Dairy traceability management system based on cloud platform

Abstract

The invention relates to the technical field of cold-chain logistics internet of things and food safety data intelligent analysis, in particular to a dairy traceability management system based on a cloud platform, which comprises a multi-source heterogeneous data aggregation unit used for generating an environment state vector sequence; the mass manifold space mapping unit is used for fitting the discrete state vectors into an actual transportation track curve by using a manifold projection algorithm; the trajectory curvature deviation analysis unit is used for extracting Riemannian curvature tensor features and generating damage scalar flow; the dynamic entropy increase index evaluation unit is used for generating a dynamic entropy increase index, a graded quality early warning signal and a dynamic shelf life suggestion value; the self-adaptive risk blocking and feedback unit is used for feeding back the abnormal actual transportation track curve data as a negative sample to the mass manifold space mapping unit; according to the invention, the defect that the traditional discrete threshold alarm cannot identify the process accumulated damage is overcome, and the accurate control of the dairy product quality is realized.

Description

Technical Field

[0001] This invention relates to the field of cold chain logistics Internet of Things and intelligent data analysis technology for food safety, specifically a dairy product traceability management system based on a cloud platform. Background Technology

[0002] With the rapid development of the cold chain logistics industry, the timeliness and accuracy requirements for environmental monitoring in dairy product transportation are increasing. This high standard has brought many challenges to the traditional logistics monitoring system in terms of data processing capabilities and risk identification mechanisms. Currently, existing technologies generally employ discrete threshold alarms for quality monitoring, relying on sensors at logistics terminals to collect single-dimensional data such as temperature and humidity. An alarm is triggered when the monitored values ​​exceed a preset fixed upper limit, or deviations are assessed by calculating the simple Euclidean distance between the actual trajectory and the ideal path. However, traditional discrete monitoring methods struggle to capture continuous trends in environmental parameters and cannot identify latent damage caused by drastic fluctuations in environmental stress during transportation that eventually return to normal. This makes it difficult to quantify the continuous cumulative damage suffered by dairy products. Furthermore, to preserve second-level fluctuation characteristics, existing solutions often require storing massive amounts of raw log data, facing a deadlock between data granularity and storage costs. They also lack deep integration with enterprise business systems, resulting in slow response times and an inability to automatically halt inventory disruptions when risks occur. Therefore, how to accurately identify latent cumulative damage during transportation while reducing data storage costs has become a pressing issue in this field. Summary of the Invention

[0003] The purpose of this invention is to provide a cloud-based dairy product traceability management system that avoids the problems of traditional discrete threshold alarms failing to identify continuous cumulative damage and the high cost of storing massive amounts of original logs. Furthermore, it can accurately identify hidden damage—where temperature recovery has been achieved but quality loss has occurred—using the Riemann curvature tensor, enabling automated blocking of inventory status and dynamic adjustment of shelf life. Specifically, the technical solution of this invention is as follows: The multi-source heterogeneous data aggregation unit is used to receive raw sensor data streams and geographic location time stamps from logistics and transportation terminals. It uses the built-in spatiotemporal alignment module to perform data cleaning and time axis synchronization operations, converting analog signals into digital sequences and generating environmental state vector sequences. The quality manifold space mapping unit is used to receive the environmental state vector sequence, perform data dimension normalization operation to construct state vectors, call the preset lossless cold chain manifold model, and use manifold projection algorithms such as isometric mapping ISOMAP or local linear embedding LLE to fit the discrete state vectors into the actual transportation trajectory curve. The trajectory curvature deviation analysis unit is used to receive the actual transportation trajectory curve and the preset ideal geodesic, calculate the covariant derivative difference between the actual transportation trajectory curve and the ideal geodesic, extract the Riemann curvature tensor features, and generate the damage scalar flow through tensor product operation and norm calculation. The dynamic entropy increase index evaluation unit is used to receive the damage scalar flow, perform time integration to generate a dynamic entropy increase index, compare the second time derivative of the entropy increase index with a preset mass phase transition critical threshold, and generate graded quality warning signals and dynamic shelf life recommendations. The adaptive risk blocking and feedback unit is used to activate the digital electronic fence function according to the quality early warning signal, modify the inventory status field in the enterprise ERP system, and feed back the abnormal actual transportation trajectory curve data as a negative sample to the quality manifold space mapping unit.

