A temperature over-standard self-adjusting alarm method for an unmanned cold chain vehicle with Beidou positioning

Abstract

This invention relates to the field of unmanned cold chain vehicle technology, specifically to a self-adjusting alarm method for temperature exceeding limits in unmanned cold chain vehicles equipped with BeiDou positioning. The method includes: spatiotemporally and synchronously fusing multi-source monitoring data of the unmanned cold chain vehicle based on spatiotemporal information provided by BeiDou positioning to obtain spatiotemporal reference collaborative data; predicting changes in the environmental heat load of the unmanned cold chain vehicle based on the location information in the spatiotemporal reference collaborative data, and calculating the permissible temperature range by combining preset cargo temperature control requirements and transportation stage information to obtain the temperature threshold at the current moment; performing correlation analysis based on the spatiotemporal reference collaborative data and the temperature threshold, and triggering corresponding self-adjusting actions and generating alarm information based on the analysis results. This method solves the problem of isolated spatiotemporal multi-source monitoring data, lack of collaborative analysis, leading to inaccurate self-adjusting decisions and high false alarm rates.

Description

Technical Field

[0001] This invention relates to the field of unmanned cold chain vehicle technology, specifically to an unmanned cold chain vehicle with Beidou positioning that automatically adjusts and alarms when the temperature exceeds the limit. Background Technology

[0002] In the field of unmanned cold chain transportation, the automatic adjustment and alarm of the temperature inside the vehicle is a core technology to ensure that high-quality goods such as refrigerated medicines and fresh food are kept "uninterrupted" throughout the entire process, which is directly related to the safety and economic value of the goods.

[0003] Currently, the system mainly relies on several sensors installed inside the vehicle. These sensors, together with the vehicle's built-in Beidou positioning module, door switch sensors, and refrigeration unit status sensors, form a monitoring network. The typical workflow is as follows: each sensor collects data and uploads it to the control unit. The control unit compares the received temperature data with a pre-set fixed threshold. When the reading of a sensor exceeds the threshold, an alarm is triggered and an attempt is made to start the preset self-adjustment program.

[0004] However, because the information from multiple sources such as temperature monitoring, vehicle positioning, and equipment status is isolated and fragmented at the acquisition and processing level, the system cannot accurately distinguish whether local temperature fluctuations are caused by refrigeration system failure, abnormal door opening, drastic changes in the external environment, or temporary changes in refrigerant distribution caused by vehicle movement. This results in a lack of precision in adjustment actions, often leading to false alarms or inappropriate responses to complex fault root causes. Summary of the Invention

[0005] To address the technical problem of isolated multi-source monitoring data in time and space, lack of collaborative analysis, leading to inaccurate self-regulation decisions and high false alarm rates, this application provides a self-regulating alarm method for unmanned cold chain vehicles with BeiDou positioning to detect temperature exceeding limits.

[0006] The method for automatic temperature over-limit alarm of unmanned cold chain vehicles with Beidou positioning provided in this application adopts the following technical solution: A self-adjusting alarm method for temperature exceeding limits in unmanned cold chain vehicles equipped with BeiDou positioning includes: Based on the spatiotemporal information provided by BeiDou positioning, spatiotemporal synchronous fusion of multi-source monitoring data of unmanned cold chain vehicles is performed to obtain spatiotemporal reference collaborative data. Based on the location information in the spatiotemporal reference collaborative data, the environmental heat load change of the unmanned cold chain vehicle is predicted. Then, combined with the preset cargo temperature control requirements and transportation stage information, the temperature permissible range is calculated to obtain the temperature threshold at the current moment. After performing correlation analysis based on spatiotemporal reference collaborative data and temperature thresholds, corresponding self-adjusting actions are triggered and alarm information is generated based on the analysis results.

[0007] Furthermore, the spatiotemporal information includes a time reference and geographic location coordinates. Based on the spatiotemporal information provided by BeiDou positioning, the steps for spatiotemporally fusion of multi-source monitoring data from unmanned cold chain vehicles to obtain spatiotemporal reference collaborative data include: Based on the time reference and geographic coordinates provided by BeiDou positioning, the multi-source monitoring data of multiple sensors deployed in the compartment of the unmanned cold chain vehicle are time-aligned and position-calibrated, resulting in synchronized multi-source monitoring data marked with time and position. After obtaining real-time altitude information corresponding to geographical coordinates and acquiring the three-dimensional motion state of the unmanned cold chain vehicle based on the time reference, and combining the vehicle's compartment structure, motion compensation and altitude compensation are performed on the synchronous multi-source monitoring data to obtain spatial calibration data. Based on the time reference and geographical coordinates, the spatial calibration data, door status and refrigeration unit status are recombined and packaged to obtain spatiotemporal reference collaborative data.

[0008] Furthermore, the steps for performing motion compensation and altitude compensation on the synchronous multi-source monitoring data to obtain spatial calibration data include: Based on the three-dimensional linear acceleration and three-dimensional angular velocity in the three-dimensional motion state, and according to the installation position coordinates of each sensor, the theoretical inertial overload vector of each sensor is calculated. The motion influence of temperature readings is obtained by vector superimposing the theoretical inertial overload vector, the refrigerant flow direction in the compartment, and the cargo loading distribution map in the compartment. Based on real-time altitude information, the absolute value of atmospheric pressure and the pressure difference per unit time are calculated. Based on the installation location coordinates, the distribution of air pressure difference inside and outside the carriage and the gradient of air pressure change inside and outside the carriage are calculated. Based on the distribution of air pressure difference inside and outside the carriage and the gradient of air pressure change inside and outside the carriage, the predicted amount of deformation of the carriage wall panel and the correction amount of equivalent thermal resistance of the carriage insulation layer are coupled and calculated to obtain the influence of altitude on temperature readings. Based on the influence of temperature readings on motion and the influence of temperature readings on altitude, vector synthesis and compensation correction are performed on synchronous multi-source monitoring data to obtain spatial calibration data.

[0009] Furthermore, based on the location information in the spatiotemporal reference collaborative data, the steps for predicting the environmental heat load changes of the unmanned cold chain vehicle include: Based on the geographic location coordinates in the spatiotemporal reference collaborative data and the obtained historical geographic location coordinates, the driving trajectory is extrapolated by combining the preset driving path to obtain the predicted path and vehicle position in the future time period. After obtaining the future altitude and solar radiation intensity corresponding to the predicted location of the vehicle, the atmospheric density change factor is calculated based on the future altitude, and the solar radiation heat flux is calculated based on the future solar radiation intensity. The atmospheric density change factor and the solar radiation heat flux are then weighted and fused to obtain the change in environmental heat load.

[0010] Furthermore, the steps for calculating the temperature threshold at the current moment, based on preset cargo temperature control requirements and transportation stage information, include: The nominal temperature permissible range is obtained from the preset cargo temperature control requirements, and the first temperature coefficient corresponding to the current stage is determined based on the transportation stage information; Based on historical temperature sequences, the cumulative duration and deviation of the current transport batch of goods that are not within the nominal temperature permissible range are calculated, and the temperature compensation coefficient is calculated based on the cumulative duration and deviation. The nominal temperature allowable range is widened based on the first temperature coefficient to obtain the stage tolerance range. Then, the stage tolerance range is corrected based on the temperature compensation coefficient to obtain the target temperature range. Based on the direction and magnitude of changes in environmental heat load, the target temperature range is directionally shifted to obtain the temperature threshold at the current moment.

[0011] Furthermore, the steps for correlation analysis based on spatiotemporal reference collaborative data and temperature thresholds include: Based on the temperature threshold, the spatial calibration data in the spatiotemporal reference collaborative data are compared. After identifying the abnormal monitoring data, the location attributes and timestamps corresponding to the abnormal monitoring data are extracted to obtain a set of candidate abnormal temperature points. Based on the location attributes and timestamps of each anomaly monitoring data in the candidate anomaly temperature point set, clustering and trend analysis are performed on the spatial distribution and time series changes of spatiotemporal reference collaborative data in the unmanned cold chain vehicle to obtain temperature anomaly feature vectors. The temperature anomaly feature vector is matched with the door status and refrigeration unit status in the spatiotemporal reference collaborative data, and multi-source information matching is performed under the time reference in the spatiotemporal reference collaborative data to obtain state events that are correlated with the current temperature anomaly. Based on a pre-defined fault mode knowledge base, state events are categorized and diagnosed to identify the root cause of the current temperature anomaly and its corresponding confidence level. The analysis results include the root cause and the confidence level.

[0012] Furthermore, based on the location attributes and timestamps of each anomaly monitoring data in the candidate anomaly temperature point set, the steps for clustering and trend analysis of the spatial distribution and time series changes of the spatiotemporal reference collaborative data within the unmanned cold chain vehicle to obtain the temperature anomaly feature vector include: Based on location attributes, spatial distance-based clustering is performed on each anomaly monitoring data to obtain spatial anomaly clusters. Then, the centroid location, spatial range, average temperature within the cluster, and temperature standard deviation of the spatial anomaly clusters are extracted to obtain a spatial clustering feature set. Based on the timestamp, the anomaly monitoring data belonging to the same spatial anomaly cluster are sorted according to their timestamps, and linear and nonlinear trend fitting based on time series is performed to obtain the temperature-time change equation and the corresponding rate of change of the spatial anomaly cluster, and generate a time-domain trend feature set. The spatial clustering feature set and the temporal trend feature set are associated and fused according to the corresponding spatial anomaly clusters, and then arranged and normalized according to the preset feature dimension order to generate a temperature anomaly feature vector.

[0013] Beneficial effects achieved: This application provides a self-adjusting alarm method for temperature exceeding limits in unmanned cold chain vehicles with BeiDou positioning, comprising: spatiotemporally and synchronously fusing multi-source monitoring data of the unmanned cold chain vehicle based on the spatiotemporal information provided by BeiDou positioning to obtain spatiotemporal reference collaborative data; predicting the environmental heat load change of the unmanned cold chain vehicle based on the location information in the spatiotemporal reference collaborative data, and calculating the temperature permissible range by combining preset cargo temperature control requirements and transportation stage information to obtain the temperature threshold at the current moment; and triggering corresponding self-adjusting actions and generating alarm information based on the analysis results after performing correlation analysis on the spatiotemporal reference collaborative data and the temperature threshold.

