A risk perception and identification system for key indicators of a cold chain system

Abstract

This invention relates to the field of risk perception and identification technology, specifically a risk perception and identification system for key indicators of cold chain systems. It includes collecting multi-point environmental data inside refrigerated trucks, extracting disturbance features and path trends, simultaneously analyzing the relationship between temperature, humidity, and wind changes, identifying boundary drift and trajectory overlap, labeling risk path levels, and outputting cold chain transportation risk identification results. This invention improves disturbance identification accuracy and structural analysis capabilities by introducing maximum difference node location, wind speed jump extraction, and humidity gradient segment analysis. It achieves collaborative identification by combining path unidirectionality and synchronous change relationships, constructs a drift identification mechanism through fluctuation ratio and boundary crossing judgment, establishes a concern level system based on trajectory overlap and directional consistency, and forms a classification structure based on continuity and coverage breadth, thereby enhancing the accuracy of path risk identification and trend judgment capabilities.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of risk perception identification, and particularly relates to a cold chain system key indicator risk perception identification system. BACKGROUND

[0002] The technical field of risk perception identification includes monitoring, analyzing and identifying potential risks, aiming to discover potential risks in systems or environments in a timely manner through real-time monitoring and data analysis of key indicators. Common applications include risk assessment and early warning in logistics, transportation, production processes and other fields. Its core content usually involves data collection, processing and analysis, which can identify and feedback potential risks, and thus provide support for related decision-making. This technical field is widely used in cold chain logistics, financial risk management, health monitoring and other industries, and has important practical value.

[0003] Among them, the cold chain system key indicator risk perception identification system refers to the real-time identification of risk factors in the cold chain process through the collection and analysis of key data in the cold chain logistics. It includes the monitoring of environmental conditions such as temperature and humidity in the cold chain transportation process, combined with transportation time, equipment operating status and other information, to identify potential risks in the cold chain system. Specifically, by setting a risk identification model, the key indicators are calculated and compared based on real-time data to analyze whether there are abnormal conditions and judge the risk level. This process usually relies on continuous monitoring of cold chain environmental parameters and data processing technology to realize timely identification of system operation status and risk warning.

[0004] The existing technology mainly relies on static threshold setting and single indicator trend analysis for abnormal identification of cold chain key parameters, lacks dynamic combination judgment method based on disturbance characteristic evolution path, and in actual transportation process, it is difficult to establish effective linkage judgment logic for multi-parameter synchronous fluctuation phenomenon, unable to form a trend screening mechanism for complex disturbance trend, also without clear division of path fluctuation attention level and risk distribution level, causing difficulty in timely integrated identification of multi-point joint risk, when risk continues to spread in the interference frequent area, the feedback ability lags behind, affecting the comprehensive capture ability of the monitoring system to the risk state. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and to provide a cold chain system key indicator risk perception identification system.

[0006] In order to achieve the above purpose, the present application adopts the following technical scheme: a cold chain system key indicator risk perception identification system, the system comprises:

[0007] An environmental disturbance capturing module acquires temperature, humidity and wind data of a middle section of a refrigerated truck rack, an air duct opening and a tail stacking area, extracts temperature section difference nodes, screens humidity curve turning sections, extracts wind speed jump trajectories, compares fluctuation amplitudes in each type of path and summarizes fluctuation ranges, and generates a multi-point disturbance feature set;

[0008] An index trend screening module constructs wind speed paths according to isotropy within a period and matches their variation directions according to temperature and humidity paths in the multi-point disturbance feature set, extracts path combinations consistent in direction within the same time period, screens data satisfying each variation direction synchronization, and obtains a synchronous trend variation structure;

[0009] A limit drift identification module constructs temperature and humidity change ratio trajectories based on the synchronous trend variation structure, calls delayed paths of the cargo stacking area, judges whether the ratio appears across the border within the original boundary, identifies offset sections and marks limit adjustment positions, and outputs a drift trigger boundary structure;

[0010] A degree of attention distribution module constructs an overlap mapping diagram according to offset paths in the drift trigger boundary structure, calls surface response data, divides level labels according to the overlap of temperature and humidity fluctuations and surface reaction directions, and generates a change attention inclination distribution diagram.

[0011] As a further scheme of the present application, the multi-point disturbance feature set includes temperature range node distribution, humidity fluctuation gradient features and wind speed frequency jump sections, the synchronous trend variation structure includes path direction consistent sections, time overlap combination relationships and three types of synchronous trend patterns, the drift trigger boundary structure includes fluctuation amplitude ratio intervals, boundary offset periods and response path drift features, and the change attention inclination distribution diagram includes trajectory overlap continuous sections, direction consistency levels and high-level fluctuation path distribution.

[0012] As a further scheme of the present application, the environmental disturbance capturing module includes a multi-point data acquisition sub-module, a sequence variation extraction sub-module and a fluctuation path integration sub-module.

[0013] A multi-point data acquisition sub-module acquires temperature, humidity and wind speed time series data of a middle section of a refrigerated truck rack, an air duct opening position and a tail stacking coverage area, records time labels, numerical outputs and data densities within consecutive periods of each type of data at each collection point, judges sampling interval consistency of the collected data, eliminates abnormal null values and fills in missing sections, and generates a standardized continuous data set.

[0014] The sequence variation extraction submodule constructs a difference value track of each section based on a temperature sequence in the standardized continuous data set, screens a variation node with a maximum difference value and extracts a continuous section, positions a turning point of a humidity sequence and screens a continuous change time period, simultaneously calculates a gradient intensity in the change section, extracts a jump section of wind speed data with a sliding window and counts an occurrence frequency, and obtains a fluctuation track set of sub-items;

[0015] The fluctuation path integration submodule calls a continuous range of a corresponding time period in each sequence according to the difference value section, the turning path and the jump track extracted in the fluctuation track set of sub-items, screens an overlapping interval in which change behaviors exist in temperature, humidity and wind, marks the interval as a fluctuation concentrated section, counts a joint continuous span of each type of change, and generates a multi-point disturbance feature set.

