CN120338647A

CN120338647A - A remote monitoring system for cold chain transportation safety

Info

Publication number: CN120338647A
Application number: CN202510415468.5A
Authority: CN
Inventors: 朱迪诗; 战学刚; 迟呈英
Original assignee: University of Science and Technology Liaoning USTL
Current assignee: University of Science and Technology Liaoning USTL
Priority date: 2025-04-03
Filing date: 2025-04-03
Publication date: 2025-07-18

Abstract

The present invention relates to the field of cold chain transportation technology, and specifically to a cold chain transportation safety remote monitoring system. In the present invention, by using the sequential probability ratio method, not only can the vibration acceleration of the goods during transportation be evaluated in real time, but also in the case of sudden vibration, the abnormality can be identified in advance, thereby avoiding damage to the goods caused by sharp turns and collisions of the vehicle. In addition, by combining historical data, the system can accurately set the normal characteristic parameters of the vibration, and compare the real-time data to further enhance the sensitivity to abnormal vibration. This monitoring method based on the sequential probability ratio test method avoids the shortcomings of the traditional fixed threshold and can more accurately determine whether the vibration is abnormal. In terms of environmental monitoring, the prediction method based on the Holt linear trend method can more dynamically reflect environmental changes than the traditional static threshold judgment, and initiate environmental adjustment measures in advance to avoid damage to the goods due to excessive fluctuations in temperature and humidity.

Description

Cold chain transportation safety remote supervision system

Technical Field

The invention relates to the technical field of cold chain transportation, in particular to a cold chain transportation safety remote supervision system.

Background

The cold chain transportation safety remote supervision system is used for realizing real-time monitoring and remote supervision of the safety state in the cold chain transportation process. The system mainly aims at technical matters such as temperature and humidity abnormality, cargo state abnormality, transportation route deviation and the like in the transportation process, data such as temperature, humidity and vibration are collected through deployment environment sensors, the data are transmitted to a background supervision platform in real time by utilizing a wireless communication means, and abnormality judgment and alarm pushing are carried out by combining a preset threshold value. In addition, the system also collects the position information of the transport vehicle based on a positioning technology, and judges whether the operation deviates from a planned route or not in a path matching mode, so that the full-flow dynamic supervision of the transport process is realized.

In the vibration monitoring aspect of the cold chain transportation in the prior art, a fixed threshold is generally used for judging whether abnormal vibration exists, however, the method is easily affected by transient vibration fluctuation, for example, when a vehicle encounters a sharp turn or an uneven road section, the system is difficult to accurately capture short-time and local vibration fluctuation, so that precautions cannot be taken when the vibration just happens, and the risk of damage to goods is increased. In the aspect of temperature and humidity monitoring, most of environment monitoring systems in the prior art judge whether an abnormality occurs or not based on a preset fixed threshold value. When the external environment changes drastically (such as sudden temperature rise or sudden humidity drop), the existing system is difficult to judge future temperature and humidity fluctuation through dynamic trend prediction, so that refrigeration or ventilation equipment cannot be adjusted timely, goods are in unsuitable environments for a long time, and the risk of damaging goods is increased. In addition, in the aspect of sensor fault diagnosis, the prior art also depends on threshold judgment, and in practical application, a plurality of temperature and humidity sensors are often deployed in cold chain transportation to ensure accuracy and comprehensiveness of data, but due to environmental changes or equipment faults, data among the sensors may be inconsistent. The prior art can not effectively solve the problem, so that when sensor data is abnormal, a system can not accurately judge a fault node, and the system misjudgment is easy to cause, so that the overall monitoring effect is influenced.

Disclosure of Invention

The invention aims to solve the defects in the prior art and provides a remote supervision system for cold chain transportation safety.

In order to achieve the purpose, the invention adopts the following technical scheme that the cold chain transportation safety remote supervision system comprises:

The vibration monitoring module acquires the vibration acceleration of the goods on the smooth road section at the initial stage of transportation through the vibration sensor as a normal vibration characteristic parameter, and calculates and records the sequential probability ratio between the current continuously acquired vibration acceleration sequence of the goods and the normal vibration characteristic parameter;

the vibration anomaly determination module compares the sequential probability ratio with a preset probability detection threshold value, and determines the anomaly of the vibration acceleration of the goods to obtain vibration anomaly event information;

The environment trend prediction module collects the temperature and humidity of the internal environment of the cold chain transport container and integrates the temperature and humidity into a corresponding temperature and humidity sequence, predicts the temperature and humidity fluctuation trend of the temperature and humidity sequence in the future period, and forms an environment trend prediction sequence;

The environment abnormality confirmation module compares the temperature and humidity sequence in the current container with the environment trend prediction sequence in real time, judges that the environment is abnormal, and obtains environment continuous abnormality information;

the sensor node diagnosis module acquires node information and corresponding measurement data of temperature and humidity sensors deployed in the cold chain transport container, calculates mutual information entropy of the measurement data among the sensor nodes, judges whether the temperature and humidity sensors are abnormal according to the mutual information entropy, and acquires a sensor node diagnosis result;

The transportation risk response module sends out a reminding signal or adjusts the refrigeration equipment and the sensor nodes based on the vibration abnormal event information, the environment continuous abnormal information and the sensor node diagnosis result to obtain a risk response regulation result.

As a further aspect of the present invention, the step of obtaining the sequential probability ratio specifically includes:

The vibration sensor is used for collecting the vibration acceleration of the goods in the initial stable road section of the transportation and integrating the vibration acceleration into a historical vibration acceleration sequence of the goods;

according to the counted historical cargo vibration acceleration sequence, determining normal vibration characteristic parameters of cargo vibration acceleration in a conventional transportation state, wherein the normal vibration characteristic parameters comprise a mean value and a standard deviation in the historical cargo vibration acceleration sequence;

And calculating and recording the sequential probability ratio of the current continuously acquired cargo vibration acceleration sequence to the historical normal vibration data sequence at the appointed moment in real time by taking the normal vibration characteristic parameters as a reference and adopting a sequential probability ratio test method to evaluate whether the current continuously acquired cargo vibration acceleration sequence has deviation from the normal vibration characteristics.