[0014] In this application, by utilizing the spatiotemporal information provided by BeiDou positioning, multi-source monitoring data is spatiotemporally and synchronously fused to generate spatiotemporal reference collaborative data, effectively suppressing the isolation of various sensor data in the temporal and spatial dimensions. Next, based on the location information in this spatiotemporal reference collaborative data, changes in environmental heat load are predicted, and temperature thresholds are calculated by combining preset cargo temperature control requirements and transportation stage information. This ensures that the temperature judgment standard is no longer an isolated fixed value, but a dynamic value closely related to the specific spatiotemporal environment of the vehicle, cargo characteristics, and transportation process, thus avoiding false alarms caused by neglecting dynamic changes in the external environment and differences in internal states. Finally, correlation analysis is performed based on the spatiotemporal reference collaborative data and temperature thresholds to correlate temperature readings with vehicle status and equipment operation data at the same time and location, thereby identifying the cause of temperature anomalies and triggering targeted self-regulation actions and generating alarms including root cause diagnostic information. This achieves the transformation from isolated data to collaborative intelligence, fundamentally solving the problems of inaccurate self-regulation decisions and high false alarm rates caused by isolated data and lack of collaborative analysis. Attached Figure Description

[0015] Figure 1This is a flowchart illustrating the steps of a self-adjusting alarm method for temperature exceeding limits in an unmanned cold chain vehicle with Beidou positioning, as described in this application. Figure 2 A flowchart illustrating the steps involved in generating spatiotemporal reference collaborative data for this application; Figure 3 This is a schematic diagram illustrating the steps involved in examining changes in environmental heat load and the current temperature threshold in this application. Figure 4 This is a schematic diagram illustrating the steps of this application to perform correlation analysis between spatiotemporal reference collaborative data and temperature thresholds. Detailed Implementation

[0016] The following combination Figures 1 to 4 This application will be described in further detail.

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative positional relationship and movement of the components in a specific posture. If the specific posture changes, the directional indications will also change accordingly.

[0019] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the use of "and / or" or "and / or" throughout the text includes three parallel solutions. For example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0020] This application discloses a method for automatic temperature adjustment alarm of unmanned cold chain vehicles with Beidou positioning.

[0021] Please refer to Figure 1The method for automatic temperature over-limit alarm of unmanned cold chain vehicles with Beidou positioning proposed in this embodiment includes steps S10~S30: Step S10: Based on the spatiotemporal information provided by BeiDou positioning, the multi-source monitoring data of the unmanned cold chain vehicle is spatiotemporally and synchronously fused to obtain spatiotemporal reference collaborative data.

[0022] In this step, by introducing the spatiotemporal information provided by BeiDou positioning as the time and space alignment benchmark for multi-source monitoring data, time-scale alignment and location calibration processing are forced on data from temperature sensors, door status sensors, and refrigeration unit status sensors at different locations within the vehicle. This constructs a fused data set at the source of data generation, where all monitoring data are attached to the same spatiotemporal coordinate system, i.e., spatiotemporal benchmark collaborative data. This allows any subsequent analysis to be based on multiple state information at the "same time and the same location," effectively eliminating analytical errors and causal misjudgments caused by time asynchrony and unclear location. It enables the integration of all originally isolated and fragmented monitoring data into a single spatiotemporal benchmark collaborative data set with consistent internal logic and clear spatiotemporal relationships, laying the data foundation for environmental prediction, threshold calculation, and correlation analysis in subsequent steps.

[0023] Step S20: Based on the location information in the spatiotemporal reference collaborative data, the environmental heat load change of the unmanned cold chain vehicle is predicted. Then, combined with the preset cargo temperature control requirements and transportation stage information, the temperature permissible range is calculated to obtain the temperature threshold at the current moment.

[0024] Based on the spatiotemporal reference coordination data established in the previous step, the location information is used to predict the environmental heat load changes that the unmanned cold chain vehicle will face. This incorporates the future impact of the external environment into the current decision-making considerations, making the adjustment more predictable. Next, by combining the preset cargo temperature control requirements and transportation stage information, the rigid requirements of the cargo itself are unified with the management of the transportation process. The temperature permissible range is calculated, and finally, a temperature threshold is generated based on the specific geographical location of the unmanned cold chain vehicle, the upcoming climate conditions, the specific characteristics of the transported cargo, and the current transportation stage. This temperature threshold is no longer a rigid fixed value, but a judgment value that can adaptively adjust according to the current state. Thus, under the core premise of ensuring cargo safety, unnecessary alarms and adjustments triggered by normal environmental fluctuations or transportation stage transitions are minimized, significantly improving the accuracy and intelligence of decision-making.

[0025] Step S30: After performing correlation analysis based on spatiotemporal reference collaborative data and temperature threshold, trigger the corresponding self-adjustment action and generate alarm information according to the analysis results.

[0026] Based on the spatiotemporal reference coordinated data and temperature thresholds calculated in the above steps, the spatiotemporal reference coordinated data and temperature thresholds are correlated and analyzed. This allows for the matching of abnormal monitoring data exceeding or nearing the limits with the door status and refrigeration unit status at the same time and spatial coordinates, thereby determining the true cause of the data anomaly. Based on this, the system triggers the most matching self-adjusting action according to the diagnosed true cause, and simultaneously generates a diagnostic alarm message containing the time and location of the anomaly, the root cause of the diagnosis, and the countermeasures already implemented. This elevates the system from a passive alarm device that responds to a single parameter to an intelligent operation that can understand the problem, handle it accurately, and report clearly. It achieves targeted self-adjusting actions and operable alarm information, effectively solving the problem of inaccurate decision-making caused by the lack of data collaborative analysis.

[0027] It should be noted that, based on the analysis results including the root cause and confidence level, and the diagnostic result-action policy mapping rule base preset by the adaptive alarm system, the diagnostic result-action policy mapping rule base defines the execution instructions with priority order corresponding to different root causes.

[0028] When an automatic adjustment action is triggered, the system determines whether the execution threshold is met based on the confidence level. For example, automatic execution is only performed when the confidence level is high or medium; otherwise, manual confirmation is requested. When the execution conditions are met, the system queries the root cause diagnosis result-action strategy mapping rule base to obtain the corresponding action instruction set and converts it into control commands that can be sent to the vehicle actuators. For example, if the root cause is diagnosed as "local air conditioning leakage due to poor sealing of the left rear door," the triggered automatic adjustment actions include "immediately increasing the speed of the circulating fan in that area to the highest level" and "sending an auxiliary command to the vehicle control system to 'check the left rear door latch'." If the root cause is diagnosed as "decreased evaporator fan efficiency," the triggered automatic adjustment actions are "starting the backup fan circuit" and "increasing the main fan power by 10%." All control commands are sent to the corresponding actuators (such as frequency converters, solenoid valves, and fan controllers) in real time via the vehicle bus.

[0029] Simultaneously, the process of generating alarm information is initiated: based on a predefined structured alarm template, key fields are automatically populated. The values ​​of these fields are directly extracted from the analysis process, including the "time of anomaly occurrence" (event baseline) and "location" (specific geographical coordinates and description of the location within the vehicle) extracted from spatiotemporal baseline collaborative data, the text of the root cause from the analysis results, and the self-regulating actions that have been issued and executed. In addition, the confidence level is added to the alarm as metadata. Finally, the generated alarm information, with complete format and comprehensive information, is simultaneously sent to the remote monitoring center and the vehicle's local human-machine interface via a wireless communication module, thus completing a closed loop from intelligent analysis to precise execution and clear reporting.

[0030] In one feasible implementation, refer to Figure 2 As shown, step S10 may specifically include steps S11 to S13: Step S11: Based on the time reference and geographic coordinates provided by BeiDou positioning, perform time alignment and location calibration on the multi-source monitoring data of multiple sensors deployed inside the unmanned cold chain vehicle, resulting in synchronized multi-source monitoring data marked with time and location.

[0031] First, a high-precision and consistent time reference and current geographical coordinates (including longitude, latitude, and altitude) are obtained in real time from the BeiDou positioning system. Inside the unmanned refrigerated truck, temperature sensors, door status sensors, and refrigeration unit status sensors each generate monitoring data at their respective sampling frequencies.

[0032] The self-adjusting alarm system marks each monitoring data point with a time reference obtained from BeiDou positioning at the instant it is generated. This ensures that events that occur physically at the same time from different sensors (such as temperature changes, door opening, and compressor starting) are marked as the exact same time point at the data level, eliminating time deviations caused by asynchronous clocks or interleaved sampling periods within the sensors. Meanwhile, the vehicle structure model clearly defines a vehicle coordinate system with the vehicle's center of mass or a fixed point as the origin. This vehicle coordinate system is fixedly connected to the vehicle. Each temperature sensor, door status sensor, and refrigeration unit status sensor has a pre-measured and recorded installation position coordinate in this vehicle coordinate system. This is a position vector relative to the origin of the vehicle coordinate system. The geographical coordinates obtained from Beidou positioning are essentially the precise longitude, latitude, and altitude of the origin of the vehicle coordinate system (or a known Beidou antenna installation point with a fixed offset from the origin) in a geographic coordinate system (such as the WGS-84 coordinate system). At this time, the current heading angle of the unmanned cold chain vehicle is obtained (which can be provided by Beidou carrier phase observation or inertial navigation unit). This heading angle defines the rotation angle of the vehicle coordinate system relative to geographic north. Using the principle of spatial coordinate transformation, the installation position coordinates of the sensors in the vehicle coordinate system are rotated according to the heading angle (usually around the vertical axis) so that their direction is aligned with the direction of the geographic coordinate system. Then, the rotated installation position coordinates are vector-added with the geographic location coordinates obtained from BeiDou (which have been converted to three-dimensional rectangular coordinates) to obtain the absolute position coordinates (longitude, latitude, and altitude) of the sensor in the geographic coordinate system. This absolute position, along with the corresponding time reference, is then appended to the data record of the corresponding sensor.

[0033] Through the above process, all the originally isolated multi-source monitoring data, which only contained their own physical quantity readings, were forcibly given a unified time reference and a clear geographical location attribute, thereby generating synchronous multi-source monitoring data marked with time and location. This constructed a reference system in which all monitoring data were completely aligned in time and space and were traceable. This enabled any judgment on the status of the carriage in subsequent analysis (e.g., whether the temperature exceeded the standard at a specific time and at a specific geographical coordinate point) to be based on completely synchronous and clearly located multi-source information (temperature, door status, and unit status). This laid the data foundation for subsequent correlation analysis (e.g., determining whether the temperature rise was related to a door opening event near the same location at the same time), and fundamentally solved the problem that multi-source monitoring data could not be effectively collaboratively analyzed due to inconsistencies in the collection time and location reference.

[0034] Step S12: Obtain real-time altitude information corresponding to the geographical coordinates, and after obtaining the three-dimensional motion state of the unmanned cold chain vehicle based on the time reference, combine the vehicle's compartment structure to perform motion compensation and altitude compensation on the synchronous multi-source monitoring data to obtain spatial calibration data.