[0016] As a further scheme of the application, the index trend screening module comprises a direction consistent identification submodule, a wind speed trend superposition submodule and a time section screening submodule.

[0017] The direction consistent identification submodule obtains a temperature path and a humidity path marked in the multi-point disturbance feature set, detects the relative positions of trend directions in respective periods based on the change direction sequences of the two paths, performs path pairing on a temperature and humidity combination with consistent directions in a time section, screens combinations according to a trend consistent proportion, and generates a temperature and humidity trend matching group.

[0018] The wind speed trend superposition submodule calls a combined path in the temperature and humidity trend matching group, collects a wind speed path change sequence in a corresponding time period, performs time section corresponding registration according to the fluctuation direction of the wind speed path and the direction sequence of the temperature and humidity combination, extracts a wind speed path with consistent fluctuation direction and time range as the temperature and humidity combination, and obtains three direction consistent structures.

[0019] The time section screening submodule judges whether the three types of paths have completely overlapping time sections in a continuous time period according to the time labels of the paths in the three direction consistent structures, performs section consistency judgment on the duration length and start and end time of the overlapping section, marks a combined path that satisfies a full synchronous trend and time consistency, and generates a synchronous trend variation structure.

[0020] As a further scheme of the application, the limit drift recognition module comprises an amplitude ratio construction submodule, a boundary out-of-bound judgment submodule and a drift section extraction submodule.

[0021] The amplitude ratio construction submodule obtains maximum and minimum values of each path in a same time period based on the temperature and humidity combined paths contained in the synchronous trend variation structure, constructs a fluctuation amplitude ratio according to the value difference and a reference length, generates a graph sequence arranged according to corresponding time periods, and generates a time period amplitude ratio graph.

[0022] The boundary overrun judgment submodule calls each ratio section in the time-amplitude ratio atlas, locates the upper and lower boundary sections corresponding to each time section according to the response delay path recorded by each monitoring point in the cargo stacking area, judges whether each ratio section crosses the upper and lower boundary sections, and obtains a boundary overrun interval section;

[0023] The drift section extraction submodule extracts the path offset trend under the corresponding time slice according to the overrun time section marked in the boundary overrun interval section, identifies the start and end change range of the original boundary position in the time section, compares the difference between the upper and lower limit paths before and after the change and makes offset amount marking, and generates a drift trigger boundary structure.

[0024] As a further scheme of the present application, the attention degree distribution module comprises a trajectory coincidence extraction submodule, a time direction label division submodule and an attention level confirmation submodule.

[0025] The trajectory coincidence extraction submodule calls the cargo surface response change data in the corresponding time section based on the temperature and humidity combined path identified as offset in the drift trigger boundary structure, obtains the coincidence section of the two types of paths and matches the start and end times of each other, constructs a section coincidence duration mapping diagram according to the time section intersection, and generates a response overlap duration atlas.

[0026] The time direction label division submodule calculates the direction consistent proportion value of the two paths in the coincidence section according to the time intersection length corresponding to each overlap section in the response overlap duration atlas, compares the time length and the direction proportion with a unified interval threshold respectively, sets the corresponding label level according to the value ratio level, and generates an overlap label level division result.

[0027] The attention level confirmation submodule calls the high-level label path information in the overlap label level division result, binds this part of the path with the original temperature and humidity combined sequence, extracts the offset frequency value and surface change intensity value of each bound path, constructs an attention distribution diagram in the inclined direction of the corresponding position in the time section space, and obtains a change attention inclined distribution diagram.

[0028] As a further scheme of the present application, the system further comprises:

[0029] The risk state output module calls the attention path in the change attention inclined distribution diagram, counts the frequency and distribution of the path in the section, filters the path with long change duration and large fluctuation area coverage range, and outputs a risk awareness recognition result in the cold chain transportation process.

[0030] The risk awareness recognition result comprises a high-frequency fluctuation path set, a continuous continuation section distribution and a coverage area range statistics.

[0031] As a further scheme of the present application, the risk state output module comprises an attention path identification submodule, a path section statistics submodule, and a risk trend classification submodule.

[0032] The attention path identification submodule calls the fluctuation path marked as an attention item in the change attention inclination distribution map, extracts the time period, number, and involved transportation section sequence corresponding to each path, records the start and end time positions of the path in each section, arranges and constructs a path time span table according to the path number, and generates a path section distribution table.

[0033] The path section statistics submodule extracts the appearance frequency, time span ratio, and distribution area number of each path in different sections according to the distribution of each path in the path section distribution table in the continuous transportation section, filters the paths according to the appearance number threshold and distribution range threshold, and obtains a covered distribution path set.

[0034] The risk trend classification submodule calls the number, time span, and fluctuation direction of the path in each section in the covered distribution path set, extracts attribute similar paths according to time continuity and regional consistency, merges and classifies the paths according to the time sequence consistency and section intersection number, and generates a risk sense identification result.

[0035] Compared with the prior art, the present application has the following advantages and positive effects:

[0036] In the present application, by introducing the maximum difference node positioning, wind speed jump path extraction, and humidity gradient section analysis method, the time accuracy and structural integrity of disturbance capture are enhanced, the path trend homodromy screening and multi-path synchronous change matching are used, the collaborative identification of environmental disturbance trend is realized, the fluctuation amplitude ratio and boundary crossing path are combined to construct a response drift identification mechanism, the trajectory overlap mapping and direction consistency hierarchical labeling are used to establish an attention level system, and further relying on the continuity and distribution breadth characteristics, a multi-dimensional classification structure is formed, and the resolution ability, trend judgment accuracy, and path evaluation depth of the fluctuation path risk identification are improved. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The system flowchart of the present application is shown in the figure;

[0038] Figure 2 The acquisition flowchart of the environmental disturbance capture module of the present application is shown in the figure;

[0039] Figure 3 The acquisition flowchart of the index trend screening module of the present application is shown in the figure;

[0040] Figure 4 The acquisition flowchart of the limit drift identification module of the present application is shown in the figure;

[0041] Figure 5 Flowchart for obtaining the degree distribution module of the application;

[0042] Figure 6 Flowchart for obtaining the risk state output module of the application. DETAILED DESCRIPTION

[0043] The technical solutions in the application will be described below with reference to the drawings.

