effectiveness. Summary of the Invention

[0003] This invention proposes a collaborative contract method for food quality control based on quality stage identification. By aligning time, repairing missing data, and unifying scales of multi-source observations of fermentation and distribution, a comprehensive quality observation measurement and credibility description are constructed. Under stage continuity constraints, stage beliefs are updated to output stage labels and confidence levels. Information sharing trigger quantities are generated based on confidence levels and observation changes, and the accuracy of quality information is adaptively adjusted. Combined with finite-dimensional quality state estimation, a sequential linkage decision is formed between producer input and retailer pricing. Based on quality state estimation, stage labels, and information accuracy, dynamic benefit sharing, cost sharing, and quality compliance reward / penalty parameters are calculated, achieving self-adaptive updates and consistent execution of the collaborative contract as quality stages switch.

[0004] The present invention provides a collaborative contract method for food quality control based on quality stage identification, comprising the following steps:

[0005] S1: Acquire multi-source observation data and bind traceability identifiers, and set a unified discrete-time grid. ,in Indicates the alignment start point, Indicates the alignment interval, where i represents the time index and satisfies 'I' represents the end index of the window, the raw record for the m-th observation channel. Mapped to Alignment observations formed above

[0006]

[0007] in This is the original timestamp. These are the original observations. For the number of records, Indicates the time kernel width. This represents an exponential function, where m is the channel index and M represents the number of channels;

[0008] Convert the aligned sequence into a normalized sequence with a uniform scale. The standardized channel value is expressed as

[0009]

[0010] The repaired channel value is denoted as , Indicates the robust center parameter of the channel. Indicates the robustness scale parameter of the channel. This represents an operator that truncates the input u to the interval [-1, 1].

[0011] S2: Map the standardized sequence to a single-channel quality contribution according to the target interval, and generate a confidence level based on the channel stability. Then, perform weighted fusion on each channel based on the confidence level to output the comprehensive quality observation and comprehensive confidence level.

[0012] S3: Update stage beliefs based on the support of stage continuity priors and comprehensive quality observations, output the current quality stage label and stage confidence, and suppress frequent switching at the boundary.

[0013] S4: Calculate the triggering driving quantity based on the stage confidence level, the observed change range and the overall credibility, and generate the information sharing triggering quantity. Adaptively adjust the quality information accuracy according to the triggering quantity so that the information accuracy and uncertainty are linked in a consistent manner.

[0014] S5: Under the constraints of stage labels and information accuracy, a quality state estimate is generated recursively. Based on the estimate, the producer's input decision and the retailer's pricing decision are generated to ensure the consistency of the sequential response.

[0015] S6: Calculate the dynamic benefit-sharing ratio based on the quality status estimate, stage label, and information accuracy, simultaneously generate cost-sharing parameters and quality achievement reward and punishment parameters, and output executable collaborative contract results.

[0016] According to a specific implementation of an embodiment of the present invention, the specific steps of S2 are as follows:

[0017] S2. Construct a single-channel quality contribution for each observation channel, denoted by .

[0018]

[0019] in, Indicates availability tag, This means it is available. This represents an invalid position. and The lower and upper boundaries of the target interval are defined as follows: For smoothing scale, It is an exponential function;

[0020] Generate channel credibility and complete fusion accordingly; construct channel credibility based on fluctuations and changes.

[0021]

[0022] in, Indicates that channel m is in Local fluctuation measurement at the location Indicates the measure of adjacent changes. The fluctuation penalty coefficient, This is the jump penalty coefficient;

[0023] Normalized fusion weights are obtained based on credibility.

[0024]

[0025] in, It is a numerically stable term;

[0026] Calculate the overall quality observation With overall credibility Output

[0027]

[0028]

[0029] The overall credibility score can be obtained by aggregating the credibility scores of each channel under weights and then outputting it. .

[0030] According to a specific implementation of an embodiment of the present invention, the specific steps of S3 are as follows:

[0031] S3. Output the confidence level of the output stage for subsequent triggering decisions, and define the quality stage set. }, where 1 represents the boost period, 2 represents the steady period, and 3 represents the decay period. Let} Indicates time Comprehensive quality observation, Indicates overall credibility;

[0032] Defining the phase belief vector ,in The belief weights at stage h and satisfying ;

[0033] Perform prior updates to reflect phase continuity and obtain prior beliefs.

[0034]

[0035] in For a priori beliefs, For the belief of the previous moment, This is the stage transition matrix, and its diagonal dominance varies with confidence level. Enhancement and strengthening;

[0036] Correct priors using current observations, and set up observation centers for each stage. With extended parameters Construct observational support

[0037]

[0038] in, The equivalent noise scale for the credibility mapping. To observe the support for stage h, It is an exponential function;

[0039] Update posterior beliefs based on support and prior knowledge.