[0004] Optionally, the raw sensor data stream received by the multi-source heterogeneous data aggregation unit includes temperature data, humidity data, triaxial vibration data, and light intensity data; the environmental state vector sequence includes the instantaneous values ​​of environmental parameters and the rate of change of the environmental parameters over time.

[0005] Optionally, the process of constructing the state vector by the mass manifold space mapping unit is as follows: mapping the environmental state vector sequence to an n-dimensional Euclidean space; the lossless cold chain manifold model is a high-dimensional surface trained based on historical lossless data, and the high-dimensional surface defines the mass conservation constraint domain of dairy products under different combinations of environmental parameters; the ideal geodesic is the optimal parameter evolution path with the least fluctuation from the transportation start point to the transportation end point under the premise of satisfying the mass conservation constraint domain.

[0006] Optionally, the process of generating the damage scalar flow using the trajectory curvature deviation analytical unit is as follows: Calculate the tangent vectors of the actual transportation trajectory curve and the ideal geodesic at the same moment, where the tangent vectors represent the rate and direction of change of the environmental parameters; The second derivatives of the actual transportation trajectory curve in each dimension are calculated to extract the curvature tensor, which represents the degree of severe environmental fluctuations. The extracted curvature tensor features are subjected to tensor product operation to obtain the damage scalar flow that records the product's historical path-dependent damage.

[0007] Optionally, the entropy increase index in the dynamic entropy increase index evaluation unit is a dimensionless value. The initial value of the entropy increase index is zero and it increases monotonically with the transportation process. The critical threshold for mass phase change is a critical point determined by regression analysis of historical deterioration accident data and corresponding curvature characteristics. The critical point corresponds to the situation where the second derivative of the entropy increase index exceeds a preset value.

[0008] Optionally, the process by which the adaptive risk blocking and feedback unit executes positive control logic is as follows: When the quality warning signal indicating severe deterioration or slight fatigue is received, an instruction is sent to the enterprise ERP system via the API interface. Change the inventory status field of the corresponding batch of products from "in transit and available for sale" to "frozen pending inspection" or "scrapped"; The system automatically updates the expiration date data based on the suggested dynamic shelf life value and triggers a price reduction promotion suggestion.

[0009] Optionally, the adaptive risk blocking and feedback unit executes the reverse optimization logic as follows: The actual transportation trajectory curve data that was determined to be abnormal in this instance was marked as a negative sample. The negative sample is fed back to the mass manifold space mapping unit; The negative samples are used to fine-tune and correct the boundary of the non-destructive cold chain manifold model.

[0010] Compared with the prior art, the present invention has the following beneficial effects: 1. This system extracts the Riemann curvature tensor features by calculating the difference between the covariant derivative of the actual trajectory and the ideal geodesic, and transforms instantaneous environmental shocks into cumulative path-dependent damage. This method can not only quantify the degree of severe environmental fluctuations, but also identify hidden damages where the temperature has recovered but the quality has been compromised. It solves the problem that traditional discrete threshold alarms cannot identify process-accumulated damages, and achieves precise control over milk quality. 2. This system uses mass manifold space mapping units to transform linear time series into geometric spatial features, and fits discrete state vectors into continuous trajectories through manifold projection algorithms. This technology overcomes the contradiction between data granularity and storage cost, and uses geometric features to replace massive amounts of raw logs. While significantly reducing storage costs, it fully preserves second-level fluctuation features, ensuring the data isomorphism and integrity required for high-dimensional analysis. 3. This system is deeply coupled with the enterprise resource planning system through an adaptive risk blocking mechanism. Once a quality warning is generated, the digital electronic fence is immediately activated and the inventory status field is modified through the interface to automatically block the circulation of unqualified products. This millisecond-level linkage mechanism eliminates the lag of manual decision-making and ensures that the quality inspection results are immediately converted into offline inventory locking actions, which significantly improves the response speed and risk control capabilities of food safety management. 4. This system constructs an evaluation model based on the dynamic entropy increase index. By monitoring the abrupt change of the second derivative of the entropy increase index, it locks the critical point of qualitative change, thereby generating accurate dynamic shelf-life recommendations. At the same time, the system has self-evolution capabilities, which can feed back abnormal trajectories as negative samples to the mapping unit to fine-tune the model boundary. As data accumulates, it continuously improves the fit to real biochemical characteristics, reducing the false alarm rate and the false negative rate. Attached Figure Description