[0035] In this step, in order to resolve the physical interference caused by the three-dimensional motion state of the unmanned cold chain vehicle and the changes in the altitude of the external environment on the sensor readings, so as to ensure that the multi-source monitoring data can accurately reflect the heat exchange state inside the vehicle, and provide a basis for diagnostic decisions in subsequent steps.

[0036] To achieve the above objectives, the system first acquires real-time altitude information corresponding to the geographic coordinates and obtains the three-dimensional motion state of the unmanned refrigerated truck in three-dimensional space based on a time reference (including linear acceleration and angular velocity along three axes). The dynamic information of the unmanned refrigerated truck (i.e., its three-dimensional motion state) and its static environmental information (i.e., real-time altitude information) are used as inputs. Combined with the vehicle structure of the unmanned refrigerated truck (such as the installation coordinates of each sensor, refrigerant pipeline layout, and cargo loading distribution), the system compensates for the synchronous multi-source monitoring data generated in the previous step: ① During motion compensation, its function is to quantify and eliminate the instantaneous impact on local monitoring data caused by vehicle acceleration / deceleration, bumps, etc., resulting in refrigerant inertial flow, cargo displacement, and pressure changes; ② During altitude compensation, its function is to quantify and eliminate the systemic impact on sensor measurements caused by air pressure changes due to altitude variations, leading to deformation of the container walls, changes in the thermal resistance characteristics of insulation materials, and refrigerant phase change points.

[0037] By compensating the synchronous multi-source monitoring data as described above, the resulting spatial calibration data can remove the noise generated by the unmanned cold chain vehicle's own movement and the external environment, thereby accurately revealing the true situation of the cooling effect and heat distribution inside the vehicle. This allows subsequent analysis to be based on a more fundamental thermodynamic state, greatly improving the authenticity of temperature anomaly detection and the accuracy of root cause analysis.

[0038] Furthermore, step S12 also includes steps S121 to S125: Step S121: Based on the three-dimensional linear acceleration and three-dimensional angular velocity in the three-dimensional motion state, and according to the installation position coordinates of each sensor, calculate the theoretical inertial overload vector of each sensor.

[0039] First, two key physical quantities characterizing the overall motion state of the unmanned cold chain vehicle are obtained: the three-dimensional linear acceleration (acceleration values ​​along the three axes of front-to-back, left-to-right, and up-down) and the three-dimensional angular velocity (rotational angular velocity values ​​around the above three axes) at the center of mass of the unmanned cold chain vehicle. At the same time, the installation position coordinates of each sensor relative to the center of mass of the vehicle are retrieved. These installation position coordinates are a position vector pointing from the center of mass of the vehicle to the sensor installation point.

[0040] Because the unmanned cold chain vehicle has rotational motion (three-dimensional angular velocity), the sensor's location will rotate around the vehicle's center of mass, generating additional acceleration. This acceleration is a combination of tangential acceleration (related to the three-dimensional angular acceleration) and centripetal acceleration (related to the square of the three-dimensional angular velocity and the position vector).

[0041] Secondly, in a unified vehicle coordinate system, let the linear acceleration vector of the vehicle's center of mass be... Angular velocity vector is The position vector of the sensor relative to the vehicle's center of mass is The acceleration caused by the rotational motion of the sensor is calculated, which consists of two parts: tangential acceleration and centripetal acceleration. By angular acceleration (Right now (derivative with respect to time) and position vector The cross product is given, i.e. Centripetal acceleration From angular velocity With linear velocity The cross product is given, i.e. By vector combining these two accelerations caused by rotation, the rotational acceleration at the sensor location can be obtained. Then, the linear acceleration of the vehicle's center of gravity is... Rotational acceleration at the sensor location By performing vector addition, the total absolute acceleration of the sensor mounting point in three-dimensional space can be obtained. Since the sensors themselves measure the local environmental state at their mounting points attached to the vehicle body, in order to separate the overload caused purely by the vehicle's non-uniform motion (i.e., non-inertial frame effect), it is necessary to subtract the gravitational acceleration from the total absolute acceleration to obtain the theoretical inertial overload vector at the corresponding sensor location. This theoretical inertial overload vector describes the magnitude and direction of the inertial force acting on a unit mass object at the corresponding location due to the vehicle's acceleration, deceleration, turning, or bumping. It provides a spatiotemporal distribution map of the inertial force at each sensor point for subsequent analysis. This quantitative description is the fundamental physical input for evaluating how the motion of the unmanned refrigerated truck interferes with sensor readings by affecting refrigerant flow, cargo displacement, and internal air convection. This allows subsequent compensation to correct the measurement errors caused by the motion of the unmanned refrigerated truck in principle rather than just empirically, providing a computational basis for obtaining spatial calibration data that accurately reflects the true thermal state inside the vehicle.

[0042] Among them, gravitational acceleration is a known physical constant generated by the Earth's gravitational field.

[0043] Step S122: The theoretical inertial overload vector, the refrigerant flow direction of the carriage, and the cargo loading distribution map of the carriage are vector superimposed to obtain the motion influence of the temperature reading.

[0044] First, retrieve the pre-stored refrigerant flow direction and cargo loading distribution map of the unmanned refrigerated vehicle. The refrigerant flow direction describes the vector field information of the refrigerant (such as refrigerant or cold air) in the refrigeration system pipes of the unmanned refrigerated vehicle, which can indicate how the refrigerant flows through each sensor in a static or equilibrium state. The cargo loading distribution map is a three-dimensional model describing the space occupation, stacking density and relative positional relationship of the cargo in the vehicle with each sensor.

[0045] When performing vector superposition, the theoretical inertial overload vector is first decomposed into two influencing components. The first influencing component is the projection of the theoretical inertial overload vector onto the local refrigerant flow direction near the sensor. By calculating the magnitude and direction of the projection of this theoretical inertial overload vector onto the refrigerant flow direction, the disturbance assessment value of the unmanned refrigerated truck's movement on the local refrigerant flow velocity, pressure, and even phase change process is quantified. Specifically, by calculating the scalar projection of the theoretical inertial overload vector onto the refrigerant flow direction, the inertial force component along the refrigerant flow direction is obtained. This inertial force component directly corresponds to the pressure gradient that drives or hinders the refrigerant flow. The calculated inertial force component along the refrigerant flow direction is divided by the characteristic cross-sectional area of ​​the refrigerant pipeline and the refrigerant density to transform it into an equivalent pressure gradient. Then, the pressure gradient Substituting this into the flow rate calculation model based on the Darcy-Visbach formula in fluid mechanics, this model expresses the relationship that the pressure gradient is proportional to the square of the volumetric flow rate in steady-state flow, specifically in the form of: ,in, For the pre-calibrated friction coefficient, The equivalent length-to-diameter ratio of the pipe near the sensor. For refrigerant density, For the flow velocity, the pressure gradient and flow velocity under the current steady state are used as a reference, and the pressure gradient is introduced by solving for it. The new flow velocity corresponding to the new pressure gradient The instantaneous change rate of the cold medium's accumulation flow can then be calculated. Next, based on the refrigerant's physical properties (such as the saturated pressure-temperature curve), the temperature offset of the pressure gradient relative to the phase change temperature (such as the evaporation temperature) at the current evaporation pressure is evaluated. Then, the calculated instantaneous rate of change of the refrigerant's cumulative flow rate and the evaluated temperature offset are used as input vectors to a transfer function constructed through experimental calibration. This transfer function is a mathematical mapping (such as a trained artificial neural network or an empirical formula), whose internal parameters characterize the combined influence of changes in refrigerant flow rate and evaporation temperature at a specific sensor location on the local convective heat transfer coefficient and thus on the sensor's theoretical reading. Its output is the refrigerant flow disturbance evaluation value.

[0046] The second influencing component assesses the direction and magnitude of the theoretical inertial overload vector relative to the surface of the cargo adjacent to the sensor. Combined with the cargo mass, density, and gap between the cargo and the sensor revealed by the cargo loading distribution map, this quantifies the thermal resistance between the sensor and the cargo caused by possible micro-displacement or pressure changes in the cargo, yielding a thermal resistance disturbance value. Specifically, the theoretical inertial overload vector is decomposed into a normal component perpendicular to the surface of the cargo adjacent to the sensor. This normal component, combined with the effective mass of the cargo and the stiffness characteristics of the contact surface provided by the cargo loading distribution map, calculates the instantaneous change in contact pressure between the cargo and the carriage wall or sensor mounting surface caused by inertial force. Then, using this calculated instantaneous change in contact pressure as input, the corresponding change in contact thermal resistance is directly obtained through the contact pressure-contact thermal resistance correspondence formula. This is the thermal resistance disturbance value. The contact pressure-contact thermal resistance correspondence formula used is an empirical formula derived from fitting experimental data; its common form is... ,in, Represents contact thermal resistance. This represents the instantaneous change in contact pressure, while , and These are fitting parameters obtained through standard tests (such as steady-state heat flow method tests) on specific cargo materials and combinations of materials for the compartment wall (or sensor mounting surface). This formula describes the physical law that contact thermal resistance decreases with increasing contact pressure in a power-law relationship and asymptotically approaches a minimum. In calculations, the instantaneous change in contact pressure obtained in real time is used. Substituting into the formula, the new contact thermal resistance is calculated. The thermal resistance value under the reference contact pressure. Subtracting the two values ​​yields the thermal resistance disturbance value. .

[0047] Based on the influence coefficient mapping model constructed according to the principles of fluid mechanics and heat transfer, the specific values ​​of the weighting coefficients w1 and w2 were determined during the initialization of the adaptive alarm system through regression analysis of historical operating data or simulation analysis based on heat transfer mechanisms. The aforementioned refrigerant flow disturbance assessment value and the thermal resistance disturbance value related to cargo contact are weighted and summed to finally calculate a temperature reading motion influence quantity with a sign (indicating whether it leads to a higher or lower reading) and amplitude. During the weighted summation, a weighting coefficient w1 is assigned to the refrigerant disturbance assessment value, and a weighting coefficient w2 is assigned to the thermal resistance disturbance value (these weighting coefficients w1 and w2 are determined through historical data fitting analysis, reflecting the relative importance of the two disturbance mechanisms at the sensor location). Temperature reading motion influence quantity = (w1 * refrigerant flow disturbance) The evaluation value is calculated as (w2 * thermal resistance disturbance value). The sign of the refrigerant flow disturbance evaluation value is determined by whether the inertial force component is in the same direction or opposite to the refrigerant flow direction (if they are in the same direction, the heat transfer is enhanced and the temperature reading is lower, so the sign is negative; if they are opposite, the heat transfer is weakened and the temperature reading is higher, so the sign is positive). This transforms the motion acceleration data of the unmanned cold chain vehicle into the expected physical interference value for each sensor reading. This allows us to distinguish how much of the change in the reading is caused by false signals due to the external excitation of vehicle motion, providing a scientific basis for subsequently removing motion interference from the original readings and restoring the true state inside the compartment.