As a further aspect of the present invention, the step of obtaining the vibration abnormal event information specifically includes:

comparing the sequential probability ratio with a preset probability detection threshold, and if the sequential probability ratio exceeds the probability detection threshold, judging that the cargo vibration acceleration at the current monitoring moment is abnormal, so as to obtain an abnormal vibration analysis result;

And recording the time stamp of the abnormal occurrence in the abnormal vibration analysis result, judging the corresponding abnormal level, and summarizing the time stamp and the abnormal level to obtain vibration abnormal event information.

As a further aspect of the present invention, the step of obtaining the environmental trend prediction sequence specifically includes:

the temperature and humidity of the internal environment of the cold chain transport container are acquired in real time through a temperature and humidity sensor and are integrated into a corresponding temperature and humidity sequence;

and dynamically fitting the horizontal component and the trend component of the current temperature and humidity sequence by using a Holter linear trend method, analyzing the change rule of the fitted horizontal component and trend component in the respective time dimension, and predicting the temperature and humidity fluctuation trend in the future period to form an environment trend prediction sequence.

As a further aspect of the present invention, the step of obtaining the environment persistent abnormality information specifically includes:

the environment anomaly confirmation module compares the temperature and humidity sequence in the current container with the environment trend prediction sequence in real time, and judges the deviation direction between the current temperature and humidity and the temperature and humidity fluctuation trend in the future period according to the comparison result;

when the offset direction is continuous and stable and exceeds a preset time threshold, the current environment is considered to have continuous abnormality, and environment continuous abnormality information is obtained by verifying the corresponding state of the environment continuous abnormality.

As a further scheme of the invention, the step of obtaining the sensor node diagnosis result comprises the following steps:

The sensor node diagnosis module acquires node information of temperature and humidity sensors deployed in the cold chain transport container and corresponding measurement data, calculates mutual information entropy of the measurement data among the sensor nodes, wherein the node information is specifically a designated identifier of the temperature and humidity sensors;

and marking the sensor nodes with mutual information entropy continuously lower than a preset information threshold as abnormal sensor nodes by analyzing the real-time change rule of the mutual information entropy, and isolating the abnormal sensor nodes to obtain a sensor node diagnosis result.

As a further aspect of the present invention, the step of obtaining the risk response regulation result specifically includes:

based on the abnormal grade in the vibration abnormal event information, evaluating vibration intensity and analyzing vibration duration according to the time stamp, and transmitting a reminding signal to remind a driver to run at a reduced speed;

based on the environment continuous abnormal information, adjusting the operation mode of the refrigeration equipment or the ventilation equipment and synchronously pushing the adjustment information;

and based on the diagnosis result of the sensor node, starting a standby sensor node to replace an abnormal sensor for temperature and humidity monitoring, and synchronously correcting the configuration of the monitoring data source to obtain a risk response regulation result.

Compared with the prior art, the invention has the advantages and positive effects that:

According to the invention, by using the sequential probability ratio method, not only can the vibration acceleration of the goods in the transportation process be evaluated in real time, but also the abnormality can be identified in advance under the sudden vibration condition, so that the damage of the goods caused by sharp turns, collision and the like of the vehicle can be avoided. In addition, through combining historical data, the system can accurately set normal characteristic parameters of vibration, and compare real-time data, further enhances the sensitivity to abnormal vibration. The monitoring mode based on the sequential probability ratio test method avoids the defects of the traditional fixed threshold value, and can more accurately judge whether the vibration is abnormal or not. In the aspect of environment monitoring, compared with the traditional static threshold judgment, the prediction method based on the Hall linear trend method can dynamically reflect the change of the environment, and can start the environment adjustment measures in advance, so that the damage of goods caused by overlarge temperature and humidity fluctuation is avoided. The system can accurately identify the environment deviation and continuously abnormal conditions by comparing the temperature and humidity data with the predicted trend in real time, and timely take regulation and control measures, such as adjusting the operation of refrigeration or ventilation equipment, so as to avoid the risk caused by the environment abnormality. In addition, by calculating mutual information entropy among the sensor nodes, the system can monitor the cooperative work condition among the sensors in real time and discover the condition of sensor faults or inconsistent data in time. When the abnormality of the sensor is detected, the system can automatically start the standby sensor to replace data, so that the continuity and accuracy of data monitoring are ensured. When the sensor fails, continuous monitoring is guaranteed, the failure sensor can be isolated, the system judgment is prevented from being influenced by error data, and the stability and reliability of the whole monitoring process are improved.

Drawings

FIG. 1 is a flow chart of the system of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

Referring to fig. 1, a remote supervision system for cold chain transportation safety includes:

The sequential probability ratio obtaining step specifically includes:

At the beginning of the transportation, when the vehicle runs on a smooth road, there is usually no significant terrain change or other disturbance, so the vibration acceleration change is small, and the vehicle appears to be in a smoother state. During this time period, the vibration sensor will continuously monitor and record the vibration acceleration data of the cargo, ensuring that all the data can reflect the vibration characteristics of the smooth road section. By integrating the data of the initial smooth road section, a historical cargo vibration acceleration sequence is formed, and the data is used as a reference of normal vibration acceleration. In the data processing process, the collected original data is required to be cleaned, abnormal values and noise are removed, and the obtained vibration acceleration sequence is ensured to accurately represent the characteristics in the initial stable state of transportation. Finally, the normal vibration characteristic parameters in the conventional transportation state are obtained by calculating the mean value and standard deviation of the data.

Determining normal vibration characteristic parameters of the cargo vibration acceleration under the conventional transportation state according to the counted historical cargo vibration acceleration sequence, wherein the normal vibration characteristic parameters comprise the mean value and standard deviation in the historical cargo vibration acceleration sequence;

and determining normal vibration characteristic parameters of the cargo vibration acceleration under the conventional transportation state according to the counted historical cargo vibration acceleration sequence, wherein the normal vibration characteristic parameters comprise the mean value and standard deviation of the historical cargo vibration acceleration sequence. Firstly, statistical analysis is required for vibration acceleration data collected historically. The historical cargo vibration acceleration sequence is imported and processed using a data analysis tool such as Excel, python or Matlab software. With these tools, the mean and standard deviation of the vibration acceleration data set can be calculated to provide a basis for subsequent vibration characteristic parameters. These tools can automate the processing of large amounts of data and perform the calculation of mean and standard deviation. On this basis, the obtained mean value represents the average level of vibration acceleration in the conventional transportation state, and the standard deviation reflects the fluctuation range of the vibration acceleration, and the two represent the normal characteristics of vibration in the transportation process.