[0040]

[0041] in For prior components, It is a numerically stable term.

[0042] According to a specific implementation of an embodiment of the present invention, the specific steps of S4 are as follows:

[0043] S4. Construct the driving quantity for triggering the decision, let... Indicates time The stage confidence level, with a value ranging from [0,1], and a larger value indicates a more certain stage judgment, is given by... Indicates comprehensive quality observations. Indicates overall confidence level and defines the uncertain driving force.

[0044]

[0045] in Indicates the triggering drive quantity. This represents the amplification factor. Represents the numerically stable term. Represents absolute value;

[0046] Based on the driving quantity, generate information sharing trigger quantities and define trigger quantities. And use threshold judgment

[0047]

[0048] in This indicates an indicator function that returns 1 if the condition is true and 0 otherwise. Indicates the trigger threshold parameter;

[0049] The precision of the quality information is adaptively adjusted based on the trigger value, and a precision description for subsequent steps is output. The precision of the quality information is defined as follows:

[0050]

[0051] in, This indicates the base accuracy level when sharing is not triggered. This indicates the target accuracy level after sharing is triggered.

[0052] According to a specific implementation of an embodiment of the present invention, the specific steps of S5 are as follows:

[0053] S5. Construct a finite-dimensional quality state estimation process, assuming... To ensure information accuracy, the quality state estimate is defined as follows: Define the estimation recursion

[0054]

[0055] in, Indicates at time Quality-related state estimates, Indicates the estimated gain. This represents a stage-related prediction mapping. Represents the observation mapping, the With information accuracy Monotonous adjustment;

[0056] Based on the quality estimate, the producer's input decision is generated, and the producer's input intensity is defined as... ,definition

[0057]

[0058] in Indicates that the manufacturer is at any time The intensity of investment, As the benchmark input item, This is a vector of input sensitivity coefficients related to the stage. and The upper and lower limits of investment, It is a saturation operator;

[0059] Retailer pricing decisions are generated under sequential response relationships, and the retailer's price decision is defined as follows:

[0060]

[0061] in Indicates that retailers are at the moment The pricing results For the benchmark price item, This is a vector of price sensitivity coefficients related to different stages. The input response coefficient, and These are the upper and lower limits of the price.

[0062] According to a specific implementation of an embodiment of the present invention, the specific steps of S6 are as follows:

[0063] S6. Generate a dynamic revenue-sharing ratio, set... This represents the quality state estimate output by S5. This indicates the stage label output by S3. This indicates the accuracy of the information output by S4, and defines the revenue-sharing ratio as... And constrained within [0,1] to ensure executability, construct the sharing ratio of segmented modulation.

[0064]

[0065] in Indicates the benchmark shared item, This represents a vector of shared sensitivity coefficients related to the stage. This represents the modulation coefficients for information precision related to the stage. Represents the natural logarithm. This represents an operator that truncates a value to [0,1].

[0066] Generate cost-sharing parameters to align with the stage and link them to information accuracy, defining the cost-sharing coefficient as follows: And constrained within [0,1], construct

[0067]

[0068] in Indicates the baseline contribution item. This represents the sharing and linkage coefficient related to the stage. This indicates the modulation intensity of the information precision in relation to the load. Represents the numerically stable term;

[0069] Generate quality compliance reward and penalty parameters and output contract settlement results; define the quality compliance evaluation function. Used to extract verifiable quality scores from estimated quality status, and to set evaluation thresholds. Corresponding to the three categories of substandard, satisfactory, and excellent, the reward and penalty amounts are defined as follows:

[0070]

[0071] in Represents the hyperbolic tangent function. This is the coefficient for the intensity of rewards and punishments. Attached Figure Description

[0072] Figure 1 This is a flowchart of the method;

[0073] Figure 2 This is the architecture diagram of this method. Detailed Implementation

[0074] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to examples and accompanying drawings.

[0075] As attached Figure 1 and attached Figure 2 As shown, a collaborative contract method for food quality control based on quality stage identification includes the following steps:

[0076] Step 1: Acquire multi-source observation data and bind traceability identifiers, and set a unified discrete-time grid. ,in Indicates the alignment start point, Indicates the alignment interval, where i represents the time index and satisfies 'I' represents the end index of the window, the raw record for the m-th observation channel. Mapped to Alignment observations formed above

[0077]

[0078] in This is the original timestamp. These are the original observations. For the number of records, Indicates the time kernel width. This represents an exponential function, where m is the channel index and M represents the number of channels;

[0079] Convert the aligned sequence into a normalized sequence with a uniform scale. The standardized channel value is expressed as

[0080]

[0081] The repaired channel value is denoted as , Indicates the robust center parameter of the channel. Indicates the robustness scale parameter of the channel. This represents an operator that truncates the input u to the interval [-1, 1].