[0048] Step S123: Calculate the absolute value of atmospheric pressure and the pressure difference per unit time based on real-time altitude information, and calculate the distribution of air pressure difference inside and outside the carriage and the gradient of air pressure change inside and outside the carriage based on the installation location coordinates.

[0049] Based on the pre-set standard atmospheric pressure model in the adaptive alarm system, such as using the International Standard Atmosphere (ISA) formula... ,in, Standard sea level pressure, For temperature lapse rate, The standard temperature at sea level It is the acceleration due to gravity. The molar mass of dry air. This is the universal gas constant. Altitude. This refers to the real-time altitude information of the unmanned refrigerated trucks. Substituting this into the formula as an input variable, the absolute value of the atmospheric pressure at the current location can be directly calculated. Subsequently, based on the known relationship between air pressure and altitude, i.e., the vertical pressure gradient... ,in, The air density is combined with the rate of change of altitude obtained in real time from the location information. (i.e., the change in altitude per unit time), the pressure difference per unit time is calculated using the chain rule: pressure difference per unit time The calculation results directly characterize the instantaneous rate of change of ambient atmospheric pressure over time caused by altitude changes due to vehicle movement.

[0050] Next, based on the design premise of good airtightness of the carriage structure, it is assumed that the interior of the carriage is a relatively enclosed space, where the internal air does not directly and rapidly exchange macroscopic mass with the external atmosphere. However, due to the intermittent operation of the refrigeration system, slight permeation of the carriage material, and possible ventilation openings, the total internal air pressure will fluctuate around an equilibrium point. This equilibrium point and the external atmospheric pressure have a reference difference related to altitude due to the difference in air resistance caused by altitude changes. It can be calibrated to altitude through experiments. The function of pressure; based on the hydrostatic relationship, pressure in a static fluid changes linearly with height, as shown by the formula: ,in, As a reference point pressure, This represents the vertical height difference relative to a reference point. The calculation uses the altitude of a predetermined reference point inside the carriage (usually the geometric center of the carriage) as the reference point. and corresponding internal air pressure (Its value is the current external atmospheric pressure) Add the benchmark difference Using this as a reference, for each sensor, its installation position coordinates are read, especially its vertical height difference relative to a preset reference point inside the carriage. Substituting into the statics formula, calculate the theoretical air pressure inside the compartment at the sensor location. ,in, The internal air density is estimated based on the average temperature inside the compartment. Finally, the pressure difference distribution between the inside and outside of the compartment is obtained by calculating the difference between the absolute value of the external atmospheric pressure at each sensor location and the theoretical air pressure inside the compartment. ,in, Based on the actual altitude corresponding to the sensor location By matching the pre-set parameters, the specific air pressure difference values ​​at each sensor installation point in the three-dimensional space inside the carriage can be obtained. This creates a pressure difference between the inside and outside of the carriage.

[0051] And, the pressure difference per unit time calculated above As the input source driving changes in external ambient air pressure, the adaptive alarm system uses a pre-set lumped parameter model to describe the dynamic response characteristics of the cabin air pressure. The core parameter of this lumped parameter model is the overall "sealing parameter" of the cabin structure (usually expressed as a time constant). The dynamic response of the air pressure inside the carriage is characterized by the time constant required for the internal air pressure to reach equilibrium in response to an external step change (which can be obtained experimentally or calculated based on the total equivalent leakage area of ​​the carriage gaps) and the status of each ventilation opening (such as the opening degree of ventilation valves and the filter resistance coefficient, which determine the preferred path of airflow and local flow resistance). In simulating the transmission process, the change in external air pressure is considered as an excitation to the closed carriage system, and the dynamic response of the air pressure inside the carriage is... It can be approximated by a first-order inertial element, and its transitivity is expressed in the Laplace domain as follows: ,in, This refers to changes in external air pressure. In the time domain, it means the rate of change of internal air pressure. Pressure difference between the outside and inside Proportional, that is On a short timescale, the current rate of change of external air pressure can be... Treating this as an excitation, the rate of change of the average air pressure inside the carriage is calculated by solving the dynamic equation or by discretizing and iterating. However, due to the presence of air vents and the internal structure of the carriage, the transmission of air pressure changes is not uniform. This non-uniformity needs to be characterized by a spatial influence weighting coefficient matrix, which combines the condition of the air vents and their installation location coordinates. This spatial influence weighting coefficient matrix defines the coordinates of each installation location. Internal pressure change rate , Rate of change of average air pressure inside the carriage The relationship between them is usually expressed as ,in, It is a location The impact factor (e.g., location near vents or corners) A value greater than 1 indicates a faster response; located behind a sealed cargo stack. (Possibly less than 1, indicating a response lag) The air pressure difference between the inside and outside of the carriage at the sensor location The change at a certain point, i.e., the gradient of air pressure change inside and outside the carriage. Calculated using the following formula: ,in, It is a known pressure difference per unit time. This is the calculated rate of change of air pressure inside the corresponding sensor installation point. This gradient value quantitatively describes the instantaneous rate of change of the pressure difference between the inside and outside of the carriage felt at each specific sensor location per unit time, due to the combined effect of external air pressure changes and the dynamic response inside the carriage. Its spatial distribution (i.e., the gradient values ​​are different at each sensor installation point) is a direct reflection of the transmission rate and spatial non-uniformity.

[0052] Step S124: Based on the distribution of air pressure difference inside and outside the carriage and the gradient of air pressure change inside and outside the carriage, the predicted deformation of the carriage wall panel and the correction of the equivalent thermal resistance of the carriage insulation layer are coupled and calculated to obtain the influence of altitude on temperature readings.

[0053] Based on the air pressure difference distribution inside and outside the carriage, each sensor reads the local air pressure difference experienced by the wall panel near its installation location. Local pressure difference As a uniformly distributed load perpendicular to the wall panel Substituting the bending control equation of the wall panel based on the theory of small deformation of thin plates, i.e. the bending stiffness of the wall panel... With the normal deflection of the wall panel Fourth-order partial differential equations Among them, bending stiffness The elastic modulus H, Poisson's ratio V, and plate thickness j in the geometric structure of the carriage wall panel material are used to formulate... The calculation yielded the normal deflection of the wall panel. Based on this, the strain-displacement geometric relationship of thin plate theory, i.e., the strain components within the plate, is used. , , With the normal deflection of the wall panel curvature , and torque relation , , Among them, curvature , Torque ,in, Given the vertical distance from the mounting point to the mid-surface of the wall panel, calculate the in-plane strain components at any mounting point within the wall panel; then, based on the in-plane stress components for isotropic materials... , , The relationship with the in-plane strain components is as follows , , By substituting the calculated in-plane strain components into these equations, the spatial distribution of the in-plane membrane stress σ caused by the deflection can be calculated. This in-plane membrane stress σ is related to the normal deflection of the wall panel. This constitutes the predicted deformation of the carriage wall panel under the local pressure differential load. This predicted deformation will alter the contact area between the carriage wall panel and the internal air, as well as the convective heat transfer conditions, thus affecting heat transfer. and These are two mutually perpendicular rectangular coordinate directions within the plane of the carriage wall panel; It is the vertical distance from the installation point to the middle surface of the wall panel; It is the internal friction force (shear stress) per unit area of ​​the wall panel. It is the relative displacement per unit length between adjacent layers inside the wall panel (shear strain).

[0054] At the same time, based on the gradient of air pressure changes inside and outside the carriage And combined with the material composition of the body insulation layer (including the thickness of the sealed air gap) ) and physical parameters, based on the thermal conductivity of air With pressure Approximately proportional (in the region where the molecular mean free path is limited by the gap size), the current average air pressure Gradient of air pressure change inside and outside the carriage Substituting this proportional relationship, calculate the rate of change of air gap thermal conductivity per unit time. Then reduce the air gap thermal resistance Thermal resistance of solid materials Adding them together gives the overall equivalent thermal resistance. Among them, air gap thermal resistance Due to air gap thickness thermal resistance of air gap The ratio is defined as follows: The thermal resistance of solid materials It is a constant value; due to the thermal resistance of the air gap. Changes over time, affecting the overall equivalent thermal resistance Taking the time derivative yields This is due to the instantaneous change in the thermal conductivity of the air gap, resulting in the overall equivalent thermal resistance of the body insulation layer. The instantaneous change; this instantaneous change Calculating the step size Internal time Integrating this equation yields the equivalent thermal resistance correction of the cabin insulation layer caused by air pressure changes within that step length. ,in, It is the differential of time t.

[0055] Next, based on the principles of heat transfer, a local one-dimensional composite heat transfer model is constructed for the sensor installation location. This local one-dimensional composite heat transfer model represents the net heat flux density penetrating the vehicle body wall. Among them, total thermal resistance The sum of the series thermal resistances of each stage along the heat flow path, External ambient temperature, The temperature inside the carriage; the calculated temperature Compared with the baseline heat flux density without altitude change The difference is used to obtain the change in heat flow. .in, The reference convective heat transfer coefficient under no deformation. The baseline equivalent thermal resistance of the body insulation layer under conditions without changes in altitude (air pressure) is given. It is the equivalent thermal resistance from the inner wall of the carriage to the internal air.

[0056] Finally, the local environment of the sensor and its mounting point is considered as having a heat capacity C and a thermal resistance. For a first-order thermal system, the change in heat flow is... The action on this first-order thermal system will cause a steady-state shift in the temperature at the sensor detection point. Through the pre-calibrated sensor thermal response coefficient (Unit: °C / (W / m)) 2 The calculated heat flow change value Directly convert to temperature readings and altitude effect This value can be positive or negative.

[0057] This method transforms abstract altitude changes into the altitude-related influence on temperature readings generated by each sensor. This allows for the identification of how much of the sensor reading variation is caused by physical factors such as changes in external air pressure, providing a basis for correction to subsequently remove interference from altitude changes from the readings and restore the true thermodynamic state inside the carriage.

[0058] Step S125: Based on the influence of temperature readings on motion and the influence of temperature readings on altitude, vector synthesis and compensation correction are performed on the synchronous multi-source monitoring data to obtain spatial calibration data.