The method comprises the steps of taking normal vibration characteristic parameters as a reference, adopting a sequential probability ratio test method, and calculating and recording the sequential probability ratio of a current continuously collected cargo vibration acceleration sequence to a historical normal vibration data sequence at a designated moment in real time so as to evaluate whether the current continuously collected cargo vibration acceleration sequence has deviation from the normal vibration characteristics;

for the sequential probability ratio, the formula is used:

Wherein S _t represents a sequential probability ratio to the sampling point t at the current time, for judging whether the acceleration sequence from the 1 st sampling point to the t sampling point is more likely to belong to H ₁ (abnormal) or H ₀ (normal) as a whole, The method is characterized in that the logarithmic probability ratio of each sampling point is accumulated item by item, t is the sampling point at the current moment, log ₁₀ is a logarithmic function based on 10 and is used for accumulation judgment, X _i is the vibration acceleration acquired at the ith time point in the current continuously acquired cargo vibration acceleration sequence, the vibration acceleration is acquired by a sensor in real time, the unit is m/s ²,P(X_i|H₀), the probability density of X _i in a normal state H ₀ is based on the normal vibration characteristic parameter, the probability density of P (X _i|H₁) is the probability density of X _i in an assumed abnormal state H ₁, the average value of the normal state plus twice the standard deviation is taken as the average value of the abnormal state, and the standard deviation is kept consistent.

Assuming 5 minutes before the start of the transportation (e.g., 300 points are sampled at 1Hz frequency), the statistics of the vibration data collected on the stationary road segment show that the mean value of the normal state H ₀ is μ ₀＝0.5m/s², the standard deviation of the normal state H ₀ is σ ₀＝0.1m/s², and the probability density function is:

Where pi is used to ensure that the total probability of the normal distribution is 1, i.e. the probability density function of the normal distribution is normalized to ensure an effective probability distribution.

The parameters are specifically as follows:

In the sequential probability ratio test, the abnormal state H ₁ is a hypothetical state, and is generally constructed based on statistical assumptions about possible behaviors deviating from the normal state, and is used for testing whether the current data deviate from the normal state significantly, and is derived from sampling of actual abnormal road segments (such as sharp turns and road surface collapse) in transportation, or is defined according to empirical rules, and the hypothetical analysis shows that the vibration acceleration is generally higher and the fluctuation is also larger in the abnormal state, so that the abnormal state is constructed as follows, the mean mu ₁＝1.2m/s² of the abnormal state H ₁ and the standard deviation sigma ₁＝0.3m/s² of the abnormal state H ₁.

The probability density function is:

the parameters are specifically as follows:

assume that the current continuously acquired cargo vibration acceleration sequence is:

X= ]0.52,0.54,0.63,0.68,0.71] (unit: m/s ²).

For X ₁ = 0.52:

Probability ratio:

Log ₁₀(R₁)≈log₁₀ (0.0026) ≡2.59.

For X ₂ = 0.54:

Probability ratio:

Log ₁₀(R₂)＝log₁₀ (0.0321) ≡1.493.

For X ₃ = 0.63:

Probability ratio:

Log ₁₀(R₃)＝log₁₀ (0.128) ≡0.893.

For X ₄ = 0.68:

Probability ratio:

Log ₁₀(R₄)＝log₁₀ (0.38) ≡0.42.

For X ₅ = 0.71:

Probability ratio:

Log ₁₀(R₅)＝log₁₀ (0.795) ≡0.1.

The cumulative log-likelihood ratio S ₅ = -2.59+ (-1.493) +(-0.893) +(-0.42) +(-0.1) = -5.496 is calculated.

Wherein, R ₁ to R ₅ are ratios of abnormal state probability to normal state probability calculated at different sampling points (such as X ₁ to X ₅), respectively, and represent the degree of contrast between each sampling point and the normal vibration feature.

The result S ₅ shows that the cumulative sequential probability ratio up to the current 5 th sampling point is-5.496.

In order to avoid the delay response of the accumulation process of the sequential probability ratio to the current state change due to the influence of early data, a sliding window mechanism can be adopted in practical application to combine with the sequential probability ratio test method, and the sensitivity to the local state change is improved by only reserving the latest period of time or the latest certain number of data segments for accumulation judgment.

The sequential probability ratio test method has the advantages that each time a vibration acceleration value is acquired, the system can add the vibration acceleration value into the current sliding window, the probability density of each data segment in the window under the normal state and the abnormal state is calculated respectively, logarithms are taken, and the logarithms are accumulated to form logarithmic probability ratio statistics in the current window. Compared with the traditional single-point threshold judgment, the method has the advantages that misjudgment caused by individual abnormal values or instantaneous fluctuation is avoided, the method has better anti-interference capability, meanwhile, the method is different from the common method of statistical mean value and standard deviation in a sliding window, the method is usually required to collect a section of data with larger quantity, the average value and fluctuation of the current data sequence are estimated in the window, the judgment result is sensitive to the data distribution in the window, and especially when the data quantity in the initial sampling stage is smaller, the method is easy to be interfered by local fluctuation. According to the method, probability judgment can be carried out on each data segment at the beginning of sampling through a preset normal state and abnormal state distribution model without depending on a real-time statistical process of current data, so that state tendency of the data segment is rapidly evaluated, and early abnormal recognition and response are realized. The mechanism is particularly suitable for stable monitoring and dynamic early warning in time sequence sensitive scenes with burstiness and continuity characteristics such as transportation vibration.

By collecting vibration acceleration data of the initial smooth road section of transportation and integrating the vibration acceleration data into a historical vibration sequence, the system can establish a reference based on vibration characteristics (mean and standard deviation) in a normal transportation state. This provides an accurate reference value for subsequent real-time monitoring. On the basis, a sequential probability ratio test method is adopted to calculate and record the deviation between the current vibration acceleration and the normal vibration characteristics, and any abnormal vibration can be detected in real time. Compared with the traditional fixed threshold judgment, the method has higher sensitivity and accuracy, can discover abnormal vibration earlier, and timely take corresponding measures to ensure the safe transportation of goods.