[0082] S2: Map the standardized sequence to a single-channel quality contribution according to the target interval, and generate a confidence level based on the channel stability. Then, perform weighted fusion on each channel based on the confidence level to output the comprehensive quality observation and comprehensive confidence level.

[0083] S3: Update stage beliefs based on the support of stage continuity priors and comprehensive quality observations, output the current quality stage label and stage confidence, and suppress frequent switching at the boundary.

[0084] S4: Calculate the triggering driving quantity based on the stage confidence level, the observed change range and the overall credibility, and generate the information sharing triggering quantity. Adaptively adjust the quality information accuracy according to the triggering quantity so that the information accuracy and uncertainty are linked in a consistent manner.

[0085] S5: Under the constraints of stage labels and information accuracy, a quality state estimate is generated recursively. Based on the estimate, the producer's input decision and the retailer's pricing decision are generated to ensure the consistency of the sequential response.

[0086] S6: Calculate the dynamic benefit-sharing ratio based on the quality status estimate, stage label and information accuracy, simultaneously generate cost-sharing parameters and quality compliance reward and punishment parameters, and output executable collaborative contract results;

[0087] Step 2: Construct a single-channel quality contribution for each observation channel, expressed as .

[0088]

[0089] in, Indicates availability tag, This means it is available. This represents an invalid position. and The lower and upper boundaries of the target interval are defined as follows: For smoothing scale, It is an exponential function;

[0090] Generate channel credibility and complete fusion accordingly; construct channel credibility based on fluctuations and changes.

[0091]

[0092] in, Indicates that channel m is in Local fluctuation measurement at the location Indicates the measure of adjacent changes. The fluctuation penalty coefficient, This is the jump penalty coefficient;

[0093] Normalized fusion weights are obtained based on credibility.

[0094]

[0095] in, It is a numerically stable term;

[0096] Calculate the overall quality observation With overall credibility Output

[0097]

[0098]

[0099] The overall credibility score can be obtained by aggregating the credibility scores of each channel under weights and then outputting it. ;

[0100] Step 3: Output the stage confidence score for subsequent triggering decisions, and define the quality stage set. }, where 1 represents the boost period, 2 represents the steady period, and 3 represents the decay period. Let} Indicates time Comprehensive quality observation, Indicates overall credibility;

[0101] Defining the phase belief vector ,in The belief weights at stage h and satisfying ;

[0102] Perform prior updates to reflect phase continuity and obtain prior beliefs.

[0103]

[0104] in For a priori beliefs, For the belief of the previous moment, This is the stage transition matrix, and its diagonal dominance varies with confidence level. Enhancement and strengthening;

[0105] Correct priors using current observations, and set up observation centers for each stage. With extended parameters Construct observational support

[0106]

[0107] in, The equivalent noise scale for the credibility mapping. To observe the support for stage h, It is an exponential function;

[0108] Update posterior beliefs based on support and prior knowledge.

[0109]

[0110] in For prior components, It is a numerically stable term;

[0111] Step 4: Construct the driving quantity for triggering the decision, let... Indicates time The stage confidence level, with a value ranging from [0,1], and a larger value indicates a more certain stage judgment, is given by... Indicates comprehensive quality observations. Indicates overall confidence level and defines the uncertain driving force.

[0112]

[0113] in Indicates the triggering drive quantity. This represents the amplification factor. Represents the numerically stable term. Represents absolute value;

[0114] Based on the driving quantity, generate information sharing trigger quantities and define trigger quantities. And use threshold judgment

[0115]

[0116] in This indicates an indicator function that returns 1 if the condition is true and 0 otherwise. Indicates the trigger threshold parameter;

[0117] The precision of the quality information is adaptively adjusted based on the trigger value, and a precision description for subsequent steps is output. The precision of the quality information is defined as follows:

[0118]

[0119] in, This indicates the base accuracy level when sharing is not triggered. Indicates the target accuracy level after sharing is triggered;

[0120] Step 5: Construct a finite-dimensional mass state estimation process, assuming... To ensure information accuracy, the quality state estimate is defined as follows: Define the estimation recursion

[0121]