[0011] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0013] Example 1: Please see Figure 1 A cloud-based dairy product traceability management system includes: The multi-source heterogeneous data aggregation unit is used to receive raw sensor data streams and geographic location time stamps from logistics and transportation terminals. It uses the built-in spatiotemporal alignment module to perform data cleaning and time axis synchronization operations, converting analog signals into digital sequences and generating environmental state vector sequences. The quality manifold space mapping unit is used to receive the environmental state vector sequence, perform data dimension normalization operation to construct state vectors, call the preset lossless cold chain manifold model, and use manifold projection algorithms such as isometric mapping ISOMAP or local linear embedding LLE to fit the discrete state vectors into the actual transportation trajectory curve. The trajectory curvature deviation analysis unit is used to receive the actual transportation trajectory curve and the preset ideal geodesic, calculate the covariant derivative difference between the actual transportation trajectory curve and the ideal geodesic, extract the Riemann curvature tensor features, and generate the damage scalar flow through tensor product operation and norm calculation. The dynamic entropy increase index evaluation unit is used to receive the damage scalar flow, perform time integration to generate a dynamic entropy increase index, compare the second time derivative of the entropy increase index, i.e. the abrupt acceleration that characterizes the trend of accelerated deterioration, with a preset critical threshold for quality phase transition, and generate graded quality warning signals and dynamic shelf life recommendations. The adaptive risk blocking and feedback unit is used to activate the digital electronic fence function according to the quality early warning signal, modify the inventory status field in the enterprise ERP system, and feed back the abnormal actual transportation trajectory curve data as a negative sample to the quality manifold space mapping unit.

[0014] To address the fundamental deficiency in existing technologies where discrete threshold alarms cannot identify continuous cumulative damage, this system reconstructs the quality monitoring logic using high-dimensional manifold mapping technology. After receiving the raw sensor data stream and geographic location time stamp from the cold chain transportation terminal, the multi-source heterogeneous data aggregation unit initiates a data cleaning and synchronization mechanism through its built-in spatiotemporal alignment module. Given the inherent differences in sampling frequencies of different physical sensors, this module uses interpolation algorithms to fill in the time gaps in low-frequency data and filter out high-frequency electromagnetic noise, converting the analog signal into a digital sequence with a unified time reference. This process generates a standardized environmental state vector sequence, ensuring the data isomorphism of subsequent high-dimensional analysis. The mass manifold space mapping unit transforms linear time series into geometric space features. This unit performs data dimension normalization, mapping heterogeneous physical quantities to the same numerical range to construct state vectors. Relying on a pre-defined lossless cold chain manifold model, which is a high-dimensional surface representing the dairy product quality conservation constraint domain trained based on massive historical lossless transportation data, this unit uses a manifold projection algorithm to map and fit discrete state vector points into a continuously extending actual transportation trajectory curve. This step realizes a paradigm shift from data point monitoring to geometric path analysis. The trajectory curvature deviation analysis unit quantifies the environmental stress experienced by dairy products during transportation based on the principles of differential geometry. This unit does not calculate the simple Euclidean distance between the actual transportation trajectory curve and the preset ideal geodesic, but instead analyzes the difference in the covariant derivatives of the two in the same spatiotemporal coordinates. This operation aims to extract the Riemann curvature tensor features, which can sensitively characterize the high curvature traces left by drastic fluctuations in environmental parameters in geometric space. This unit performs tensor product operations on the extracted curvature features to generate a damage scalar flow, thereby transforming the instantaneous environmental impact into an accumulative historical path-dependent damage record. The dynamic entropy increase index evaluation unit is responsible for decoding abstract tensor data into executable business indicators. This unit receives the damage scalar stream and performs time integration to generate a dynamic entropy increase index. The system monitors the evolution rate of the entropy increase index in real time and compares the second time derivative of the entropy increase index with a preset critical threshold for quality phase transition. Once the calculated value of the second time derivative exceeds the threshold, the system determines that an irreversible quality change has occurred inside the dairy product, and then generates a graded quality warning signal and a precise dynamic shelf life recommendation value. In response to a quality warning signal, the adaptive risk blocking and feedback unit immediately activates the digital electronic fence function, uses the API interface to penetrate to the underlying layer of the enterprise ERP system, and forcibly modifies the inventory status field of the batch of products involved. At the same time, in order to enhance the system's evolutionary capability, the unit marks the actual transportation trajectory curve data that is determined to be abnormal as a negative sample and feeds it back to the quality manifold space mapping unit to correct the boundary parameters of the model and achieve iterative optimization of detection accuracy. This invention breaks through the deadlock between data granularity and storage cost in traditional cold chain logistics. It uses manifold mapping technology to store geometric features instead of massive amounts of raw logs, significantly reducing storage costs while fully preserving second-level fluctuation characteristics. Compared with existing technologies, this solution uses the Riemann curvature tensor to accurately identify hidden damage where the temperature has recovered but the quality has been compromised. Through deep coupling with the ERP system, it achieves automated blocking of inventory status and dynamic adjustment of shelf life, significantly improving the response speed and accuracy of food safety management.