The step of obtaining the vibration abnormal event information specifically comprises the following steps:

And comparing the sequential probability ratio with a preset probability detection threshold, and if the sequential probability ratio exceeds an upper limit or is lower than a lower limit, carrying out real-time judgment on the vibration state of the goods at the current monitoring moment so as to obtain an analysis result of whether abnormal vibration exists. For example, if the accumulated logarithmic probability ratio is S ₅ = -5.496 by the 5 th sampling point, and is far lower than the judgment lower limit-1.0, the current vibration data can be definitely judged to be obviously in accordance with the statistical characteristics of the normal state (H ₀), which indicates that the current transportation is in a stable state without triggering early warning.

To ensure the stability and confidence of the judgment, the judgment range of the sequential probability ratio is set as follows:

If S _t is more than or equal to A, receiving an abnormal state assumption H ₁, and judging that the current vibration is abnormal;

if S _t is less than or equal to B, receiving a normal state assumption H ₀, and judging that the current vibration is normal;

if B < S _t < A, the judgment result is not determined, sampling is continued and updating is accumulated.

The setting of the thresholds a and B is based on an error control principle in the SPRT theory, so as to balance a first type of error (normal error is determined as abnormal, probability is α) and a second type of error (abnormal error is determined as normal, probability is β), and the fault-tolerant design is usually performed by adopting α=0.1 and β=0.1, and the corresponding calculation formula is as follows:

Upper threshold:

lower threshold:

To facilitate the simplification of the operation, it is generally set approximately to a symmetry threshold ±1.0.

In addition, if sudden strong shock occurs in the sampling process, such as when a vehicle passes through a hollow road section, makes a sharp turn or is loaded loose, so that the shock acceleration obviously deviates from a normal state, the probability density of a plurality of corresponding sampling points in an abnormal state model is obviously higher than that of the normal model, and the logarithmic ratio is positive and continuously increases. For example, if the data continuously deviates from the normal range from the 20 th sampling point, when the data is accumulated to the 25 th sampling point, S ₂₅ =1.287 >1.0 reaches the upper limit of abnormal judgment, at this time, an abnormal vibration alarm is immediately triggered, and the abnormal vibration analysis result at the current moment is output to assist the system in decision and transportation intervention.

Recording a time stamp of the occurrence of the abnormality in the abnormal vibration analysis result, judging the corresponding abnormality level, and summarizing the time stamp and the abnormality level to obtain vibration abnormal event information;

And recording a time stamp of the occurrence of the abnormality in the abnormal vibration analysis result, and comprehensively judging the abnormality level based on the vibration data fragments in the sliding window where the time point is located. The abnormal grading is not dependent on subjective judgment, but is quantitatively graded according to preset vibration amplitude and persistence indexes. Specifically, comparing the acceleration value sequence in the sliding window with the abnormal level interval set by the system, firstly calculating the average amplitude of the data in the window And the number of continuous superscript points n _exceed, then the grade judgment can be carried out according to the following standard, namely, the assumption is thatAnd n _exceed E [3,5 ] is determined as a first level (low) anomaly ifAnd n _exceed E [5,8 ] is determined as a secondary (medium) anomaly, ifAnd n _exceed is more than or equal to 8, judging that the three-level (high) abnormality exists. And once the abnormal level is confirmed, recording a time stamp corresponding to the current sliding window end point, and writing the time stamp and the determined abnormal level into an abnormal event log to finally form vibration abnormal event information containing contents such as abnormal occurrence time, intensity grading, duration time and the like for subsequent traceability analysis and transportation safety response.

By comparing the sequential probability ratio with a preset probability detection threshold, the system can accurately judge whether abnormal vibration exists or not, and misjudgment and hysteresis in the traditional method are avoided. When the sequential probability ratio exceeds a threshold, an abnormality judgment is immediately triggered, and a time stamp and an abnormality level of occurrence of the abnormality are recorded. The processing mode not only ensures the timely identification of abnormal vibration, but also can grade abnormal events, and improves the response speed and accuracy of the system.

The environmental trend prediction sequence comprises the following steps:

The method comprises the steps that a temperature and humidity sensor-based numerical value acquisition module performs periodic data reading on the internal environment of a container, sampling frequency is set to be 1Hz, an embedded acquisition control chip is connected into a temperature and humidity signal channel, an analog-to-digital conversion unit is used for converting analog signals into digital quantities, sampled data are uploaded to a data processing unit in real time through a communication interface, temperature data are marked in units of degrees centigrade (° C) and humidity data are marked in units of relative humidity percentage (%RH), a receiving end sequentially stores temperature and humidity original data into a cache queue according to an acquisition time sequence and marks a current sampling time stamp to form a two-dimensional data structure composed of data values and time stamps, real-time updating of the temperature and humidity acquisition data is maintained through a FIFO mechanism, at any time T ', the temperature sequence can be expressed as T (T' ₁),T(t'₂),…,T(t′_n), the humidity sequence is H '(T' ₁),H'(t'₂),…,H'(t′_n), n is the current sampling quantity, and finally the temperature and humidity sequence is integrated into a temperature and humidity sequence for subsequent trend prediction processing.

Dynamically fitting the horizontal component and the trend component of the current temperature and humidity sequence by using a Holter linear trend method, analyzing the change rule of the fitted horizontal component and trend component in the respective time dimension, and predicting the temperature and humidity fluctuation trend of the future period to form an environment trend prediction sequence;