[0044] In the embodiments of the application, the words such as “example”, “for example” are used to represent an example, illustration or description. Any embodiment or design scheme described as “example” in the application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the word “example” is intended to present the concept in a specific manner. In addition, in the embodiments of the application, the meaning expressed by “and / or” can be both, or can be one of the two.

[0045] In the embodiments of the application, “image” and “picture” can be used interchangeably at times. It should be pointed out that when the distinction is not emphasized, the meanings expressed are consistent. “Of”, “corresponding” and “corresponding” can be used interchangeably at times. It should be pointed out that when the distinction is not emphasized, the meanings expressed are consistent.

[0046] In the embodiments of the application, sometimes the subscript such as W1 can be written in the form of non-subscript such as W1. When the distinction is not emphasized, the meanings expressed are consistent.

[0047] To make the technical problems, technical solutions and advantages to be solved by the application clearer, specific embodiments will be described in detail below with reference to the drawings.

[0048] Please refer to Figure 1 The application provides a technical solution: a cold chain system key indicator risk awareness identification system, the system comprising:

[0049] An environmental disturbance capturing module acquires temperature, humidity and wind speed time series data of the middle section of the shelf, the air duct opening position and the tail stacking coverage area in the refrigerated vehicle compartment, extracts the maximum difference change node in the temperature sequence and identifies the duration, extracts the fluctuation duration section based on the turning point of the humidity curve and extracts the gradient section, extracts the jump section path according to the fluctuation frequency of the wind speed in the sliding window, summarizes and locates the maximum fluctuation interval in various change sequences, and generates a multi-point disturbance feature set;

[0050] The index trend screening module performs same direction judgment in a period according to the temperature and humidity sequence of the located fluctuation path in the multi-point disturbance feature set, extracts a combination consistent with the trend, constructs the direction trajectory of the wind speed path and the combination in the same time period, judges whether there is a completely synchronous trend change in each type of trajectory in the corresponding section, screens out a data set that meets the time overlap and direction consistency conditions, and obtains a synchronous trend change structure;

[0051] The limit drift identification module constructs a fluctuation amplitude ratio-time period matching graph based on the temperature and humidity combination path contained in the synchronous trend change structure, judges whether each ratio section in the delay path crosses the original upper and lower boundary section, extracts the time period corresponding to the crossing section and performs boundary offset identification, and outputs a drift trigger boundary structure;

[0052] The attention degree distribution module constructs an overlapping section persistence mapping graph of the two trajectories according to the temperature and humidity combination change path identified as offset in the drift trigger boundary structure, performs hierarchical label division according to the time overlap length and direction consistency ratio, records the fluctuation path with a high label level as an attention item, and generates a change attention inclination distribution graph;

[0053] The risk state output module calls the fluctuation path marked as an attention item in the change attention inclination distribution graph, performs time matching processing on the continuation span of each path in the continuous transportation section, performs partition statistics on the frequency and distribution range of the path in each section, classifies and labels the fluctuation path with a continuous occurrence trend and a large coverage area range, and outputs a risk awareness recognition result in the cold chain transportation process.

[0054] The multi-point disturbance feature set includes temperature range node distribution, humidity fluctuation gradient features, and wind speed frequency jump section, the synchronous trend change structure includes path direction consistent section, time overlap combination relationship, and three types of synchronous trend patterns, the drift trigger boundary structure includes fluctuation amplitude ratio interval, boundary offset period, and response path drift features, the change attention inclination distribution graph includes trajectory overlapping persistence section, direction consistency level, and high-level fluctuation path distribution, and the risk awareness recognition result includes high-frequency fluctuation path set, continuous continuation section distribution, and coverage area range statistics.

[0055] Please refer to Figure 2 , the environmental disturbance capturing module includes a multi-point data acquisition submodule, a sequence change extraction submodule, and a fluctuation path integration submodule;

[0056] The multi-point data acquisition sub-module acquires temperature, humidity and wind speed time series data of the middle section of the goods shelf, the air duct opening position and the tail stacking coverage area in the refrigerated vehicle compartment, records the time tag, numerical output and data density in each collection point in a continuous period, judges the sampling interval consistency of the collected data, eliminates abnormal null values and fills in the missing sections, and generates a standardized continuous data set.

[0057] The temperature, humidity and wind speed time series data of the middle section of the goods shelf, the air duct opening position and the tail stacking coverage area in the refrigerated vehicle compartment are acquired. Specifically, Pt100 platinum resistance temperature sensors with an accuracy of ±0.1℃, capacitive humidity sensors with an accuracy of ±2%RH and hot-wire anemometers with an accuracy of ±0.1m / s are respectively deployed at the middle section of the goods shelf (point A), the air duct opening (point B) and the tail stacking coverage area (point C). The data is continuously collected for 900 periods at a fixed sampling interval of 10 seconds. Each data point is associated with the collection point identifier (A, B or C), the data type identifier (T, H or W), the Unix timestamp (accurate to seconds) and the numerical value, forming an original data set. The data density in each 60-second continuous period is calculated, i.e. the actual number of collected points divided by the expected number of collected points (6). If the data density is less than 0.9, the period is marked as unreliable. The sampling interval consistency of all collected data is judged by calculating the difference between adjacent timestamps to verify whether it is 10 seconds. For data points with a difference other than 10 seconds, they are marked and removed from the data set if the value is -999 or exceeds the physical possible range (e.g. humidity exceeds 100%RH). For missing data segments caused by removal or collection failure, linear interpolation is used to fill in the missing data. For example, if the temperature at point A at timestamp 1663905600 is 4.2℃ and the temperature at timestamp 1663905620 is 4.4℃, the missing temperature value at timestamp 1663905610 is calculated and filled in as (4.2+4.4) / 2=4.3℃. This process is repeated for all data points to generate a standardized continuous data set without null values and with a strict 10-second time interval.