[0059] The motion effect and altitude effect of temperature readings for each temperature sensor, calculated by the above steps, are obtained, and vector synthesis and compensation correction are performed on the synchronous multi-source monitoring data.

[0060] For each temperature sensor's raw reading, the motion effect of the temperature reading and the altitude effect of the temperature reading under the same spatiotemporal reference are algebraically added together to obtain an estimate of the total physical interference experienced by a corresponding temperature sensor reading.

[0061] Subsequently, compensation and correction are performed: the estimated value calculated above is subtracted from the synchronous multi-source monitoring data of the temperature sensor, thereby removing the noise components (i.e., estimated values) introduced by known external physical factors (vehicle movement, altitude changes) that have been identified and quantified from the contaminated synchronous multi-source monitoring data, and obtaining spatial calibration data. This spatial calibration data can more realistically reflect the state and changes of heat exchange inside the vehicle compartment itself, providing a reliable data foundation that almost reflects the "intrinsic thermal state" for subsequent accurate anomaly diagnosis, dynamic threshold comparison, and intelligent decision-making based on the temperature field. This greatly reduces the risk of false alarms and misjudgments caused by sensor readings being interfered with by non-thermodynamic factors such as vehicle bumps, turns, or mountain crossings.

[0062] Step S13: Based on the time reference and geographical coordinates, the spatial calibration data, door status and refrigeration unit status are recombined and packaged to obtain spatiotemporal reference collaborative data.

[0063] Based on the unified time reference and corresponding geographic coordinate sequence continuously acquired from BeiDou positioning as a global benchmark, for each time reference and its corresponding geographic coordinates, the spatial calibration of all temperature sensors in the vehicle under that time reference, after motion and altitude compensation, is extracted from the spatial calibration data. At the same time, the vehicle door status (such as open, closed, and locked status) and refrigeration unit status (such as compressor start / stop, fan speed, operating current, set temperature, and other operating parameters) recorded at the same time reference are extracted from the synchronous multi-source inspection data.

[0064] Next, the three types of information—spatial calibration data representing the internal thermal state, door state representing the boundary conditions, and refrigeration unit state representing the working state of the refrigeration system—are bound to their common time reference and geographical coordinates, and packaged into a data unit with a unified structure. This data unit ensures that all the state information contained therein strictly corresponds to the same instant and the same location.

[0065] By continuously performing the above operations at each sampling moment, the generated series of data units are arranged and encapsulated according to the time sequence to form spatiotemporal reference collaborative data. This constructs a data set with completely consistent internal spatiotemporal logic and precise alignment of multi-dimensional information. This ensures that when any subsequent analysis program retrieves data at a certain moment, the obtained data on the temperature field distribution of the carriage, the opening and closing status of the doors, the operating parameters of the refrigeration unit, and the geographical location of the vehicle are necessarily synchronous and correlated. This provides a data foundation for predicting changes in environmental heat load in step S20 and for conducting symptom-root cause correlation analysis in step S30.

[0066] In one feasible implementation, refer to Figure 3As shown, step S20 may specifically include steps S21 to S26: Step S21: Based on the geographic location coordinates in the spatiotemporal reference collaborative data and the obtained historical geographic location coordinates, the driving trajectory is extrapolated in combination with the preset driving path to obtain the predicted path and vehicle predicted position in the future time period.

[0067] A series of time-referenced geographic location coordinates are extracted from the spatiotemporal reference collaborative data to form a spatiotemporal point sequence describing the recent movement trajectory of the unmanned cold chain vehicle. Then, the most recent (e.g., the last 10) consecutive data points are extracted from this sequence. Each data point contains a timestamp tk provided by BeiDou and its corresponding geographic location coordinates. To calculate the motion characteristics, the latitude and longitude coordinates are first converted to local tangent plane rectangular coordinates (such as the ENU coordinate system) to obtain a coordinate sequence. When calculating the average velocity, the displacement between the two most recent consecutive points is taken. Divide by time interval To obtain instantaneous velocity Then, calculate the arithmetic mean of the instantaneous velocity sequences within a short time window (e.g., the most recent 5) to obtain the average velocity within that short time window. Next, a linear fit is performed on the instantaneous velocity sequence with respect to the timestamp tk, and the slope of the fitted line is the average linear acceleration. And calculate the heading angle based on the rectangular coordinates of the continuous installation points. (With true north as 0 degrees), then calculate the change in the continuous heading angle. Then, the average rate of change of the heading angle sequence within the short-time window is calculated with respect to the time interval to obtain the average heading rate of change. .

[0068] Through the above calculations, motion features describing the current motion state of the unmanned cold chain vehicle can be quantitatively extracted from the original spatiotemporal reference collaborative data.

[0069] Simultaneously, a preset driving path is obtained, which is the planned route based on the current transportation task of the unmanned cold chain vehicle, typically defined by a series of waypoints or road link sequences. When extrapolating the driving trajectory, it is not a blind calculation in free space, but rather, under the spatial constraints of the preset driving path, the average speed is calculated using the current position of the unmanned cold chain vehicle as the starting point of the path. and current heading angle As the initial kinematic state, and after accessing a digital map including road topology information, the system obtains the attributes of the upcoming road segment on the preset driving path, such as the road curvature radius, the location and type of the upcoming intersection, and the speed limit of the segment. At the beginning of each iterative calculation, the system first determines the road segment the driver will enter based on the projection of the starting point onto the preset path, and then reads the curvature radius and speed limit of that road segment. Based on the physical constraints of cornering, the system determines the upper limit of the safe speed allowed for that corner. The current speed is compared with the upper limit of the safe speed and the speed limit, and the minimum value is taken as the theoretical target speed for that step. If an upcoming intersection is detected, the theoretical target speed is further dynamically adjusted based on the intersection type (such as traffic light control, stop and yield) and the preset passage strategy (such as cautious passage requiring a certain reduction in speed). Then, based on the current speed... Calculate the step size based on the theoretical target velocity. Actual acceleration within Afterwards, update the speed of the unmanned cold chain vehicle. and the distance traveled along the road path ,in, This represents the current distance.

[0070] Based on the updated forward distance Interpolating along the geometric alignment of the preset driving path, the corresponding latitude and longitude coordinates are calculated, yielding the predicted position for the future time. This position, speed, and timestamp are recorded as a predicted point, and updated... The road attributes, reassessed at the new location, are used as input for the next iteration. This process is repeated until the entire future time period is covered, thereby generating a predicted path.

[0071] The predicted location of the unmanned cold chain vehicle at a series of equally spaced future time points (e.g., one point every 5 seconds) along the predicted path is thus provided as a reliable estimate of the spatial-temporal framework of the unmanned cold chain vehicle in the short term. This enables subsequent steps to obtain environmental parameters (e.g., altitude, solar radiation) based on the predicted vehicle location, thereby elevating the perception and decision-making of the self-adjusting alarm system from the current passive state to a proactive predictive state for the future, laying the spatiotemporal foundation for the dynamic adjustment of the temperature threshold.

[0072] Step S22: After obtaining the future altitude and solar radiation intensity corresponding to the predicted location of the vehicle, calculate the atmospheric density change factor based on the future altitude and the solar radiation heat flux based on the future solar radiation intensity. Then, perform weighted fusion of the atmospheric density change factor and the solar radiation heat flux to obtain the change in environmental heat load.

[0073] Based on the predicted vehicle locations obtained in the previous step, the elevation value corresponding to each predicted vehicle location is retrieved, i.e., the future altitude. At the same time, the latitude and longitude, date and time, and possible weather forecast information of the predicted vehicle locations are input from the integrated meteorological data service interface to obtain the corresponding future solar radiation intensity.

[0074] Next, because atmospheric density decreases exponentially with increasing altitude, the system utilizes future altitudes, substituting the altitude of each predicted future vehicle location into the standard formula of the International Standard Atmosphere (ISA) model, which describes atmospheric density. The pattern of changes with altitude is commonly expressed in the following form: ,in, and These are the standard air density and temperature at a reference sea level (0 altitude), T is the temperature at the current altitude, g is the acceleration due to gravity, and R is the atmospheric constant. The atmospheric density at the predicted vehicle location can be directly calculated using this formula. Then compare it with the reference sea level density By comparison, the atmospheric density variation factor is obtained. This directly affects the convective heat transfer coefficient between the outer surface of the unmanned refrigerated vehicle and the air. Then, in calculating the solar radiation heat flux based on future solar radiation intensity, it is necessary to consider the solar radiation intensity, the angle of sunlight incident on the outer surface of the vehicle (calculated from the vehicle's location, time, and date), and the absorptivity of the vehicle's outer surface to solar radiation. The solar radiation intensity is multiplied by the absorptivity, then by the cosine of the angle of sunlight incident, and finally by the effective area exposed to sunlight to calculate the solar radiation heat flux absorbed by the vehicle's surface per unit time.

[0075] Finally, because atmospheric density changes (affecting convective heat dissipation) and solar radiation (increasing heat load) contribute to the total heat load of the carriage differently in terms of their contribution mechanisms and dimensions, the atmospheric density change factor is converted into its impact on the basic heat exchange coefficient based on pre-calibrated weighting coefficients (which can be determined through historical data). Specifically, an experimentally determined correlation is used, for example... ,in, It is the actual convective heat transfer coefficient at the current atmospheric density at the current altitude. The reference convective heat transfer coefficient is based on the atmospheric density at sea level, from which the influence of the convective heat transfer coefficient can be calculated. ( -1), This is the value of the influence on the basic heat exchange coefficient, converted from the atmospheric density variation factor, and its unit is W / (m³). 2 ·K), which can be directly used for heat balance calculations.

[0076] The environmental heat load change is obtained by weighting and summing the influencing value with the solar radiation heat flux in units of power. This environmental heat load change quantifies the expected change in net heat load on the carriage caused by the external environment in the future period (positive value is an increase in heat load, and negative value is a decrease in heat load).

[0077] By transforming future geographical location information into an environmental heat load change with clear physical meaning, an external disturbance input is provided for step S26 to dynamically adjust the temperature threshold. This enables the adaptive alarm system to detect and prepare to deal with the thermal challenges caused by changes in altitude and sunlight on the preset driving path in advance, realizing the transformation from passively responding to the environment to actively adapting to the environment.

[0078] It should be noted that the effective area exposed to sunlight is the equivalent area used to calculate the solar radiation heat flux absorbed by the surface of the carriage. It is equal to the projected area of ​​the exposed surface of the carriage in the direction perpendicular to sunlight multiplied by the solar radiation absorptivity of that surface.

[0079] Step S23: The nominal temperature permissible range is obtained from the preset cargo temperature control requirements, and the first temperature coefficient corresponding to the current stage is determined based on the transportation stage information.