[0015] Example 2: The raw sensor data stream received by the multi-source heterogeneous data aggregation unit includes temperature data, humidity data, triaxial vibration data, and light intensity data; the environmental state vector sequence contains the instantaneous values ​​of environmental parameters and the rate of change of the environmental parameters over time.

[0016] The multi-source heterogeneous data aggregation unit is configured to perform full-dimensional environmental perception. In order to fully capture the causes of dairy product spoilage, the raw sensor data stream received by this unit covers four core dimensions: temperature and humidity data are used to assess the thermodynamic environment for microbial growth, triaxial vibration data is used to quantify the risk of fat globule rupture caused by physical shaking, and light intensity data is used to monitor the occurrence of photo-oxidation reaction. The environmental state vector sequence generated by this unit possesses high-order features; this sequence not only encapsulates the instantaneous values ​​of the aforementioned environmental parameters at a specific timestamp, but also embeds the rate of change features of the environmental parameters over time through differential calculation; for example, this vector sequence also includes the current temperature. With the rate of temperature change ; By introducing multidimensional data and rate of change features, the system can not only grasp the current state of the environment, but also capture the trend of environmental change. This vector structure containing the rate of change provides the necessary mathematical basis for constructing a directional vector field in the manifold space, which significantly enhances the system's ability to predict the trend of environmental deterioration.

[0017] Example 3: The process of constructing state vectors using the mass manifold space mapping unit is as follows: the environmental state vector sequence is mapped to a high-dimensional phase space, and a topological structure based on manifold learning is embedded in this phase space; the lossless cold chain manifold model is a high-dimensional surface trained based on historical lossless data, and the high-dimensional surface defines the mass conservation constraint domain of dairy products under different combinations of environmental parameters; the ideal geodesic pipeline is defined as a tubular topological region within the mass conservation constraint domain, with the optimal parameter evolution path as the central axis, allowing for small environmental disturbances within a preset range; this region covers multiple equivalent lossless transportation paths.

[0018] When constructing state vectors, the mass manifold space mapping unit performs a high-dimensional embedding operation, mapping the sequence of environment state vectors containing multiple physical parameters to a single dimension. In Vioclimatic space, specifically, to eliminate the dimensional differences between heterogeneous data such as temperature, humidity, and vibration acceleration, a weighted Mahalanobis distance is systematically constructed as the Riemannian metric tensor of this manifold space. The weight parameters of the metric tensor are initialized by performing principal component analysis (PCA) on each environmental factor in the historical lossless data to maintain quality, thereby giving the environmental parameters specific distance and angle meanings in geometric space; making the environmental state at each time point a coordinate point in that space. In this spatial architecture, the system's pre-defined lossless cold chain manifold model is defined as a data-driven high-dimensional surface. The model is constructed based on topological learning of massive historical lossless transportation data, and its geometric boundaries strictly define the quality conservation constraint domain that enables dairy products to maintain stable quality under the coupling effects of different environmental parameters. Any point that deviates from the surface constraint domain is considered to have deviated from the quality conservation state. An ideal geodesic is defined as an energy functional minimum path connecting the starting point and the ending point of transportation, under the premise of strictly satisfying the boundary conditions of the mass conservation constraint domain; this path represents the optimal parameter evolution path that theoretically minimizes product damage. By defining the mass conservation constraint domain and ideal geodesics, the complex biochemical degradation mechanism is transformed into a precise geometric topological problem. This transformation enables the system to use mature geometric algorithms to handle the nonlinear environmental coupling effects, providing a unique and objective mathematical benchmark for quantifying latent damage.