When the dynamic fitting is carried out on the current temperature and humidity sequence by utilizing the Holter linear trend method, a temperature sequence T (T ' ₁),T(t'₂),…,T(t'_n) and a humidity sequence H ' (T ' ₁),H'(t'₂),…,H'(t'_n) are firstly obtained, and the sequences are derived from continuous sampling of the temperature and humidity sensor in the paragraph 1 at equal interval time and are integrated into a time sequence in time sequence. The trend modeling of the current sequence is performed using the following formula in the fitting process:

l_t′＝α′x_t′+(1-α′)(l_t′-1+b_t′-1)

b_t′＝β'(l_t′-l_t′-1)+(1-β′)b_t'-1

Wherein x _t′ is the observed value corresponding to a time point t ' in a temperature sequence or a humidity sequence, the observed value is acquired in real time by a temperature and humidity sensor, l _t′ is the current horizontal component corresponding to the time point t ' in the temperature sequence or the humidity sequence, the current estimated reference value of the temperature sequence or the humidity sequence is represented, l _t'-1 is the horizontal component corresponding to the time point t ' -1 before the time point t ' in the temperature sequence or the humidity sequence and used for recursive updating, b _t′ is the current trend component corresponding to the time point t ' in the temperature sequence or the humidity sequence and reflecting the change speed of the temperature and humidity in unit time, b _t'-1 is the trend component corresponding to the time point t ' -1 at the time point t ' in the temperature sequence or the humidity sequence, Is an estimated value of temperature or humidity of a corresponding time point t '+h in a predicted temperature sequence or humidity sequence, h is a predicted step length, the unit is consistent with the sampling frequency, which indicates how many time units are predicted forward, α' is a smoothing coefficient of a horizontal component, β 'is a smoothing coefficient of a trend component, and the smoothing coefficient α' controls the weight of a new observed value in horizontal estimation. The setting basis is that the fluctuation and the change speed of the historical temperature and humidity sequence are adjusted. In a scene with larger fluctuation (such as a vehicle-mounted environment with high-frequency change), alpha 'is appropriately larger, so that the model can respond suddenly more quickly, and in a stable environment (such as a constant-temperature warehouse), alpha' can be appropriately smaller, so that the stability is improved. In practical setting, first collecting one hour (360 samples, sampling period 10 seconds) temperature data, and searching the optimal alpha' in 0.1,0.9 intervals through grid by fitting the minimum Mean Square Error (MSE) principle. The smoothing coefficient β 'of the trend component controls the sensitivity of the trend variation in the same manner as α'. And comparing the prediction errors under different beta', and selecting the parameter with the minimum error. Taking a transport cold chain box as an example, humidity trends per minute over 7 days are collected, the increase and decrease amplitude per minute is calculated, and a proper beta' value is selected through trend error sensitivity analysis.

Let the observation temperature sequence (unit: ° C) be, when the current sampling point t ' =1, the temperature observation value is x _t′＝1 =4.5, the observation value of the current sampling point t ' =2 is x _t′＝2 =4.7, the observation value of the current sampling point t ' =3 is x _t′＝3 =5.0, and let α ' =0.6 be α ' =0.5.

The initial horizontal component, i.e. the reference horizontal value of the first sample point, is i _t′＝1＝x_t′＝1 =4.5.

The initial trend component, i.e. the average change value per unit time between the first two points, b _t′＝1＝x_t′＝2-x_t′＝1 =4.7-4.5=0.2.

Updating the current horizontal component l _t′＝2, wherein the current observed value is x _t′＝2 =4.7, the horizontal value at the previous moment is l _t′＝1 =4.5, the trend at the previous moment is b _t′＝1 =0.2, and the current horizontal component is obtained by substituting the formula of l _t′＝2 =0.6.4.7+0.4 (4.5+0.2) =4.70.

Updating the current trend component b _t′＝2, calculating the trend from the new level value l _t′＝2 =4.70 and the old level value l _t′＝1 =4.5, b _t′＝2 =0.5· (4.70-4.5) +0.5·0.2=0.2.

Updating the current horizontal component l _t′＝3, wherein the current observed value is x _t′＝3 =5.0, the horizontal value at the last moment is l _t′＝2 =4.70, and the previous trend value is b _t′＝2 =0.2, namely l _t′＝3 =0.6.5.0+0.4 (4.70+0.2) =4.96.

The current trend component b _t′＝3 is updated, and the new and old horizontal components are l _t′＝3＝4.96、l_t′＝2＝4.70：b_t′＝3 =0.5· (4.96-4.70) +0.5·0.2=0.23, respectively.

Temperature prediction future h=2 steps (i.e. 20 seconds later):

Similarly, let the humidity observation sequence (unit:%RH) be that the humidity observation value of the 1 st sampling point is x _t′＝1 =70.0, the humidity observation value of the 2 nd sampling point is x _t′＝2 =68.5, and the humidity observation value of the 3 rd sampling point is x _t′＝3 =66.0.

Initial horizontal component, l _t′＝1＝x_t′＝1 =70.0.

The initial trend component b _t′＝1＝x_t′＝2-x_t′＝1 =68.5-70.0= -1.5.

Update the current horizontal component l _t′＝2：l_t′＝2 =0.6·68.5+0.4· (70.0-1.5) =68.5.

Update the current trend component b _t′＝2：b_t′＝2 = 0.5- (68.5-70.0) +0.5 (-1.5) = -1.5.

Update the current horizontal component l _t′＝3：l_t′＝3 =0.6·66.0+0.4· (68.5-1.5) =66.4.

Update the current trend component b _t′＝3：b_t′＝3 = 0.5 · (66.4-68.5) +0.5 · (-1.5) = -1.8.

Humidity prediction h=after 2 steps (i.e. after 20 seconds):

the results show that the trend component was continuously positive (warming) for the temperature sequence, predicted future temperature was 5.42 ℃, and negative (dehumidifying) for the humidity sequence, predicted future humidity was 62.8% rh.

And (3) adopting a linear trend extrapolation mode, and sequentially calculating the current reference level value and the change trend by dynamically updating the observed value of each time point. In the calculation process, each time new temperature or humidity data is acquired, the data is combined with the estimated value at the previous moment, the current estimated level is updated, and the current trend estimated value is further corrected by combining the trend change at the previous moment. And then, according to the current level value and the trend value, the temperature and humidity value at any time point in the future can be extrapolated and predicted, so that a continuously updatable short-term prediction mechanism is formed. Compared with the traditional method relying on integral sequence statistics or sliding window mean, the method can complete trend modeling at the initial acquisition stage without waiting for accumulating a large amount of data, and update the judgment result in real time when each new data arrives. The method has the advantages of being capable of capturing tiny fluctuation and trend change of temperature and humidity rapidly, being particularly suitable for dynamic environments sensitive to time sequence data such as cold chain transportation and the like, and achieving prospective judgment and risk early warning of future environmental states.