[0080] The nominal temperature permissible range is obtained by parsing the preset cargo temperature control requirements. These requirements are usually pre-existing in digital form in the RFID tag information associated with the cargo or in the data order of this transportation task. The content clearly specifies the permissible storage temperature range for this batch of cargo. By reading and parsing this data field, the upper and lower limits of the temperature range can be directly extracted to form the nominal temperature permissible range.

[0081] It should also be noted that the transportation stage information is determined by a combination of factors, including vehicle location, task progress logic, and external instructions. For example, it is divided into different stages such as "loading at the station," "mainline transportation," "waiting at the intersection," and "final unloading." The adaptive alarm system has a pre-built mapping table that defines the allowable degree of relaxation relative to the nominal temperature permissible range for each transportation stage. This relaxation is quantified as a coefficient, namely the first temperature coefficient. For example, in the "mainline transportation" stage, this coefficient is 1.0 (indicating strict adherence to the nominal temperature permissible range), while in the "brief unloading" stage, the first temperature coefficient is 1.2 (indicating a 20% relaxation of the allowable boundary). By querying the current transportation stage, the corresponding first temperature coefficient can be obtained from the mapping table.

[0082] This step establishes an initial decision-making framework for calculating temperature thresholds, with cargo safety requirements as the rigid core and the actual operability of the transportation process as the flexible adjustment basis. This lays the rule foundation for subsequent temperature management to not only safeguard the bottom line of cargo quality (the nominal temperature permissible range) but also to achieve intelligent and humanized temperature control based on the actual situation of transportation operations (by providing reasonable flexibility through the first temperature coefficient).

[0083] Step S24: Based on the historical temperature sequence, calculate the cumulative duration and deviation of the current transport batch of goods that are not within the nominal temperature permissible range, and calculate the temperature compensation coefficient based on the cumulative duration and deviation.

[0084] Obtain the historical temperature sequence of the current transportation batch since the start of the mission. This historical temperature sequence is composed of temperature readings in the spatiotemporal reference collaborative data generated in step S10 arranged in chronological order.

[0085] When calculating the cumulative duration of time outside the nominal temperature permissible range, the adaptive alarm system traverses each historical temperature reading in the historical temperature sequence, compares each historical temperature reading with the upper and lower limits of the nominal temperature permissible range obtained in the previous step, identifies all historical abnormal temperature readings that exceed the nominal temperature permissible range, calculates the time span covered from the first historical abnormal temperature reading to the last historical abnormal temperature reading based on the timestamp corresponding to the historical abnormal temperature reading, and accumulates the duration of all continuous or intermittent historical abnormal temperature readings within this time span to obtain a cumulative duration representing the total length of time the goods were exposed to non-permissible temperatures.

[0086] Simultaneously, for each identified historical abnormal temperature reading, the absolute difference between it and the nearest boundary of the nominal temperature permissible range is calculated. For example, for a historical abnormal temperature reading higher than the upper limit of the nominal temperature permissible range, the difference between it and the upper limit is calculated. Then, from the differences corresponding to all historical abnormal temperature readings, the absolute value of the largest difference is selected as the deviation amplitude. Next, a compensation coefficient calculation function is preset. This compensation coefficient calculation function takes the cumulative duration and deviation amplitude as input variables. Its internal logic is: the longer the cumulative duration and the larger the deviation amplitude, the greater the cumulative quality damage, the smaller the subsequent allowable temperature fluctuation elasticity, and the stricter the control. That is, the compensation coefficient calculation function outputs a monotonically decreasing temperature compensation coefficient (for example, decreasing downward from the base value of 1.0) as the cumulative duration and deviation amplitude increase. This enables the adaptive alarm system to dynamically assess the historical quality damage of goods and proactively adjust the future control strictness accordingly. The temperature compensation coefficient is introduced as a feedback adjustment into the temperature threshold decision, thereby tightening the subsequent temperature permissible range after the goods have experienced a certain temperature drift, providing stricter protection for potentially more vulnerable goods.

[0087] Step S25: After relaxing the nominal temperature allowable range according to the first temperature coefficient to obtain the stage tolerance range, the stage tolerance range is corrected according to the temperature compensation coefficient to obtain the target temperature range.

[0088] First, obtain the upper limit of the nominal temperature permissible range. and lower limit value And the first temperature coefficient applicable to the current transportation phase. The nominal temperature permissible range is relaxed, wherein the first center point of the nominal temperature permissible range is calculated. and the first half width Using the first temperature coefficient Multiply by the first half width The second half width is obtained after widening. Then, using the second half width and the center point Calculate the new upper and lower limits to obtain the upper limit of the stage tolerance interval. and lower limit This step is mathematically equivalent to proportionally expanding the nominal temperature permissible range symmetrically, using the center of the interval as a reference.

[0089] Next, obtain the temperature compensation coefficient. This is used to correct the stage tolerance interval obtained above, and to obtain the second center point of the stage tolerance interval. Second half width Then, use the temperature compensation coefficient. Multiply by the second half width The new half-width after tightening is obtained Afterwards, based on the new half-width Second center point Calculate the final upper and lower limits, which gives the upper limit of the target temperature range. and lower limit .

[0090] It should be noted that the stage tolerance range provides the necessary operational flexibility based on the transportation stage, while the target temperature range introspectively tightens this flexibility based on the cargo's own historical temperature drift. This allows it to both respect the flexibility requirements of actual operations and provide targeted protection for cargo quality that may become more vulnerable due to historical exposure. It achieves a dynamic and reasonable balance between operational convenience and cargo safety, and provides a benchmark range for adjusting to changes in the external environment in the next step.

[0091] Step S26: Based on the direction and magnitude of the change in ambient heat load, the target temperature range is directionally shifted to obtain the temperature threshold at the current moment.

[0092] When the change in ambient heat load is positive (indicating an increase in heat load), the upper limit of the target temperature range is subtracted by an offset proportional to the magnitude of the change in ambient heat load, while the lower limit remains unchanged. The offset is calculated using a pre-calibrated mapping function that converts the change in ambient heat load into an impact on the temperature threshold. The mapping relationship is established based on models of the thermal inertia of the carriage and the response characteristics of the refrigeration system.

[0093] Conversely, when the change in ambient heat load is negative (indicating a decrease in heat load), the upper limit of the target temperature range is increased by an offset proportional to the magnitude of the change in ambient heat load, and the upper limit is appropriately widened to reduce unnecessary cooling energy consumption, while the lower limit adjustment strategy is determined based on the characteristics of the goods.

[0094] After the aforementioned targeted adjustments, the upper and lower limits of the target temperature range have been updated. This new pair of upper and lower limits together constitute the temperature threshold at the current moment. The temperature threshold used for monitoring and alarms has the ability to dynamically resist or adapt to foreseeable environmental shocks. It can proactively raise control standards (tighten the upper limit) before the arrival of heat threats, or intelligently reduce energy consumption (relax the upper limit) when the environment is favorable. This fundamentally transforms temperature control from a reactive, post-event alarm and correction mechanism to a proactive, pre-event prevention and adaptation mechanism, significantly improving the robustness of the adaptive alarm system in response to environmental changes, the level of cargo safety assurance, and overall energy efficiency.

[0095] In one feasible implementation, refer to Figure 4 As shown, step S30 may specifically include steps S31 to S34: Step S31: Based on the temperature threshold, compare the spatial calibration data in the spatiotemporal reference collaborative data, identify the abnormal monitoring data, extract the location attributes and timestamps corresponding to the abnormal monitoring data, and obtain a set of candidate abnormal temperature points.

[0096] During the comparison, each compensated temperature reading in the spatial calibration data is traversed, and its value is compared with the upper and lower limits of the temperature threshold in real time. If the compensated temperature reading is greater than the upper limit of the temperature threshold or less than the lower limit, the corresponding compensated temperature reading is marked as abnormal monitoring data.

[0097] Upon identifying abnormal monitoring data, the adaptive alarm system immediately extracts the information corresponding to the abnormal monitoring data from the spatiotemporal reference collaborative data. This information is bound during data generation and mainly includes the installation coordinates of the temperature sensor corresponding to the abnormal temperature reading in the carriage, as well as the relative positional relationship and time reference with key components such as the carriage door and evaporator.

[0098] Each identified abnormal monitoring data point, along with its extracted location attributes and timestamp, is packaged into a structured abnormal data point. All abnormal monitoring data within the current period are then collected into a list to form a candidate abnormal temperature point set.

[0099] In this step, by matching the temperature threshold as a control standard with spatial calibration data, not only are all the abnormal monitoring data that need attention quickly located, but more importantly, each abnormal monitoring data is given spatiotemporal coordinates and contextual information. This transforms a simple over-temperature alarm into a set of candidate abnormal temperature points containing "where, when, and how much it exceeds the standard", providing a data foundation for subsequent steps.

[0100] Step S32: Based on the location attributes and timestamps of each abnormal monitoring data in the candidate abnormal temperature point set, clustering and trend analysis are performed on the spatial distribution and time series changes of the spatiotemporal reference collaborative data in the unmanned cold chain vehicle to obtain the temperature abnormality feature vector.

[0101] Based on the location attributes and timestamps carried by each abnormal monitoring data in the candidate abnormal temperature point set, the distribution of abnormal monitoring data in the carriage and its evolution sequence on the time axis are jointly mined.

[0102] Specifically, its function is reflected in two aspects: First, through spatial clustering analysis based on location attributes, it extracts the three-dimensional location coordinates of all abnormal monitoring data points from the candidate abnormal temperature point set. After treating these abnormal monitoring data points as a spatial point set, it employs a density-based clustering algorithm (such as the noisy density-based spatial clustering algorithm DBSCAN). This algorithm uses the Euclidean distance between abnormal monitoring data points as a proximity metric, and presets a neighborhood radius and a minimum number of abnormal monitoring data points as core parameters. It traverses all abnormal monitoring data points, and groups any core abnormal monitoring data point (whose neighborhood contains at least the minimum number of abnormal monitoring data points) and... All anomalous monitoring data points with reachable density are grouped into the same cluster, while anomalous monitoring data points that cannot be grouped into any cluster are considered noise points. This results in different spatial anomalous clusters, revealing whether the anomalies are concentrated in specific areas such as vehicle doors, cargo gaps, or near the evaporator. This helps determine whether the anomalous monitoring data is a local event (such as cold air leakage) or a widespread impact. Secondly, through time-series trend analysis based on timestamps, for each identified spatial anomalous cluster, the data points are sorted chronologically according to the timestamps carried by each anomalous monitoring data point within the cluster, forming a time series. This time series is then fitted and analyzed, typically using the least squares method to perform first-order linear regression, resulting in a result in the form of... The trend equation, where the slope The average rate of change, the intercept As the initial offset, this equation and the average rate of change constitute the temporal trend result characterizing the temporal evolution of the spatial anomaly cluster, thereby quantifying whether the anomaly is a sudden leap, a slow climb or a periodic fluctuation, thus distinguishing between instantaneous disturbances and gradual faults.