[0122] in, Indicates at time Quality-related state estimates, Indicates the estimated gain. This represents a stage-related prediction mapping. Represents the observation mapping, the With information accuracy Monotonous adjustment;

[0123] Based on the quality estimate, the producer's input decision is generated, and the producer's input intensity is defined as... ,definition

[0124]

[0125] in Indicates that the manufacturer is at any time The intensity of investment, As the benchmark input item, This is a vector of input sensitivity coefficients related to the stage. and The upper and lower limits of investment, It is a saturation operator;

[0126] Retailer pricing decisions are generated under sequential response relationships, and the retailer's price decision is defined as follows:

[0127]

[0128] in Indicates that retailers are at the moment The pricing results For the benchmark price item, This is a vector of price sensitivity coefficients related to different stages. The input response coefficient, and These are the upper and lower limits of the price.

[0129] Step 6: Generate a dynamic revenue-sharing ratio, set... This represents the quality state estimate output by S5. This indicates the stage label output by S3. This indicates the accuracy of the information output by S4, and defines the revenue-sharing ratio as... And constrained within [0,1] to ensure executability, construct the sharing ratio of segmented modulation.

[0130]

[0131] in Indicates the benchmark shared item, This represents a vector of shared sensitivity coefficients related to the stage. This represents the modulation coefficients for information precision related to the stage. Represents the natural logarithm. This represents an operator that truncates a value to [0,1].

[0132] Generate cost-sharing parameters to align with the stage and link them to information accuracy, defining the cost-sharing coefficient as follows: And constrained within [0,1], construct

[0133]

[0134] in Indicates the baseline contribution item. This represents the sharing and linkage coefficient related to the stage. This indicates the modulation intensity of the information precision in relation to the load. Represents the numerically stable term;

[0135] Generate quality compliance reward and penalty parameters and output contract settlement results; define the quality compliance evaluation function. Used to extract verifiable quality scores from estimated quality status, and to set evaluation thresholds. Corresponding to the three categories of substandard, satisfactory, and excellent, the reward and penalty amounts are defined as follows:

[0136]

[0137] in Represents the hyperbolic tangent function. This is the coefficient for the intensity of rewards and punishments.

Claims

1. A collaborative contract method for food quality control based on quality stage identification, characterized in that... Includes the following steps: S1: Acquire multi-source observation data and bind traceability identifiers, and set a unified discrete-time grid. ,in Indicates the alignment start point, Indicates the alignment interval, where i represents the time index and satisfies 'I' represents the end index of the window, the raw record for the m-th observation channel. Mapped to Alignment observations formed above : ； in This is the original timestamp. These are the original observations. For the number of records, Indicates the time kernel width. This represents an exponential function, where m is the channel index and M represents the number of channels; Convert the aligned sequence into a normalized sequence with a uniform scale. The standardized channel value is expressed as ； The repaired channel value is denoted as , Indicates the robust center parameter of the channel. Indicates the channel robustness scale parameter. This represents an operator that truncates the input u to the interval [-1, 1]. S2: Map the standardized sequence to a single-channel quality contribution according to the target interval, and generate a confidence level based on the channel stability. Then, perform weighted fusion on each channel based on the confidence level to output the comprehensive quality observation and comprehensive confidence level. S3: Update stage beliefs based on the support of stage continuity priors and comprehensive quality observations, output the current quality stage label and stage confidence, and suppress frequent switching at the boundary. S4: Calculate the triggering driving quantity based on the stage confidence level, the observed change range and the overall credibility, and generate the information sharing triggering quantity. Adaptively adjust the quality information accuracy according to the triggering quantity so that the information accuracy and uncertainty are linked in a consistent manner. S5: Under the constraints of stage labels and information accuracy, a quality state estimate is generated recursively. Based on the estimate, the producer's input decision and the retailer's pricing decision are generated to ensure the consistency of the sequential response. S6: Calculate the dynamic benefit-sharing ratio based on the quality status estimate, stage label, and information accuracy, simultaneously generate cost-sharing parameters and quality achievement reward and punishment parameters, and output executable collaborative contract results.

2. The collaborative contract method for food quality control based on quality stage identification according to claim 1, characterized in that... The specific method for step S2 is as follows: S2. Construct a single-channel quality contribution for each observation channel, denoted by . : ； in, Indicates availability tag, This means it is available. This represents an invalid position. and The lower and upper boundaries of the target interval are defined as follows: For smoothing scale, It is an exponential function; Generate channel credibility and complete fusion accordingly; construct channel credibility based on fluctuations and changes. : ； in, Indicates that channel m is in Local fluctuation measurement at the location Indicates the measure of adjacent changes. The fluctuation penalty coefficient, This is the jump penalty coefficient; Normalized fusion weights are obtained based on credibility. : ； in, It is a numerically stable term; Calculate the overall quality observation With overall credibility Output ； ； The overall credibility score can be obtained by aggregating the credibility scores of each channel under weights and then outputting it. .