[0019] Example 4: The process of generating damage scalar flow using the trajectory curvature deviation analytical unit is as follows: Calculate the tangent vectors of the actual transportation trajectory curve and the ideal geodesic at the same moment, where the tangent vectors represent the rate and direction of change of the environmental parameters; Using the Levi-Civita link, the covariant second derivative of the actual transport trajectory curve is approximated by the finite difference method to extract geodesic curvature features, which characterize the degree to which the transport state is forced to deviate from the manifold inertial path by external forces. The extracted curvature tensor features are subjected to tensor product operation, tensor shrinkage or norm calculation is performed, and the higher-order tensor is transformed into a scalar sequence representing the instantaneous damage intensity, generating a damage scalar flow that records the product's historical path-dependent damage.

[0020] This unit calculates the tangent vectors of the actual transportation trajectory curve and the ideal geodesic at the same time node; in this technical context, the tangent vector, as a vector entity, accurately represents the rate and direction of change of the environmental parameters at that moment. This unit performs a second-order differential operation on the actual transportation trajectory curve to extract the curvature tensor. The core logic of this step is that the second derivative of the curve corresponds to its geometric curvature. In a physical sense, the curvature tensor directly maps the degree of severe environmental fluctuations. Even in a refreezing process where the value briefly returns to normal, a high curvature feature will be generated in this step because it has undergone drastic state changes. This unit performs tensor product operations on the extracted curvature tensor features of each dimension; this operation performs multilinear mapping and accumulation on the instantaneous stress features to generate a damage scalar flow; this data flow not only records the current stress state, but also contains the integral effect of the time dimension, thus fully recording the product's historical path-dependent damage. By introducing the second derivative to extract the curvature tensor, the system can capture process damage that cannot be identified by conventional threshold monitoring. This technical solution effectively solves the detection blind spot where the result is normal but the process is damaged, and ensures accurate identification of latent deterioration risks such as thermal shock.

[0021] Example 5: The entropy increase index in the dynamic entropy increase index evaluation unit is a dimensionless value. The initial value of the entropy increase index is zero and it increases monotonically with the transportation process. The critical threshold for mass phase change is a critical point determined by regression analysis of historical deterioration accident data and corresponding curvature characteristics. The critical point corresponds to the situation where the second derivative of the entropy increase index exceeds a preset value.