[0103] A multidimensional feature array is created for each spatial anomaly cluster. This feature array is filled with numerical features extracted from spatial clustering results, such as centroid coordinates, and numerical features extracted from the temporal trend results of the spatial anomaly cluster, such as average rate of change, in a preset order. All numerical features corresponding to the identified spatial anomaly clusters are concatenated end to end according to the identification order of the spatial anomaly clusters to form an original feature list. Each feature value in the original feature list is normalized, for example, by using minimum-maximum scaling to linearly transform it to the [0, 1] interval, generating a structured temperature anomaly feature vector. By refining and compressing a series of anomaly monitoring data points into a temperature anomaly feature vector that can comprehensively characterize the overall shape and dynamics of the current temperature anomaly, this temperature anomaly feature vector acts like a "digital fingerprint" of the anomaly pattern, providing digital input for the next step of associating and matching the temperature anomaly pattern with multi-source status information such as vehicle doors and generator units, thereby achieving intelligent root cause diagnosis.

[0104] Furthermore, step S32 also includes steps S321 to S323: Step S321: Based on location attributes, perform spatial distance-based clustering on each anomaly monitoring data to obtain spatial anomaly clusters. Then, extract the centroid location, spatial range, average temperature within the cluster, and temperature standard deviation of the spatial anomaly clusters to obtain a spatial clustering feature set.

[0105] Based on the location attribute of each anomaly monitoring data point, these spatially discrete anomaly monitoring data points are treated as a set of points. When performing spatial distance-based clustering, a distance threshold-driven clustering algorithm (e.g., DBSCAN density clustering or a hierarchical agglomerative clustering algorithm based on Euclidean distance) is used. The core of this algorithm is that any two anomaly monitoring data points whose installation location coordinates are less than a preset distance value are grouped into the same group. This preset distance value is reasonably set according to the size of the vehicle compartment and the density of sensor deployment, for example, 30 centimeters.

[0106] The algorithm iterates through all anomaly points, aggregating neighboring anomaly monitoring data to form a spatial anomaly cluster. Each spatial anomaly cluster corresponds to a local thermal runaway region. After obtaining the spatial anomaly clusters, spatial features are extracted for each cluster: ① The centroid position is calculated by taking the arithmetic mean of the installation position coordinates corresponding to the anomaly monitoring data within the cluster, thus obtaining the geometric center of the cluster in the carriage space; ② The spatial range is calculated by representing the difference between the maximum and minimum values ​​of all anomaly monitoring data within the cluster on each coordinate axis; ③ The average temperature within the cluster is calculated by taking the arithmetic mean of the temperature readings of all anomaly monitoring data within the cluster; ④ The temperature standard deviation is calculated by statistically analyzing the dispersion of the temperature readings at each point within the cluster relative to the above average value.

[0107] The centroid location, spatial range, average temperature within the cluster, and temperature standard deviation calculated for a spatial anomaly cluster are combined in a fixed order to form a spatial feature description of the corresponding spatial anomaly cluster. The above spatial feature extraction is performed on all currently identified spatial anomaly clusters, and all the obtained spatial feature descriptions are summarized to form spatial clustering features. In this way, scattered and isolated anomaly monitoring data are organized into spatial anomaly clusters with clear physical meaning based on their spatial correlation. Each spatial anomaly cluster is quantitatively characterized in multiple dimensions, thereby transforming it into a spatial clustering feature set that can characterize "where the anomaly monitoring data gathers, how large the range is, how hot the average temperature is, and how the internal differences are" within the carriage.

[0108] Step S322: Based on the timestamp, sort the anomaly monitoring data belonging to the same spatial anomaly cluster according to their timestamps, and perform linear and nonlinear trend fitting processing based on time series to obtain the temperature-time change equation and corresponding rate of change of the spatial anomaly cluster, and generate a time-domain trend feature set.

[0109] From each spatial anomaly cluster, the timestamps attached to all anomaly monitoring data within the cluster are obtained. Then, the anomaly monitoring data are sorted in chronological order from early to late to form an anomaly temperature reading sequence that is arranged chronologically and belongs to the same spatial anomaly cluster.

[0110] Next, the abnormal temperature reading sequence is analyzed, and its timestamps are converted to relative time (e.g., seconds) relative to the starting point of the abnormal temperature reading sequence. With the temperature readings as the dependent variable, the least squares method is used to perform linear trend fitting on the abnormal temperature reading sequence, obtaining a straight line that best represents its overall linear trend. The mathematical expression of this line is a linear temperature-time change equation, i.e. In this temperature-time equation, t is time, T is temperature, and a and b are fitting coefficients. The coefficient a of time t in this temperature-time equation is the rate of change, which represents how fast the temperature changes with time (e.g., how many degrees Celsius it rises per minute).

[0111] Simultaneously, polynomial fitting (such as quadratic) is employed to capture complex patterns such as accelerated and asymptotic changes, yielding the corresponding nonlinear "temperature-time change equation" and its instantaneous rate of change at the latest time point (calculated through differentiation). Specifically, when using polynomial fitting to capture complex patterns, the relative time in the time-sorted sequence of abnormal temperature readings is used... Temperature reading as the independent variable As the dependent variable, a quadratic polynomial (i.e., parabola) equation is fitted using the least squares method. ,in, , , The coefficients obtained from the fitting show that the equation is a nonlinear temperature-time variation equation, and its quadratic coefficients are... Characterizes the acceleration of temperature change (when A value greater than 0 indicates an accelerating upward trend. A value less than 0 indicates a decelerating upward or accelerating downward trend, thus enabling the description of complex patterns such as whether temperature anomalies are accelerating, easing, or exhibiting an inflection point of first rising and then falling; to obtain the instantaneous rate of change at the latest time point, the quadratic polynomial equation is analyzed with respect to time. The first derivative yields the rate of change function. Then the latest time point Substituting into this derivative formula, the instantaneous rate of change at that moment can be calculated. This value represents the instantaneous rate of change of temperature at the end of the anomalous time series.

[0112] Then, from the linear and nonlinear temperature-time change equations, select the equation that best represents the temporal change law of the corresponding spatial anomaly cluster based on the goodness of fit (such as R-squared value) and clearly record its equation form and key parameters (such as fitting coefficients), and record the rate of change corresponding to the equation.

[0113] Finally, the above sorting, fitting, equation extraction, and rate of change operations are sequentially performed on all current spatial anomaly clusters. The temporal trend features such as equation type, key parameters, and rate of change corresponding to each spatial anomaly cluster are summarized in the order of spatial anomaly cluster number to form a temporal trend feature set. This set quantifies the temperature anomaly behavior of each spatial anomaly cluster from the perspective of dynamic evolution, thereby expanding the spatial static snapshot of the anomaly into a dynamic profile that includes the temporal evolution law. This provides a temporal dimension quantitative basis for subsequent judgment of the urgency of the fault and differentiation between instantaneous interference and continuous failure.

[0114] Step S323: The spatial clustering feature set and the temporal trend feature set are associated and fused according to the corresponding spatial anomaly clusters, and then arranged and normalized according to the preset feature dimension order to generate a temperature anomaly feature vector.

[0115] After extracting the spatial features of the spatial anomaly cluster numbered 1 from the spatial cluster feature set, and extracting the temporal trend features corresponding to the spatial anomaly cluster with the same number from the temporal trend feature set, these two sets of features from the same cluster are concatenated together in the order of front and back to form a feature sub-vector representing the complete "spatial-temporal" attribute of the corresponding spatial anomaly cluster.

[0116] Repeat the above operation for all identified spatial anomaly clusters. Then, concatenate the feature sub-vectors generated by each spatial anomaly cluster according to the spatial anomaly cluster number to form the original feature sequence. Arrange the original feature sequence according to the preset feature dimension order. The preset feature dimension training is a predefined global feature dimension list, which specifies the fixed position of each feature in the final vector from "centroid X coordinate", "centroid Y coordinate", "centroid Z coordinate" to "temperature standard deviation", and then to "equation type" and "rate of change". According to the preset feature dimension order, each item in the original feature sequence is rearranged to ensure that the feature sub-vectors generated at different times and under different anomaly conditions have a consistent feature dimension order, so that different feature sub-vectors can be compared and calculated.

[0117] Since different features have different units and numerical ranges (e.g., location coordinates are in meters, temperature is in degrees Celsius, and rate of change is in degrees Celsius per second), the pre-stored normalization parameters for each feature dimension (such as the historical maximum and minimum values ​​of the corresponding feature) are called at this time, and the minimum-maximum scaling method is used to linearly transform the value of each feature dimension to a unified range.

[0118] After the above correlation, sorting and normalization operations, a temperature anomaly feature vector with ordered dimensions and uniform numerical range is finally output. This series of processing integrates and transforms the feature sets obtained from the two independent perspectives of space and time in steps S321 and S322 into a temperature anomaly feature vector. This temperature anomaly feature vector is the mathematical digital expression of the current temperature field anomaly pattern of the entire carriage.

[0119] Step S33: The temperature anomaly feature vector is matched with the door status and refrigeration unit status in the spatiotemporal reference collaborative data. Multi-source information matching processing is performed under the time reference in the spatiotemporal reference collaborative data to obtain the state events that are correlated with the current temperature anomaly.

[0120] While acquiring the temperature anomaly feature vector, the candidate anomaly temperature point set on which the corresponding temperature anomaly feature vector is based is extracted from the spatiotemporal reference collaborative data. This set is within the same time interval and is strictly aligned with the door state (such as the sequence of events of door opening and closing) and the refrigeration unit state (such as the time sequence of parameters such as compressor start-up and shutdown, fan speed, and operating current). Multi-source information matching processing is then performed. The core of this process is to perform spatiotemporal correlation calculation on the evolution of the temperature anomaly pattern and the changes of other states on a unified time reference.

[0121] Specifically, the door status and refrigeration unit status are first converted into feature vectors within a time interval. For example, the door status can be converted into features such as "number of times the door is opened during abnormal periods" and "duration of the most recent door opening"; the refrigeration unit status can be converted into features such as "average current", "current fluctuation variance", and "difference between set temperature and average return air temperature", forming a set of status feature vectors.