Through real-time acquisition temperature and humidity data and integration into the humiture sequence, the system can continuously monitor the change of environment in the cold chain transportation process. The temperature and humidity sequence is dynamically fitted by utilizing the Hall linear trend method, and the horizontal component and the trend component of the current temperature and humidity change can be analyzed, so that the trend of future temperature and humidity fluctuation can be accurately predicted. Compared with the traditional static judgment mode, the method can reflect the dynamic change of the environment, identify potential environment abnormality in advance, ensure the stability of temperature and humidity in the transportation process and avoid adverse effects on goods.

The environment continuous abnormal information acquisition step specifically comprises the following steps:

And comparing the temperature and humidity sequence in the current container with the environment trend prediction sequence in real time, and judging the offset direction between the current temperature and humidity and the temperature and humidity fluctuation trend in the future period according to the comparison result. For the temperature sequence, the current trend component is positive (temperature rise), the predicted future temperature is 5.42 ℃, and the current temperature sequence is consistent with the future temperature fluctuation trend, and the offset direction is rising. For the humidity sequence, the current trend component is negative (decreasing humidity), the predicted future humidity is 62.8% RH, indicating that the current humidity sequence is consistent with the trend of future humidity fluctuations, and the direction of the offset is decreasing. In addition, if the current trend component of the temperature sequence is negative (cooling) and the predicted future trend of the temperature fluctuation is rising, it indicates that the deviation direction of the current temperature sequence and the future trend is falling, possibly indicating that the current temperature is fluctuating but the trend is unknown, and further data acquisition and confirmation may be required, and if the current trend component of the humidity sequence is positive (rising humidity) and the predicted future trend of the humidity fluctuation is falling, the deviation direction of the current humidity sequence and the future trend of the humidity fluctuation is rising, possibly indicating that the environmental humidity is abnormally increased, and a dehumidification measure may be required.

When the offset direction is continuous and stable and exceeds a preset time threshold, the current environment is considered to have continuous abnormality, and environment continuous abnormality information is obtained by verifying the corresponding state of the environment continuous abnormality;

The time threshold is set according to a common period and tolerance that is typically based on environmental fluctuations during cold chain transportation. For example, natural changes in temperature and humidity may be affected by environmental factors (e.g., door opening and closing, traffic conditions, etc.), so the set time threshold is usually varied from 5 to 30 minutes, and is adjusted according to the stability requirements of the actual environment. If the time threshold is exceeded and the direction of the offset stabilizes, the system considers that there is a persistent anomaly in the environment. For example, the temperature is continuously increased, and if the system detects that the temperature is continuously increased within 10 continuous minutes and the rising trend is stable, the current temperature is considered to have continuous rising abnormality. If the rise amplitude exceeds a set threshold (e.g., exceeds 2 ℃), a sustained temperature rise abnormality is triggered and an alarm message is generated. The humidity continuously drops, namely if the humidity steadily drops within 15 continuous minutes and the variation amplitude exceeds a set value (such as the humidity drops by more than 0.5 percent RH per minute), the humidity is considered to be abnormal continuously, which possibly indicates that the environment is dry and humidification measures are needed. Abnormal temperature and humidity fluctuation, namely if the change of temperature and humidity is unstable but the deviation direction shows long-term fluctuation, such as the sudden drop of humidity and the continuous 20 minutes, the temperature is rapidly increased, and the change of the temperature and the humidity does not accord with the conventional mode, the system considers that equipment failure or abnormal transportation possibly exists, and further verifies and alarms.

By comparing the temperature and humidity sequence in the current container with the predicted environment trend sequence in real time, the system can accurately judge whether the temperature and humidity change deviates from the expected fluctuation trend. If this direction of deflection continues and stabilizes beyond a preset time threshold, the system will determine that the environment is continuing abnormal. The method can timely find out long-term temperature and humidity anomalies, avoid short-term fluctuation from being misjudged, and provide clear basis for subsequent environment adjustment and emergency treatment, so that the environment control precision in the transportation process is improved, and the safe transportation of goods under stable temperature and humidity conditions is ensured.

The step of obtaining the sensor node diagnosis result comprises the following steps:

For calculating mutual information entropy, the formula is adopted:

Wherein I (a, B) represents entropy corresponding to mutual information entropy between a first node a and a second node B of a temperature or humidity sensor in a cold chain shipping container, that is, total amount of shared information between measured data sequences thereof, Σ _a∈A∑_b∈B represents accumulation of joint probability items corresponding to each measured value a in all first node a data sequences and each measured value B in the second node B, grouping sampling sequences A, B according to value classifications, performing traversal summation after counting up each group of corresponding combined statistical frequencies, log ₂ is used for calculating information quantity of each item in information entropy, a is a specific temperature value in a measured sequence of the first node a of the temperature or humidity sensor, B is a specific temperature value in a measured sequence of the second node B of the temperature or humidity sensor, p (a, B) represents that when measured value a of the first node a of the temperature or humidity sensor is a, measured value B of the second node B of the temperature or humidity sensor is measured value B, the joint probability of the pair of values in the sequence is the first node a, and p (a) represents that the joint probability of the first node B of the temperature or humidity sensor is B is p.

Assuming that two temperature and humidity sensors, first and second nodes a and B, deployed within a cold chain container, are synchronously sampled during 15:00:00-15:00:40, a sampling period of 10 seconds, collecting temperature and humidity data for a total of 5 time points.

Temperature data mutual information entropy calculation (first and second nodes a and B), assuming that the temperature sampling data is a first node a temperature sequence: a _T = {4.6,4.7,4.8,4.8,4.9}, a second node B temperature sequence: B _T = {4.6,4.6,4.9,5.0,5.0}.

Each joint data pair (a _T,b_T) represents a temperature value acquired by a first node a and a second node B at the same time, where a _T represents a measured value of the first node a at a certain point in time, and B _T represents a measured value of the second node B at the same point in time. For example, the first joint data pair is (4.6), i.e., the first node A measures a temperature of 4.6℃at a certain point in time, while the second node B also measures a temperature of 4.6℃at the same point in time. Since the temperature data series of the first node A and the second node B are in one-to-one correspondence, the joint data pair is as follows ：(a₁,b₁)＝(4.6,4.6),(a₂,b₂)＝(4.7,4.6),(a₃,b₃)＝(4.8,4.9),(a₄,b₄)＝(4.8,5.0),(a₅,b₅)＝(4.9,5.0).