[0058] Table 1: Standardized continuous data representation example

[0059] Time stamp Collection point Data type Numerical value 1663905600 A Temperature 4.2℃ 1663905600 B Humidity 85.1% RH 1663905600 C Wind speed 1.5 m / s 1663905610 A Temperature 4.3℃ 1663905610 B Humidity 85.3% RH 1663905610 C Wind speed 1.4 m / s

[0060] As shown in Table 1, some standardized continuous data after processing is listed, showing the recording format of different collection points and data types at continuous time points.

[0061] The sequence variation extraction sub-module constructs difference value trajectories of each section based on temperature sequences in the standardized continuous data set, screens the maximum difference value variation node and extracts the continuous section, locates the turning point of the humidity sequence and screens the continuous change time period, calculates the gradient intensity in the change section, extracts the jump section of the wind speed data by using the sliding window, and obtains the sub-item fluctuation trajectory set;

[0062] The sequence variation extraction sub-module constructs difference value trajectories of each section based on temperature sequences in the standardized continuous data set, screens the maximum difference value variation node and extracts the continuous section, locates the turning point of the humidity sequence and screens the continuous change time period, calculates the gradient intensity in the change section, extracts the jump section of the wind speed data by using the sliding window, and obtains the sub-item fluctuation trajectory set; ​​​​​​​​​​​​​​When the 60-second period is recorded as a jump segment, the frequency of all jump segments in each hour is counted, and the extracted temperature duration section, humidity continuous change period, gradient, wind speed jump segment, and frequency information are integrated to obtain a sub-item fluctuation trajectory set.

[0063] Table 2: Sub-item fluctuation trajectory set example

[0064] Trajectory type Collection point Start time stamp End time stamp Key indicator Temperature continuous section A 1663905800 1663905920 Difference symbol: + Humidity continuous change B 1663905820 1663905950 Gradient intensity: 0.01% RH / s Wind speed jump section C 1663905810 1663905870 Occurrence frequency: 4 times / hour

[0065] Table 2 gives an example of a sub-item fluctuation trajectory set, which contains fluctuation events and their core attributes extracted from different sensor data.

[0066] The fluctuation path integration submodule calls the duration range of the corresponding time period in each sequence according to the difference section, turning path and jump trajectory extracted in the sub-item fluctuation trajectory set, filters out the overlapping interval where there is a change in temperature, humidity and wind, marks the interval as a fluctuation concentrated section, and counts the joint duration span of each type of change to generate a multi-point disturbance feature set;

[0067] According to the difference section, the turning path and the jump trajectory extracted from the sub-wave fluctuation trajectory set, the continuous range of the corresponding time period in each sequence is called, for example, the temperature continuous section of an A point is extracted from the set, the time range of which is [1663905800, 1663905920], a humidity continuous change time period, the time range of which is [1663905820, 1663905950], and a wind speed jump section, the time range of which is [1663905810, 1663905870], the overlapping interval in which the change behaviors exist in temperature, humidity and wind are screened out, and the screening process is realized by calculating the intersection of the three time sections, that is, the start time of the overlapping interval takes the maximum value of the three start times, and the end time takes the minimum value of the three end times. For the above example, the start time of the overlapping interval is max(1663905800, 1663905820, 1663905810) = 1663905820, and the end time is min(1663905920, 1663905950, 1663905870) = 1663905870. If the calculated start time is later than or equal to the end time, there is no overlapping interval, otherwise, the overlapping interval [1663905820, 1663905870] is marked as a fluctuation set section, and the joint continuous span of various changes is counted, that is, the end time of the overlapping interval is subtracted from the start time. In this example, the joint continuous span is 1663905870-1663905820 = 50 seconds. This process is performed for all fluctuation trajectories of all collection points (A, B, C). The fluctuation set sections found and their corresponding collection points, start and end times, joint continuous spans and other information are collected to generate a multi-point disturbance feature set.

[0068] Please refer to Figure 3 , the index trend screening module includes a direction consistent identification submodule, a wind speed trend superposition submodule, and a time section screening submodule.

[0069] The direction consistent identification submodule obtains the temperature path and the humidity path marked in the multi-point disturbance feature set, detects the relative position of the trend direction in each cycle based on the change direction sequence of the two paths, performs path pairing on the temperature and humidity combination with consistent direction in the time section, and screens the combination according to the trend consistent proportion to generate a temperature and humidity trend matching group.

[0070] The labeled temperature path and humidity path in the multi-point disturbance feature set are obtained. Specifically, for a fluctuation segment, such as the temperature and humidity data sequence of point A in the time period [1663905820, 1663905870], based on the change direction sequence of the two paths, the relative position of the trend direction in each period is detected, and the direction label is calculated for each 10-second interval of the two paths. If the current value is greater than the previous value, the label is +1, less than -1, and equal to 0. Thus, two sequences composed of +1, -1, and 0 are obtained, for example, the temperature direction sequence is [+1, +1, +1, -1, +1], and the humidity direction sequence is [+1, +1, -1, -1, +1]. The temperature and humidity combination with the same direction in the time segment is paired, that is, at the same time point, if the direction labels of temperature and humidity are both +1 or both -1, the time point is regarded as a direction consistent point. In the above example, the 1st, 2nd, and 5th time points are direction consistent points. According to the trend consistent proportion, the combination is screened, and the proportion is calculated as the number of direction consistent points divided by the total number of time points in the segment. In the example, the proportion is 3 / 5=0.6, and the trend consistent proportion threshold is set . The threshold is set by analyzing 50 sets of historical transportation data, of which 25 sets have cargo corrosion and 25 sets are normal. Analysis shows that the trend consistent proportion of the corrosion group is generally concentrated above 0.8, while the normal group is below 0.7. To ensure the accuracy of identification, 0.8 is selected as the threshold, that is . If the trend consistent proportion of a combination is less than 0.8, the combination is discarded. If it is greater than or equal to 0.8, the temperature and humidity combination path and related information are retained, and a temperature and humidity trend matching group is generated.