3. The collaborative contract method for food quality control based on quality stage identification according to claim 1, characterized in that... The specific method in step S3 is as follows: S3. Output the confidence level of the output stage for subsequent triggering decisions, and define the quality stage set. }, where 1 represents the boost period, 2 represents the steady period, and 3 represents the decay period. Let} Indicates time Comprehensive quality observation, Indicates overall credibility; Defining the phase belief vector ,in The belief weights at stage h and satisfying ; Perform prior updates to reflect phase continuity and obtain prior beliefs. ； in For a priori beliefs, For the belief of the previous moment, This is the stage transition matrix, and its diagonal dominance varies with confidence level. Enhancement and strengthening; Correct priors using current observations, and set up observation centers for each stage. With extended parameters Construct observational support ； in, The equivalent noise scale for the credibility mapping. To observe the support for stage h, It is an exponential function; Update posterior beliefs based on support and prior knowledge. ； in For prior components, It is a numerically stable term.

4. The collaborative contract method for food quality control based on quality stage identification according to claim 1, characterized in that... The specific method in step S4 is as follows: S4. Construct the driving quantity for triggering the decision, let... Indicates time The stage confidence level, with a value range of [0,1], and a larger value indicates a more certain stage judgment, defines the uncertainty driving quantity. ； in Indicates the triggering drive quantity. This represents the amplification factor. Represents the numerically stable term. Represents absolute value; Based on the driving quantity, generate information sharing trigger quantities and define trigger quantities. And use threshold judgment ； in This indicates an indicator function that returns 1 if the condition is true and 0 otherwise. Indicates the trigger threshold parameter; The precision of the quality information is adaptively adjusted based on the trigger value, and a precision description for subsequent steps is output. The precision of the quality information is defined as follows: : ； in, This indicates the base accuracy level when sharing is not triggered. This indicates the target accuracy level after sharing is triggered.

5. The collaborative contract method for food quality control based on quality stage identification according to claim 1, characterized in that... The specific method in step S5 is as follows: S5. Construct a finite-dimensional quality state estimation process, assuming... To ensure information accuracy, the quality state estimate is defined as follows: Define the estimation recursion ； in, Indicates at time Quality-related state estimates, Indicates the estimated gain. This represents a stage-related prediction mapping. Represents the observation mapping, the With information accuracy Monotonous adjustment; Based on the quality estimate, the producer's input decision is generated, and the producer's input intensity is defined as... ,definition ； in Indicates that the manufacturer is at any time The intensity of investment, As the benchmark input item, This is a vector of input sensitivity coefficients related to the stage. and The upper and lower limits of investment, It is a saturation operator; Retailer pricing decisions are generated under sequential response relationships, and the retailer's price decision is defined as follows: : ； in Indicates that retailers are at the moment The pricing results For the benchmark price item, This is a vector of price sensitivity coefficients related to different stages. The input response coefficient, and These are the upper and lower limits of the price.

6. The collaborative contract method for food quality control based on quality stage identification according to claim 1, characterized in that... The specific method in step S6 is as follows: S6. Generate a dynamic revenue-sharing ratio, set... This represents the quality state estimate output by S5. This indicates the stage label output by S3. This indicates the accuracy of the information output by S4, and defines the revenue-sharing ratio as... And constrained within [0,1] to ensure executability, construct the sharing ratio of segmented modulation. ； in Indicates the benchmark shared item, This represents a vector of shared sensitivity coefficients related to the stage. This represents the modulation coefficients for information precision related to the stage. Represents the natural logarithm. This represents an operator that truncates a value to [0,1]. Generate cost-sharing parameters to align with the stage and link them to information accuracy, defining the cost-sharing coefficient as follows: And constrained within [0,1], construct ； in Indicates the baseline contribution item. This represents the sharing and linkage coefficient related to the stage. This indicates the modulation intensity of the information precision in relation to the load. Represents the numerically stable term; Generate quality compliance reward and penalty parameters and output contract settlement results; define the quality compliance evaluation function. Used to extract verifiable quality scores from estimated quality status, and to set evaluation thresholds. Corresponding to the three categories of substandard, satisfactory, and excellent, the reward and penalty amounts are defined as follows: : ； in Represents the hyperbolic tangent function. This is the reward / punishment intensity coefficient.

Images