The frequency of occurrence of each pair of data is 1 time and the total number of samples is 5, so the joint probability of each joint pair of data is:

In the temperature data of the first node a, the occurrence frequencies of the measured values are respectively: In the measurement data representing the first node a, a value of 4.6 ℃ appears 1 time, a value of 4.7 ℃ appears 1 time, a value of 4.8 ℃ appears 2 times, and a value of 4.9 ℃ appears 1 time.

In the temperature data of the second node B, the occurrence frequencies of the measured values are respectively: in the measurement data representing the second node B, the value of the temperature of 4.6 ℃ appears 2 times, the value of the temperature of 4.9 ℃ appears 1 time, and the value of the temperature of 5.0 ℃ appears 2 times.

Mutual information entropy calculation:

In the formula, mutual information entropy between the first node A and the second node B is calculated, and the mutual information entropy represents the information quantity of data sharing between the two nodes. Each calculation represents the contribution of each pair of combined data pairs and is derived from a probability distribution.

First term calculation:

second term calculation:

third calculation:

fourth calculation:

Fifth calculation:

all calculation results were summed to I (a _T,B_T) =0.264+0.264+0.264+0.064+0.264=1.12.

The result shows that the mutual information entropy between the temperature data of the first node a and the second node B is 1.12.

Humidity data mutual information entropy calculation (first and second nodes a and B), first node a humidity sequence: a _H = {68.0,68.5,69.0,69.5,70.0}, second node B humidity sequence: B _H = {73.0,73.2,73.7,74.0,74.2}.

Each joint data pair (a _H,b_H) represents a humidity value acquired by a first node a and a second node B at the same time, where a _H represents a humidity value of the first node a at a certain point in time, and B _H represents a humidity value of the second node B at the same point in time.

For example, the first joint data pair is (68.0,73.0), i.e., the first node A measures 68.0% humidity at a certain point in time, while the second node B measures 73.0% humidity at the same point in time.

Since the humidity data sequences of the first node a and the second node B correspond one-to-one, the joint data pair ：(a₁,b₁)＝(68.0,73.0),(a₂,b₂)＝(68.5,73.2),(a₃,b₃)＝(69.0,73.7),(a₄,b₄)＝(69.5,74.0),(a₅,b₅)＝(70.0,74.2). as shown below reflects the relationship of humidity measurements of the first node a and the second node B at different points in time.

in the humidity data of the first node a, the occurrence frequencies of the measured values are respectively: the humidity measurements representing the first node a each occur equally frequently, each accounting for 20%.

In the humidity data of the second node B, the occurrence frequencies of the measured values are respectively: The humidity measurements representing the second node B each occur equally frequently, each accounting for 20%.

Mutual information entropy formula:

First term calculation:

second term calculation:

third calculation:

fourth calculation:

Fifth calculation:

All calculation results were summed to I (a _H,B_H) =0.464+0.464+0.464+0.464+0.464=2.32.

The result shows that the mutual information entropy between the humidity data of the first node a and the second node B is 2.32.

The logic of the mutual information entropy calculation formula consists in quantifying the degree of measured data sharing between two sensor nodes. In application scenarios such as cold chain transportation, a plurality of sensor nodes generally need to work cooperatively to acquire data such as temperature and humidity. However, due to environmental fluctuations or external disturbances (such as door opening, transportation shocks, etc.), there may be some difference in measurement data between different nodes. Through the calculation of mutual information entropy, the consistency and the relevance of the data among the nodes can be effectively measured. When the mutual information entropy value is higher, the data change trend of the two nodes is highly consistent, and the data synchronism of the sensor during measurement is good and the sensor works normally. When the mutual information entropy is low, it may indicate that a problem occurs in a certain node, such as a sensor failure, a calibration error, or data instability. However, if the gap between node data is large, mutual information entropy calculation can still help us distinguish whether it is a fault or not. For example, the humidity data of two nodes is widely separated, but since the temperature data trend is consistent, it can be inferred that this difference may be caused by environmental factors (such as humidity fluctuations due to door opening) rather than sensor failure. By the method, the mutual information entropy can identify real fault nodes, can distinguish changes caused by normal external environment factors, and avoid misjudging the state of the sensor, so that the stability and reliability of the system are optimized.

Through analyzing the real-time change rule of mutual information entropy, marking the sensor nodes with mutual information entropy continuously lower than a preset information threshold as abnormal sensor nodes, and isolating the abnormal sensor nodes to obtain a sensor node diagnosis result;

And according to the mutual information entropy between the temperature data of the first node and the second node is 1.12, and the mutual information entropy between the humidity data of the first node and the second node is 2.32. These mutual information entropy values can be used to evaluate the degree of data sharing between two sensor nodes. A lower mutual information entropy value may indicate that there is a large deviation in measured data between nodes, or that a certain node is faulty. For a temperature mutual information entropy value of 1.12, a relatively low value indicates that the temperature data sharing between nodes is moderate, and although certain consistency exists, possible errors or drift still exist, and the temperature mutual information entropy value is worth focusing on. If the mutual information entropy of the temperature data continues to be below a preset threshold (e.g., 0.9) over consecutive sampling periods, it may be determined that one of the sensor nodes may be faulty, resulting in insufficient synchronicity of its measurement data with the other node. The threshold value is set according to the change rule of temperature data of the system under normal working conditions and the performance characteristics of the temperature and humidity sensor. And (3) determining the normal fluctuation range of the mutual information entropy value under the stable working condition through analysis of the historical data. In practice, the threshold setting may be determined based on calibration data provided by the device manufacturer, the accuracy of the sensor, and the sampling frequency. For example, if the accuracy of the temperature sensor in the system is

And under normal conditions, the mutual information entropy of temperature change is generally between 1.1 and 1.5, 1.0 can be set as a reasonable minimum threshold value for judging whether abnormality exists. And if the mutual information entropy obtained by continuous multiple times of calculation is lower than the threshold value, indicating that the sensor node has larger deviation. On the other hand, the humidity mutual information entropy value of 2.32 indicates that the information shared by the humidity data among the nodes is strong, the humidity change trend between the two is consistent, and the mutual information entropy is high. Even if the humidity measurement value is large in phase difference, the mutual information entropy is high due to the fact that the change trend of the two nodes is similar, so that the data synchronism of the two nodes in the humidity measurement is good, and normal cooperative work can be achieved. Therefore, in this case, no abnormal fluctuation of the humidity data occurs, and no further abnormality diagnosis is required. When the mutual information entropy value is continuously lower than a preset threshold value, the system starts a sensor fault detection mechanism to isolate abnormal sensor nodes. For the temperature data in this example, if the mutual information entropy is lower than the threshold value 1.0 for a plurality of times, the system will automatically mark the node as "abnormal" and isolate the measurement data of the node so as to avoid that the error data affects the system decision.