[0071] The wind speed trend superposition submodule calls the combination path in the temperature and humidity trend matching group, collects the wind speed path change sequence in the corresponding time period, and performs time period corresponding registration according to the fluctuation direction of the wind speed path and the direction sequence of the temperature and humidity combination. The wind speed path with the same fluctuation direction and time range as the temperature and humidity combination is extracted, and three direction consistent structures are obtained.

[0072] The combination path in the temperature and humidity trend matching group is called, for example, a temperature and humidity combination with a trend consistent proportion of 0.9 in the time period [1663906200, 1663906300]. The wind speed path change sequence is collected in the corresponding time period, that is, the wind speed data in the 100 seconds is extracted, and its corresponding direction label sequence is calculated, obtaining a wind speed direction sequence composed of +1, -1, and 0. According to the fluctuation direction of the wind speed path and the direction sequence of the temperature and humidity combination, the time period is correspondingly registered. Specifically, in the temperature and humidity combination, the time points with consistent temperature and humidity directions are identified, for example, at , the temperature and humidity directions are both +1, so the common direction at this time is set to +1. In the wind speed direction sequence, the The direction labels of the time points are judged to see if they are equal, for example, if The time point is inconsistent if the combined direction of temperature and humidity is +1 and the wind speed direction is -1, and the time point is consistent if the wind speed direction is also +1. The wind speed path is extracted by repeating the registration process for all time points in the 100 seconds, filtering out all time points where the three direction labels of temperature, humidity, and wind are the same and not 0, and connecting the extracted time points into one or more sub-paths, and integrating the sub-paths with the original temperature and humidity combined path to obtain a three-direction consistent structure.

[0073] The time section screening submodule judges whether the three types of paths have completely overlapping time sections in continuous time periods according to the time labels of the paths in the three-direction consistent structure. The duration and start and end times of the overlapping sections are subjected to section consistency judgment, and the combined path that meets the full synchronization trend and time consistency is marked to generate a synchronization trend change structure.

[0074] The time section screening submodule judges whether the three types of paths have completely overlapping time sections in continuous time periods according to the time labels of the paths in the three-direction consistent structure. The duration and start and end times of the overlapping sections are subjected to section consistency judgment, and the combined path that meets the full synchronization trend and time consistency is marked to generate a synchronization trend change structure. The duration of each overlapping section is calculated, for example, the first section is 40 seconds long and the second is 20 seconds long. A minimum duration threshold is set, which is 30 seconds according to the cargo thermal inertia experiment, which shows that for the specific fruits and vegetables carried, the internal temperature is basically not affected when the external environmental disturbance lasts for less than 30 seconds. Therefore, the threshold is set to 30 seconds. All overlapping sections are screened using the threshold, and sections with a duration of less than 30 seconds are removed. In the above example, the second section with a length of 20 seconds is removed, and only the first section with a length of 40 seconds is retained. The combined path that meets the full synchronization trend and time consistency is marked, i.e., the section [1663906210, 1663906240] retained after screening and the corresponding temperature, humidity, and wind data paths are marked to generate a synchronization trend change structure.

[0075] Please refer to Figure 4The limit drift identification module comprises an amplitude ratio construction submodule, a boundary out-of-limit judgment submodule, and a drift section extraction submodule.

[0076] The amplitude ratio construction submodule obtains the maximum and minimum values of each path in the same time period based on the temperature and humidity combination paths contained in the synchronous trend variation structure, constructs the fluctuation amplitude ratio based on the value difference and the reference length, arranges to generate a spectrum sequence according to the corresponding time period, and generates a time period amplitude ratio spectrum.

[0077] Based on the temperature and humidity combination paths contained in the synchronous trend variation structure, for example, a synchronous path with a time period of [1663906210, 1663906240], the maximum and minimum values of each path in the same time period are obtained. In this 40-second section, the temperature rises from 4.8℃ to 5.3℃, and the humidity rises from 88%RH to 91%RH, therefore, , , , The fluctuation amplitude ratio is constructed based on the value difference and the reference length. The reference length is set to 60 seconds. The calculation method of the fluctuation amplitude ratio is to divide the value difference by the actual duration and multiply it by the reference length, so as to standardize it into the change rate per minute. The fluctuation amplitude ratio of the temperature , and the fluctuation amplitude ratio of the humidity is 0.0005 / minute. According to the corresponding time period, a spectrum sequence is arranged to associate the calculated value of each synchronous variation structure with the corresponding time period to form a time sequence, and a time period amplitude ratio spectrum is generated.

[0078] The boundary out-of-limit judgment submodule calls each ratio section in the time period amplitude ratio spectrum, locates the upper and lower limit sections of the boundary corresponding to each time period according to the response delay path recorded by each monitoring point in the cargo stacking area, judges whether each ratio section crosses the upper and lower boundary sections, and obtains the boundary out-of-limit interval section.

[0079] Each ratio section in the time period amplitude ratio spectrum is called, for example, the ratio section corresponding to the time period [1663906210, 1663906240] mentioned above , according to the response delay path recorded by each monitoring point in the cargo stacking area, the boundary upper and lower limit section corresponding to each time period is located, specifically, the boundary upper and lower limit is preset according to the type of goods, for the batch of transported lettuce, the upper limit of the allowed temperature fluctuation rate is 1.0℃ / min, and the lower limit is-1.0℃ / min, the upper limit of the humidity fluctuation rate is 5.0%RH / min, and the lower limit is-5.0%RH / min, the above boundary value is determined by 10 times of simulation transportation experiment data, the environmental parameters are changed at different rates, and the quality of the goods is monitored to determine the critical change rate that does not cause quality degradation, whether each ratio section crosses the upper and lower boundary section is judged, the calculated fluctuation amplitude ratio is compared with the boundary value, for temperature, , no crossing, for humidity, , also no crossing, if the calculation in another time period is , then because , the temperature ratio section of the time period is judged to cross the upper boundary, the time period will be marked and recorded, and all the time periods marked as crossing the boundary are summarized to obtain the boundary crossing interval section.