[0122] Then, the temperature anomaly feature vector and the various state feature vectors extracted and constructed from the same time window are correlated and quantified. The core is to calculate the similarity or correlation between two vectors in the multivariate feature space. Cosine similarity is usually used as the metric. This calculation takes the cosine similarity between two multidimensional vectors as the correlation degree. Its value range is [-1, 1]. The closer the value is to 1, the more consistent the direction of the two vectors are and the more similar the pattern is. It is used to quantify the degree of matching between temperature anomalies and each state change.

[0123] By setting a correlation threshold on the adaptive alarm system, state changes with a correlation exceeding the correlation threshold are filtered out. The filtered state events are determined to be correlated with the current temperature anomaly. For example, the left rear door opening event that occurs within 30 seconds before the temperature rises rapidly, and the compressor operating current event that is consistently lower than the rated value during the continuous temperature rise.

[0124] Breaking through the limitations of traditional methods that rely solely on human experience or simple temporal proximity to determine causal relationships, this method uses quantitative feature matching to identify the state events that are statistically and pattern-wise most relevant to the current temperature anomaly and most likely to constitute a causal relationship from numerous concurrent state changes. This provides a direct, reliable, and interpretable chain of intermediate evidence for the next step of intelligent root cause diagnosis, greatly improving the accuracy and efficiency of fault tracing.

[0125] Step S34: Based on the preset fault mode knowledge base, classify and diagnose the state events to obtain the root cause of the current temperature anomaly and the corresponding confidence level. The analysis results include the root cause and the confidence level.

[0126] Access the preset fault mode knowledge base, which is a structured database that pre-stores various typical fault modes defined by domain expert knowledge and historical case data. Each fault mode clearly defines its corresponding typical pattern of state event combination and temperature anomaly feature vector with causal relationship. For example, a record in the preset fault mode knowledge base defines that if the state event is a door opening event and the temperature anomaly feature vector shows a rapidly rising spatial anomaly cluster located near the door, the root cause may be that the door is not sealed properly or the abnormal opening caused the cold air to leak.

[0127] When performing classification and diagnostic identification, the identified state events and the currently calculated temperature anomaly feature vector are used as a whole input. The system searches and matches the system in a preset fault mode knowledge base. The whole input is compared with each fault mode in the preset fault mode knowledge base. When the key features in the whole input, such as event type, spatial location relationship, and time domain change trend, are highly consistent with the description of a certain fault mode, it is determined that the fault mode has been matched.

[0128] Next, it should be noted that the root cause is the specific fault described by the matched fault mode (such as "evaporator fan stall"), and the confidence level is obtained through a quantitative calculation. The basis includes the matching degree between the overall input and the fault mode (such as the proportion of matched features), the accuracy statistics of the same diagnosis in history, etc. Based on these factors, a confidence level (such as "high", "medium" or "low") representing the reliability of the diagnosis is calculated.

[0129] The above operations have achieved a leap from raw data to intelligent cognition, so that the output is no longer a simple high temperature alarm, but a clear and actionable diagnostic conclusion ("what equipment or link has a problem") and a reliability measure of the conclusion. This allows subsequent "self-adjustment actions" to be initiated based on high-confidence diagnostic results. At the same time, the generated alarm information can directly guide maintenance personnel to take the most effective countermeasures, greatly improving the decision-making intelligence and operation and maintenance efficiency of the entire system.

[0130] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for automatic temperature over-limit alarm in unmanned cold chain vehicles equipped with Beidou positioning, characterized in that, include: Based on the spatiotemporal information provided by BeiDou positioning, spatiotemporal synchronous fusion of multi-source monitoring data of unmanned cold chain vehicles is performed to obtain spatiotemporal reference collaborative data. Based on the location information in the spatiotemporal reference collaborative data, the environmental heat load change of the unmanned cold chain vehicle is predicted. Then, combined with the preset cargo temperature control requirements and transportation stage information, the temperature permissible range is calculated to obtain the temperature threshold at the current moment. After performing correlation analysis based on the spatiotemporal reference collaborative data and the temperature threshold, the corresponding self-adjustment action is triggered and alarm information is generated according to the analysis results.

2. The method for automatic temperature over-limit alarm of unmanned cold chain vehicle with Beidou positioning according to claim 1, characterized in that, The spatiotemporal information includes a time reference and geographic location coordinates. The step of spatiotemporally and synchronously fusing multi-source monitoring data of unmanned cold chain vehicles based on the spatiotemporal information provided by BeiDou positioning to obtain spatiotemporal reference collaborative data includes: Based on the time reference and geographical coordinates provided by the BeiDou positioning, the multi-source monitoring data of multiple sensors deployed inside the unmanned cold chain vehicle are time-aligned and position-calibrated, resulting in synchronized multi-source monitoring data marked with time and position. After obtaining the real-time altitude information corresponding to the geographical coordinates and obtaining the three-dimensional motion state of the unmanned cold chain vehicle based on the time reference, and combining the vehicle body structure of the unmanned cold chain vehicle, motion compensation and altitude compensation are performed on the synchronous multi-source monitoring data to obtain spatial calibration data. Based on the time reference and the geographical coordinates, the spatial calibration data, door status and refrigeration unit status are recombined and packaged to obtain the spatiotemporal reference collaborative data.

3. The method for automatic temperature adjustment and alarm of unmanned cold chain vehicles with Beidou positioning according to claim 2, characterized in that, The step of performing motion compensation and altitude compensation on the synchronous multi-source monitoring data to obtain spatial calibration data includes: Based on the three-dimensional linear acceleration and three-dimensional angular velocity in the three-dimensional motion state, and according to the installation position coordinates of each sensor, the theoretical inertial overload vector of each sensor is calculated. The theoretical inertial overload vector, the refrigerant flow direction in the carriage, and the cargo loading distribution diagram in the carriage are vectorically superimposed to obtain the motion influence of the temperature reading; Based on the real-time altitude information, the absolute value of atmospheric pressure and the pressure difference per unit time are calculated, and based on the installation location coordinates, the distribution of air pressure difference inside and outside the vehicle compartment and the gradient of air pressure change inside and outside the vehicle compartment are calculated. Based on the pressure difference distribution inside and outside the carriage and the pressure change gradient inside and outside the carriage, the predicted deformation of the carriage wall panel and the correction of the equivalent thermal resistance of the carriage insulation layer are coupled and calculated to obtain the influence of altitude on temperature readings. Based on the motion influence of the temperature readings and the altitude influence of the temperature readings, vector synthesis and compensation correction are performed on the synchronous multi-source monitoring data to obtain the spatial calibration data.

4. The method for automatic temperature over-limit alarm of unmanned cold chain vehicle with Beidou positioning according to claim 1, characterized in that, The step of predicting the environmental heat load change of the unmanned cold chain vehicle based on the location information in the spatiotemporal reference collaborative data includes: Based on the geographic location coordinates in the spatiotemporal reference collaborative data and the obtained historical geographic location coordinates, the driving trajectory is extrapolated by combining the preset driving path to obtain the predicted path and vehicle position in the future time period. After obtaining the future altitude and solar radiation intensity corresponding to the predicted location of the vehicle, the atmospheric density change factor is calculated based on the future altitude, and the solar radiation heat flux is calculated based on the future solar radiation intensity. The atmospheric density change factor and the solar radiation heat flux are then weighted and fused to obtain the change in environmental heat load.

5. The method for automatic temperature adjustment alarm of unmanned cold chain vehicle with Beidou positioning according to claim 4, characterized in that, The step of calculating the temperature permissible range by combining preset cargo temperature control requirements and transportation stage information to obtain the temperature threshold at the current moment includes: The nominal temperature permissible range is obtained from the preset cargo temperature control requirements, and the first temperature coefficient corresponding to the current stage is determined based on the transportation stage information; Based on historical temperature sequences, the cumulative duration and deviation of the current transport batch of goods that are not within the nominal temperature permissible range are calculated, and a temperature compensation coefficient is calculated based on the cumulative duration and the deviation. After relaxing the nominal temperature allowable range according to the first temperature coefficient to obtain the stage tolerance range, the stage tolerance range is then corrected according to the temperature compensation coefficient to obtain the target temperature range. Based on the direction and magnitude of the change in the environmental heat load, the target temperature range is directionally shifted to obtain the temperature threshold at the current moment.

6. The method for automatic temperature adjustment and alarm of unmanned cold chain vehicles with Beidou positioning according to claim 1, characterized in that, The step of performing correlation analysis based on the spatiotemporal reference collaborative data and the temperature threshold includes: Based on the temperature threshold, the spatial calibration data in the spatiotemporal reference collaborative data is compared, and after identifying abnormal monitoring data, the location attributes and timestamps corresponding to the abnormal monitoring data are extracted to obtain a candidate abnormal temperature point set. Based on the location attributes and timestamps of each abnormal monitoring data in the candidate abnormal temperature point set, clustering and trend analysis are performed on the spatial distribution and time series changes of the spatiotemporal reference collaborative data in the unmanned cold chain vehicle to obtain a temperature abnormality feature vector. The temperature anomaly feature vector is matched with the door status and refrigeration unit status in the spatiotemporal reference collaborative data, and multi-source information matching is performed under the time reference in the spatiotemporal reference collaborative data to obtain state events that are correlated with the current temperature anomaly. Based on a preset fault mode knowledge base, the state events are categorized and diagnosed to obtain the root cause of the current temperature anomaly and the corresponding confidence level. The analysis results include the root cause and the confidence level.

7. The method for automatic temperature adjustment alarm of unmanned cold chain vehicle with Beidou positioning according to claim 6, characterized in that, The step of clustering and trend analysis of the spatial distribution and time series changes of the spatiotemporal reference collaborative data within the unmanned cold chain vehicle based on the location attributes and timestamps of each anomaly monitoring data in the candidate anomaly temperature point set to obtain the temperature anomaly feature vector includes: Based on the location attributes, the abnormal monitoring data are clustered based on spatial distance to obtain spatial abnormal clusters. Then, the centroid location, spatial range, average temperature within the cluster, and temperature standard deviation of the spatial abnormal clusters are extracted to obtain a spatial clustering feature set. Based on the timestamp, the anomaly monitoring data belonging to the same spatial anomaly cluster are sorted according to their timestamps, and linear and nonlinear trend fitting processing based on time series is performed to obtain the temperature-time change equation and the corresponding rate of change of the spatial anomaly cluster, and generate a time-domain trend feature set. The spatial clustering feature set and the temporal trend feature set are associated and fused according to the corresponding spatial anomaly clusters, and then arranged and normalized according to a preset feature dimension order to generate the temperature anomaly feature vector.

Images