[0022] In biochemical dynamics, the sudden increase in the second derivative corresponds to the inflection point when microorganisms enter the logarithmic growth phase from the lag phase, indicating that the deterioration reaction begins to accelerate out of control. The entropy increase index is defined as a dimensionless value that characterizes the degree of disorder or potential spoilage of dairy products during their current life cycle. Following the principle of thermodynamic irreversibility, the initial value of the index is set to zero, and it shows a monotonically increasing trend with the accumulation of environmental disturbances during transportation. The critical threshold for phase transition, serving as the core criterion of the system, is not a fixed constant set by humans, but a dynamic critical point determined through big data regression analysis. Specifically, the system retrieves historical deterioration accident data and their corresponding curvature features, and determines the threshold through algorithm fitting. Mathematically, the critical point strictly corresponds to the situation where the second derivative of the entropy increase exponent exceeds a preset value. This, at the biochemical level, represents the entry of bacterial reproduction into an exponential growth phase, indicating the inevitable occurrence of qualitative change. The calculation process of the dynamic shelf life recommendation value specifically adopts a kinematic extrapolation model: the system obtains the current time... entropy increase index The first derivative of the entropy increase exponent and the second derivative The critical threshold for mass phase transition is set as follows: Considering that the biochemical reaction rate of dairy product spoilage typically follows a nonlinear Arrhenius equation or an exponential growth model, this system employs a Taylor series expansion method to approximate the spoilage trajectory at the current moment in order to simplify calculations and achieve real-time prediction. This formula essentially uses the current spoilage state as a base point, locally approximating the complex biochemical decay curve as a kinematic trajectory. The formula is based on... Solving for time variables The positive real solutions, among which, The set critical threshold for mass phase transition. For the current moment The entropy increase index, The first derivative of the entropy increase exponent. The second derivative of the entropy increase exponent. Let the time variable be the solution to be solved. This is the recommended value for the remaining dynamic shelf life under the current environmental stress state; By establishing a monotonically increasing model based on the entropy increase exponent and a threshold determination mechanism based on the mutation of the second derivative, this system is freed from the limitation of a single upper temperature limit. This scheme focuses on the acceleration of the deterioration process, thus enabling more scientific and earlier detection of the shelf life inflection point, providing a solid theoretical basis for proactively preventing risks.

[0023] Example 6: The process by which the adaptive risk blocking and feedback unit executes positive control logic is as follows: When the quality warning signal indicating severe deterioration or slight fatigue is received, an instruction is sent to the enterprise ERP system via the API interface. Change the inventory status field of the corresponding batch of products from "in transit and available for sale" to "frozen pending inspection" or "scrapped"; The system automatically updates the expiration date data based on the suggested dynamic shelf life value and triggers a price reduction promotion suggestion.

[0024] Once a quality warning signal indicating severe deterioration or slight fatigue is received, the unit immediately initiates a blocking procedure; relying on a pre-set communication protocol, the unit establishes an instruction channel with the enterprise ERP system through an API interface and sends status change instructions. This instruction acts directly on the underlying database, forcibly modifying the inventory status field of the corresponding batch of products; based on the warning level, the system will immediately change the original status from in transit to available for sale to frozen to pending inspection or scrapped, thereby blocking the flow of information before the physical logistics arrive. Based on the dynamic shelf-life recommendation value, this unit automatically updates the expiration date data in the system and triggers a price reduction and promotion suggestion when the remaining shelf life is lower than the preset warning line, notifying the sales department to take emergency measures. The millisecond-level inventory status linkage achieved through the API interface completely eliminates the lag in manual decision-making; this mechanism ensures that cloud-based quality inspection results can be instantly converted into offline inventory locking actions, which not only prevents unqualified products from entering the market, but also helps enterprises minimize economic losses through dynamic shelf-life management.

[0025] Example 7: The process of the adaptive risk blocking and feedback unit executing the reverse optimization logic is as follows: The actual transportation trajectory curve data that was determined to be abnormal in this instance was marked as a negative sample. The negative sample is fed back to the mass manifold space mapping unit; The negative samples are used to fine-tune and correct the boundary of the non-destructive cold chain manifold model.

[0026] After an abnormal transportation event ends, the unit extracts features from the actual transportation trajectory curve data that caused the quality damage and marks it as a negative sample; a negative sample specifically refers to those trajectories with specific combinations of environmental parameters that have been empirically proven to lead to product quality deterioration. The unit initiates a feedback loop to transmit the negative sample data stream back to the front-end mass manifold space mapping unit. The mapping unit uses this real negative sample data to perform fine-tuning corrections on the geometric boundaries of the lossless cold chain manifold model, such as shrinking the range of the mass conservation constraint domain or increasing the weight parameters for specific high curvature features. This reverse optimization logic endows the system with the ability to learn continuously; with the accumulation of negative samples and the iterative correction of the model boundary, the system’s fit to the real biochemical characteristics of dairy products is continuously improved, thereby effectively reducing the false alarm rate and the false negative rate, and realizing the adaptive evolution of traceability accuracy.

[0027] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A cloud-based dairy product traceability management system, characterized in that, include: The multi-source heterogeneous data aggregation unit is used to receive raw sensor data streams and geographic location time stamps from logistics and transportation terminals. It uses the built-in spatiotemporal alignment module to perform data cleaning and time axis synchronization operations, converting analog signals into digital sequences and generating environmental state vector sequences. The quality manifold space mapping unit is used to receive the environmental state vector sequence, perform data dimension normalization operation to construct state vectors, call the preset lossless cold chain manifold model, and use manifold projection algorithms such as isometric mapping ISOMAP or local linear embedding LLE to fit the discrete state vectors into the actual transportation trajectory curve. The trajectory curvature deviation analysis unit is used to receive the actual transportation trajectory curve and the preset ideal geodesic, calculate the covariant derivative difference between the actual transportation trajectory curve and the ideal geodesic, extract the Riemann curvature tensor features, and generate the damage scalar flow through tensor product operation and norm calculation. The dynamic entropy increase index evaluation unit is used to receive the damage scalar flow, perform time integration to generate a dynamic entropy increase index, compare the second time derivative of the entropy increase index with a preset mass phase transition critical threshold, and generate graded quality warning signals and dynamic shelf life recommendations. The adaptive risk blocking and feedback unit is used to activate the digital electronic fence function according to the quality early warning signal, modify the inventory status field in the enterprise ERP system, and feed back the abnormal actual transportation trajectory curve data as a negative sample to the quality manifold space mapping unit.

2. The dairy product traceability management system based on a cloud platform according to claim 1, characterized in that, The raw sensor data stream received by the multi-source heterogeneous data aggregation unit includes temperature data, humidity data, triaxial vibration data, and light intensity data; the environmental state vector sequence includes the instantaneous values ​​of environmental parameters and the rate of change of the environmental parameters over time.

3. The dairy product traceability management system based on a cloud platform according to claim 1, characterized in that, The process of constructing the state vector by the mass manifold space mapping unit is as follows: mapping the environmental state vector sequence to an n-dimensional Euclidean space; the lossless cold chain manifold model is a high-dimensional surface trained based on historical lossless data, and the high-dimensional surface defines the mass conservation constraint domain of dairy products under different combinations of environmental parameters; the ideal geodesic is the optimal parameter evolution path with the least fluctuation from the transportation start point to the transportation end point under the premise of satisfying the mass conservation constraint domain.

4. The dairy product traceability management system based on a cloud platform according to claim 1, characterized in that, The process by which the trajectory curvature deviation analysis unit generates the damage scalar flow is as follows: Calculate the tangent vectors of the actual transportation trajectory curve and the ideal geodesic at the same moment, where the tangent vectors represent the rate and direction of change of the environmental parameters; The second derivatives of the actual transportation trajectory curve in each dimension are calculated to extract the curvature tensor, which represents the degree of severe environmental fluctuations. The extracted curvature tensor features are subjected to tensor product operation to obtain the damage scalar flow that records the product's historical path-dependent damage.

5. A dairy product traceability management system based on a cloud platform according to claim 1, characterized in that, The entropy increase index in the dynamic entropy increase index evaluation unit is a dimensionless value. The initial value of the entropy increase index is zero and it increases monotonically with the transportation process. The critical threshold for mass phase change is a critical point determined by regression analysis of historical deterioration accident data and corresponding curvature characteristics. The critical point corresponds to the situation where the second derivative of the entropy increase index exceeds a preset value.

6. The dairy product traceability management system based on a cloud platform according to claim 1, characterized in that, The process by which the adaptive risk blocking and feedback unit executes positive control logic is as follows: When the quality warning signal indicating severe deterioration or slight fatigue is received, an instruction is sent to the enterprise ERP system via the API interface. Change the inventory status field of the corresponding batch of products from "in transit and available for sale" to "frozen pending inspection" or "scrapped"; The system automatically updates the expiration date data based on the suggested dynamic shelf life value and triggers a price reduction promotion suggestion.

7. The dairy product traceability management system based on a cloud platform according to claim 1, characterized in that, The process by which the adaptive risk blocking and feedback unit executes the reverse optimization logic is as follows: The actual transportation trajectory curve data that was determined to be abnormal in this instance was marked as a negative sample. The negative sample is fed back to the mass manifold space mapping unit; The negative samples are used to fine-tune and correct the boundary of the non-destructive cold chain manifold model.

Images