[0080] The drift section extraction submodule extracts the path offset trend under the corresponding time slice according to the crossing time period marked in the boundary crossing interval section, identifies the start and end change range of the original boundary position in the time period, compares the difference between the upper and lower limit paths before and after the change and makes offset amount marking, and generates the drift trigger boundary structure;

[0081] According to the crossing time period marked in the boundary crossing interval section, for example, a time period in [1663907100, 1663907150], For the 1.2℃ / min cross-border record, the path offset trend under the corresponding time slice is extracted, that is, the original temperature data sequence within 50 seconds is called, which shows that the temperature linearly rises from 4.6℃ to 5.6℃, and the starting and ending change range of the original boundary position in the time period is identified. The original boundary here refers to the absolute value range of the safe temperature of the goods, not the change rate. For lettuce, the range is set to [1.0℃, 5.0℃]. The experimental verification process of this range is to place 30 lettuce samples in a constant temperature environment from -2.0℃ to 8.0℃ with a step of 0.5℃ for 48 hours, and observe and record the cell structure, water content and signs of decay. The results show that ice crystal damage occurs below 1.0℃, and the reproduction rate of microorganisms significantly increases above 5.0℃. Therefore, this range is determined. Within the above cross-border time period, the temperature path changes from 4.6℃ (within the boundary) to 5.6℃ (outside the boundary), so the starting point of the path offset is the time when the temperature reaches 5.0℃, which is assumed to be 1663907133, and the end point of the offset is the end time 1663907150. Compare the difference between the upper and lower limit paths before and after the change and make offset annotations, that is, calculate the maximum value of the path exceeding the boundary, the offset is 5.6℃-5.0℃=0.6℃, and the offset 0.6℃ and the specific time period [1663907133, 1663907150] of the offset are annotated to generate the drift trigger boundary structure.

[0082] Please refer to Figure 5 The attention degree distribution module includes a trajectory coincidence extraction submodule, a time direction label division submodule, and an attention level confirmation submodule.

[0083] The trajectory coincidence extraction submodule, based on the temperature and humidity combination path identified as offset in the drift trigger boundary structure, calls the cargo surface response change data within the corresponding time period, obtains the coincidence sections of the two types of paths and matches their starting and ending times, constructs a persistent mapping diagram of each section according to the intersection of the time period, and generates a response overlap persistence map.

[0084] Based on the temperature and humidity combination path marked as offset in the drift-triggered boundary structure, such as the temperature path that shifted by 0.6℃ within the aforementioned time period [1663907133, 1663907150], the surface response change data of the goods within the corresponding time period is retrieved. This data is collected by a temperature patch pasted on the surface of the goods packaging box, which shows that the surface temperature rose from 4.8℃ to 5.2℃ within the time period [1663907140, 1663907160]. The overlapping sections of the two types of paths are obtained and their start and end times are matched. The ambient temperature offset path time is [1663907133, 1663907150]. The response path time is [1663907140, 1663907160]. By calculating the time intersection, the overlapping segment is obtained as [max(1663907133, 1663907140), min(1663907150, 1663907160)] = [1663907140, 1663907150]. Based on the time intersection, a persistence mapping map of the overlapping area of ​​each segment is constructed, that is, each drift event is mapped to the duration of the corresponding cargo surface response overlapping segment (10 seconds in this example), forming a list containing all drift events and their response durations, generating a response overlap persistence map.

[0085] The time direction labeling submodule calculates the directional consistency ratio of the two paths in the overlapping area based on the time intersection length of each overlapping segment in the response overlap duration map. It then compares the time length and directional ratio with a unified interval threshold, sets the corresponding label level according to the ratio level, and generates the overlapping label level classification result.

[0086] Based on the time intersection length corresponding to each overlapping segment in the response overlap duration map, such as the aforementioned 10-second overlapping segment, the directional consistency ratio of the two paths within the overlapping area is calculated. Specifically, within the time period [1663907140, 1663907150], the directional sequence of the ambient temperature path (one point every 10 seconds) is [+1], and the directional sequence of the cargo surface temperature path is [+1]. Therefore, the directional consistency ratio is 1 / 1 = 1.0. The time length and directional ratio are compared with a unified interval threshold, where a time length threshold is set here. With direction ratio threshold The threshold was set based on historical data analysis, which statistically analyzed 30 risk events that caused level 2 or higher damage to goods. It was found that the overlap duration of these events all exceeded 8 seconds, and the directional consistency was all higher than 0.9. In contrast, over 95% of drift events that did not cause damage did not meet these two conditions. Therefore, the threshold was set... It lasts for 8 seconds. The value is 0.9. The corresponding label level is set according to the ratio level, with the following rules: if the overlap duration > and the direction ratio If one of the conditions is met, the label is "medium"; if none of the conditions is met, the label is "low". In this example, the overlap duration is 10 seconds > 8 seconds, and the direction ratio is 1.0 > 0.9, so the label level is set to "high". This is performed for all overlapping sections to generate the overlap label level division result.

[0087] The attention level confirmation submodule calls the high-level label path information in the overlap label level division result, binds the part of the path with the original temperature and humidity combination sequence, and extracts the offset frequency value and surface change intensity value of each bound path to construct the attention distribution graph in the corresponding position and tilt direction in the time period space, and obtain the change attention tilt distribution graph.

[0088] The attention level confirmation submodule calls the high-level label path information in the overlap label level division result, binds the part of the path with the original temperature and humidity combination sequence, and extracts the offset frequency value and surface change intensity value of each bound path to construct the attention distribution graph in the corresponding position and tilt direction in the time period space, and obtain the change attention tilt distribution graph.

[0089] Please refer to Figure 6 The risk state output module includes an attention path identification submodule, a path section statistics submodule, and a risk trend classification submodule.

[0090] The attention path identification submodule calls the fluctuation path marked as an attention item in the change attention tilt distribution graph, extracts the time period, number, and involved transportation section sequence corresponding to each path, records the start and end time positions of the path in each section, arranges and constructs a path time span table according to the path number, and generates a path section distribution table.

[0091] The fluctuation path marked as the attention item in the change attention tilt distribution diagram is called, that is, all the paths represented by the points and lines in the diagram are identified, the time period, number, and involved transportation section sequence corresponding to each path are extracted, for example, a marked path P001, the occurrence time period is [1663907133, 1663907150], the physical location of occurrence is point A (middle section of the shelf), and the corresponding transportation section is "urban section B". The starting and ending time positions of the path in each section are recorded. If the P001 path occurs completely within the "urban section B", the start and end times of the path in the section are recorded. If it spans two sections, they are recorded respectively. The path time span table is constructed according to the path number.

[0092] Table 3: Path section distribution table

[0093] Path number Transport section Start time stamp End time stamp P001 Urban section B 1663907133 1663907150 P002 Highway section A 1663912400 1663912460 P002 Mountain section C 1663925010 1663925090

[0094] As shown in Table 3, the table records the specific occurrence time of each high-attention path in different transportation geographical sections in detail. The path section distribution table is generated based on this table.

[0095] The path section statistical submodule extracts the occurrence frequency, time span ratio, and distribution area number of each path in different sections according to the distribution of each path in the continuous transportation section in the path section distribution table. The paths are screened according to the occurrence frequency threshold and the distribution range threshold to obtain the coverage distribution path set.

[0096] According to the distribution of each path in the continuous transportation section in the path section distribution table, the occurrence frequency, time span ratio, and distribution area number of each path in different sections are extracted. For the path P002 in Table 3, the occurrence frequency is 2 times, the distribution area number is 2 (highway section A and mountainous area section C), and the time span ratio is calculated as the ratio of the total duration of the path to the total time of the section. Assuming that the total time of highway section A is 3600 seconds, the time span ratio of P002 in the section is (1663912460-1663912400) / 3600=60 / 3600≈0.0167. The paths are screened according to the occurrence frequency threshold and the distribution range threshold. The occurrence frequency threshold is set to 2 times and the distribution range threshold is set to 2. The setting of these two thresholds is based on the review of 200 historical transportation tasks. It is found that the occurrence frequency of the path of the risk event leading to the loss of bulk goods is not less than 2 times, and at least two different physical or geographical sections are affected. Therefore, this combination is selected as the screening standard. When the occurrence frequency of a path is greater than or equal to and the distribution area number is greater than or equal to ​When the path is reserved, in the example, the occurrence frequency of path P002 is 2, the number of distribution areas is 2, the condition is met, and therefore P002 is screened out, and P001 is eliminated because it only occurs once. All paths meeting the condition are summarized to obtain a covered distribution path set.

[0097] The risk trend classification submodule calls the number, time span and fluctuation direction of the paths in the covered distribution path set in each section, extracts attribute similar paths according to time continuity and regional consistency, merges and classifies them according to the time sequence consistency between paths and the number of section intersections, and generates a risk sense knowledge recognition result;

[0098] The number, time span and fluctuation direction of the paths in the covered distribution path set in each section are called, for example, path P002 occurs in highway section A and mountainous area section C, and the fluctuation direction (determined by the synchronous trend change structure) is positive (temperature and humidity rise) in both cases. According to time continuity and regional consistency, attribute similar paths are extracted, this process regards the two occurrences of P002 in different sections as attribute similar events, and according to the time sequence consistency between paths and the number of section intersections, they are merged and classified. The time sequence consistency is calculated by dynamic time warping (DTW) distance, the normalized temperature sequence of the two occurrences of P002 is calculated, and the DTW distance is 0.15. The consistency threshold is set to 0.3, because 0.15 < 0.3, the time sequence patterns of the two occurrences are considered to be highly consistent, and the number of section intersections is the total number of sections affected by the path, which is 2 here. Attribute similar paths (fluctuation direction is the same, and DTW distance is less than the threshold) that show repeatability in time and space are merged and classified into one risk trend, for example, the two occurrences of P002 are classified as "persistent warming and humidification risk", and the type, affected range (highway section A, mountainous area section C) and occurrence time sequence of the risk are output, and a risk sense knowledge recognition result is generated.

[0099] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.​

Claims

1. A cold chain system key indicator risk awareness system, characterized in that, The system comprises: An environmental disturbance capturing module acquires temperature and humidity data of the middle section, air duct opening and tail stacking area of the refrigerated truck rack, extracts temperature section difference nodes, screens humidity curve turning sections, extracts wind speed jump trajectory, compares fluctuation amplitudes in each type of path and summarizes fluctuation ranges, generates a multi-point disturbance feature set, and the multi-point disturbance feature set comprises temperature extreme difference node distribution, humidity fluctuation gradient feature and wind speed frequency jump section; An index trend screening module constructs a wind speed path according to the same direction within a period based on the temperature and humidity paths in the multi-point disturbance feature set, matches the change direction, extracts path combinations consistent in direction within the same time period, screens data satisfying each change direction synchronization, obtains a synchronous trend change structure, and the synchronous trend change structure comprises a path direction consistent section, a time overlap combination relationship and three types of synchronous trend modes; A limit drift identification module constructs a temperature and humidity change ratio trajectory based on the synchronous trend change structure, calls a delayed path of the cargo stacking area, judges whether the ratio appears across the border within the original boundary, identifies the offset section and marks the limit adjustment position, outputs a drift trigger boundary structure, and the drift trigger boundary structure comprises a fluctuation amplitude ratio interval, a boundary offset period and a response path drift feature; An attention degree distribution module calls surface response data to construct an overlap mapping diagram based on the offset path in the drift trigger boundary structure, divides level labels according to the overlap of temperature and humidity fluctuations and surface reaction directions, generates a change attention tilt distribution diagram, and the change attention tilt distribution diagram comprises a trajectory overlap duration section, a direction consistency level and a high-level fluctuation path distribution; The attention degree distribution module comprises a trajectory overlap extraction submodule, a time direction label division submodule and an attention level confirmation submodule; The trajectory overlap extraction submodule calls cargo surface response change data within the corresponding time period based on the temperature and humidity combined path identified as offset in the drift trigger boundary structure, acquires the overlap section of the two types of paths and matches the start and end times of each other, constructs a duration mapping diagram of each section overlap section according to the time period intersection, and generates a response overlap duration atlas; The time direction label division submodule calculates the direction consistency proportion value of the two paths in the overlap section according to the time intersection length corresponding to each overlap section in the response overlap duration atlas, compares the time length and the direction proportion with a unified interval threshold, sets the corresponding label level according to the value level, and generates an overlap label level division result; The attention level confirmation submodule calls high-level label path information in the overlap label level division result, binds the part of the path with the original temperature and humidity combination sequence, extracts the offset frequency value and the surface change intensity value of each bound path, constructs the attention distribution diagram in the corresponding position tilt direction in the time period space, and obtains the change attention tilt distribution diagram.

2. The cold chain system key performance indicator risk awareness system of claim 1, wherein, The environmental disturbance capturing module comprises a multi-point data acquisition submodule, a sequence change extraction submodule and a fluctuation path integration submodule. The multi-point data acquisition submodule acquires time series data of temperature, humidity and wind speed of a middle section of a goods shelf, a position of an air duct, a tail stacking coverage area in a refrigerated vehicle compartment, records time tags, numerical outputs and data density in a continuous period of each type of data on each collection point, performs sampling interval consistency judgment on the collected data, eliminates abnormal null values and fills in missing sections, and generates a standardized continuous data set; The sequence variation extraction submodule constructs a difference value trajectory of each section based on a temperature sequence in the standardized continuous data set, screens a variation node with the largest difference value and extracts a continuous section, locates a turning point of a humidity sequence and screens a continuous change time period, simultaneously calculates a gradient intensity in a change section, extracts a jump section from wind speed data by using a sliding window and counts an occurrence frequency, and obtains a fluctuation trajectory set. The fluctuation path integration submodule calls a continuous range of a corresponding time period in each sequence according to the difference value section, the turning path and the jump trajectory extracted from the fluctuation trajectory set, screens an overlapping interval in which change behaviors exist in temperature, humidity and wind, marks the interval as a fluctuation concentrated section, counts a joint continuous span of each type of change, and generates a multi-point disturbance feature set.

3. The cold chain system key performance indicator risk awareness system of claim 1, wherein, The index trend screening module includes a direction consistent identification submodule, a wind speed trend superposition submodule, and a time section screening submodule. The direction consistent identification submodule acquires a temperature path and a humidity path that have been marked in the multi-point disturbance feature set, detects the relative positions of trend directions in respective periods based on change direction sequences of the two paths, performs path pairing on a temperature-humidity combination with consistent directions in a time section, screens combinations according to a trend consistent proportion, and generates a temperature-humidity trend matching group. The wind speed trend superposition submodule calls a combination path in the temperature-humidity trend matching group, acquires a wind speed path change sequence in a corresponding time section, performs time section corresponding registration according to fluctuation directions of the wind speed path and direction sequences of the temperature-humidity combination, extracts a wind speed path with consistent fluctuation directions and time ranges as the temperature-humidity combination, and obtains three direction consistent structures. The time section screening submodule judges whether the three types of paths have completely overlapping time sections in a continuous time section according to time tags of the paths in the three direction consistent structures, performs section consistency judgment on the continuous length and start and end times of the overlapping section, marks a combination path that satisfies a full synchronous trend and time consistency, and generates a synchronous trend variation structure.

4. The cold chain system key performance indicator risk awareness system of claim 1, wherein, The limit drift recognition module includes an amplitude ratio construction submodule, a boundary out-of-bound judgment submodule, and a drift section extraction submodule. The amplitude ratio construction submodule acquires maximum and minimum values of each path in a same time section based on temperature-humidity combination paths contained in the synchronous trend variation structure, constructs fluctuation amplitude ratios according to value differences and reference time lengths, generates a sequence of atlas according to corresponding time sections, and generates a time section amplitude ratio atlas; The boundary out-of-bound judgment submodule calls each ratio section in the time section amplitude ratio atlas, locates upper and lower boundary sections corresponding to each time section according to a response delay path recorded by each monitoring point in a goods stacking area, judges whether each ratio section crosses the upper and lower boundary sections, and obtains a boundary out-of-bound interval section. The drift section extraction submodule extracts the path deviation trend under the corresponding time slice according to the marked boundary crossing time interval in the boundary crossing interval section, identifies the start and end change range of the original boundary position in the time interval, compares the difference of the upper and lower limit paths before and after the change and makes deviation amount annotation, and generates a drift trigger boundary structure.

5. The cold chain system key performance indicator risk awareness system of claim 1, wherein, The system further comprises: The risk state output module calls the attention path in the change attention inclined distribution diagram, counts the frequency and distribution of the path in the section, screens the path with long change duration and large fluctuation area coverage, and outputs the risk awareness recognition result in the cold chain transportation process; The risk awareness recognition result includes a high-frequency fluctuation path set, a continuous continuation section distribution, and a coverage area range statistics.

6. The cold chain system key performance indicator risk awareness system of claim 5, wherein, The risk state output module comprises an attention path identification submodule, a path section statistics submodule, and a risk trend classification submodule; The attention path identification submodule calls the fluctuation path marked as an attention item in the change attention inclined distribution diagram, extracts the time interval, number, and involved transportation section sequence corresponding to each path, records the start and end time position of the path in each section, arranges and constructs a path time span table according to the path number, and generates a path section distribution table; The path section statistics submodule extracts the appearance frequency, time span ratio, and distribution area number of each path in different sections according to the distribution of each path in the continuous transportation section in the path section distribution table, screens the path according to the appearance number threshold and the distribution range threshold, and obtains a coverage distribution path set; The risk trend classification submodule calls the number, time span, and fluctuation direction of the path in each section in the coverage distribution path set, extracts attribute similar paths according to time continuity and area consistency, merges and classifies according to the time sequence consistency and section intersection number between paths, and generates a risk awareness recognition result.

Images