By calculating the mutual information entropy between the sensor nodes, the system can evaluate the data consistency and the cooperative work condition between different sensors. The real-time change rule of mutual information entropy provides dynamic sensor data monitoring for the system, and helps to identify the sensors with inconsistent data. When the mutual information entropy is continuously lower than a preset threshold value, the system can judge that the sensor node possibly has faults, so that the sensor node is isolated, and the accuracy of monitoring data and the stability of the system are ensured. The method can timely find and process the sensor faults, avoid the error data from interfering the normal operation of the system, and improve the reliability of the system and the accuracy of the data.

The acquiring step of the risk response regulation result specifically comprises the following steps:

evaluating vibration intensity based on abnormal grades in the vibration abnormal event information, analyzing vibration duration according to the time stamp, and transmitting a reminding signal to remind a driver to run at a reduced speed;

The occurrence of the vibration abnormal event is usually related to sudden conditions (such as sudden braking, uneven road surface and the like) in the transportation process, and whether the abnormal risk exists in the transportation process can be timely estimated by monitoring the vibration intensity and the duration in real time. When the vibration intensity exceeds a normal preset range, the system performs real-time analysis by collecting data through the sensor, and judges the duration of vibration by combining with the time stamp information. If the vibration continuously exceeds the set threshold, the system can judge that the vibration is abnormal and send a reminding signal to the driver, so that the driver is required to run at a reduced speed, the transportation safety is ensured, and the damage to goods or equipment failure caused by excessive vibration is prevented. Through the dynamic monitoring and analysis of the vibration intensity and duration, potential risks in transportation can be effectively avoided, and the safety of goods in the transportation process is ensured.

By monitoring and analyzing the environment continuous abnormal information, the system can timely find out abnormal fluctuation of environmental parameters such as temperature and humidity and the like, and further take corresponding control measures. When the environmental anomaly data is continuously lower than a preset threshold value or exceeds a preset fluctuation range, the system starts an automatic adjusting mechanism to adjust the working state of the refrigeration equipment or the ventilation equipment. For example, if the temperature exceeds a standard, the system will automatically increase the output power of the refrigeration unit, and if the humidity drops, the system will increase the operating efficiency of the ventilation unit or activate the humidification device. Meanwhile, the adjustment information is pushed to operators and management staff through a communication network, so that the operators are ensured to know the current environment regulation and control condition, and timely respond or further adjust the equipment. Through the real-time environment abnormal response mechanism, the temperature and humidity environment in the transportation process can be ensured to meet the requirement of goods storage to the greatest extent, and the damage or other risks of goods caused by temperature and humidity abnormality are prevented.

Based on the diagnosis result of the sensor node, starting a standby sensor node to replace an abnormal sensor for temperature and humidity monitoring, and synchronously correcting the configuration of a monitoring data source to obtain a risk response regulation result;

In the diagnosis process of the sensor nodes, when the system detects that the measured data of one sensor node is continuously abnormal (such as the mutual information entropy value is lower than a set threshold value and fails to recover to be normal), the standby sensor node is started immediately. And the standby node is used for replacing the abnormal node to continuously monitor the temperature and the humidity, and synchronously updating the original configuration in the monitored data. The process ensures that the system can still maintain continuous data acquisition and environment monitoring under the condition of sensor failure or failure, and avoids influencing the monitoring and decision of the transportation process due to data interruption. Meanwhile, the system corrects the original monitoring data source configuration according to the new data of the standby sensor node, and updates the relevant alarm threshold and response mechanism. Through the automatic fault replacement and data synchronous correction, the system can rapidly respond to temperature and humidity monitoring abnormality, ensure the environmental stability in the cargo transportation process and avoid environmental monitoring loss caused by sensor faults. Finally, the system generates a risk response regulation result, including fault node substitution information, data correction conditions and the current running state of the system, and sends the risk response regulation result to relevant operators for further inspection and confirmation.

By evaluating the vibration intensity and duration based on the abnormal level in the vibration abnormal event information, the system can analyze and evaluate the severity of vibration in time and remind the driver to run at a reduced speed by transmitting a reminding signal, thereby reducing the risk of damage to goods caused by excessive vibration. For the continuous abnormal information of the environment, the system can automatically adjust the operation mode of the refrigeration equipment or the ventilation equipment according to the detected environmental problem, synchronously push the adjustment information, ensure that the environment is kept in a proper temperature and humidity range and prevent goods from being damaged. For the diagnosis result of the sensor node, the system ensures continuous temperature and humidity monitoring by starting the standby sensor node to replace an abnormal sensor, and simultaneously corrects the configuration of the monitoring data source to ensure the accuracy and the integrity of the data.

The present invention is not limited to the above embodiments, and any equivalent embodiments which can be changed or modified by the technical disclosure described above can be applied to other fields, but any simple modification, equivalent changes and modification made to the above embodiments according to the technical matter of the present invention will still fall within the scope of the technical disclosure.

Claims

1. A cold chain transportation safety remote supervision system, the system comprising:

2. The cold chain transportation safety remote supervision system according to claim 1, wherein the step of obtaining the sequential probability ratio is specifically:

3. The cold chain transportation safety remote supervision system according to claim 1, wherein the step of obtaining the vibration abnormal event information specifically comprises:

4. The cold chain transportation safety remote supervision system according to claim 1, wherein the obtaining step of the environmental trend prediction sequence specifically comprises:

5. The cold chain transportation safety remote supervision system according to claim 1, wherein the obtaining step of the environment persistent anomaly information specifically comprises:

6. The cold chain transportation safety remote supervision system according to claim 1, wherein the step of obtaining the sensor node diagnosis result specifically comprises:

7. The cold chain transportation safety remote supervision system according to claim 3, wherein the step of acquiring the risk response regulation result specifically comprises: