Chemical production and transportation safety monitoring method based on big data

Abstract

The present application relates to the technical field of characteristic gas detection and analysis, and discloses a chemical production and transportation safety monitoring method based on big data, which comprises the following steps: first, a fluid mechanics model is established according to the geometric parameters and the liquid filling state of a container, and the intrinsic time reference is calculated; then, the characteristic gas concentration sequence is collected at a preset sampling period, the positive transition event is identified, and the waiting time sequence is formed; the time turning point parameter is obtained by fitting a double segment distribution model, and the target monitoring parameter is calculated in combination with the intrinsic time reference; the relative offset parameter of the transportation process is obtained by taking the monitoring value of the loading and standing stage as the reference; when the offset parameter continuously meets the disposal condition for a set number of times, the alarm mechanism is triggered, and the accurate early warning and safety guarantee of abnormal gas release in the transportation process are realized.

Description

Big Data-Based Safety Monitoring Methods for Chemical Production and Transportation Technical Field

[0001] This invention relates to the field of characteristic gas detection and analysis technology, and more specifically, to a method for monitoring the safety of chemical production and transportation based on big data. Background Technology

[0002] Organic peroxides are classified as self-reactive hazardous chemicals and are subject to special control under international transport regulations. For a long time, industry practice has primarily relied on temperature control measures, packaging design, and transport speed limits to mitigate the risks of thermal runaway and mechanical triggering. Meanwhile, trace gas detection methods are increasingly being applied to transport safety monitoring to track decomposition products such as carbon dioxide, acetone, or organic acids. However, existing methods often assume a relatively stable transport environment, focusing primarily on temperature and average concentration levels, while neglecting the dynamic mechanical disturbances during transport.

[0003] During transportation, random vibrations, sudden impacts, and continuous friction from the vehicle act on the loading container, causing irregular sloshing of the liquid surface inside. This sloshing not only alters the contact area between the liquid surface and the gas but also triggers the periodic activation of the venting device through instantaneous pressure fluctuations. Consequently, the release of trace amounts of gas inside the container is no longer a stable, continuous process but rather exhibits a non-uniform rhythm of rapid rise → brief pause → re-release. This rhythm is influenced by container geometry, filling height, venting device design, and external vibration frequency, resulting in complex intermittent changes in the gas concentration curve.

[0004] Due to the intermittent and non-uniform nature of the release process, single gas detection signals often exhibit random, clustered increases and unpredictable recovery processes, rendering conventional methods relying on average values ​​or trends ineffective. In this situation, trace detection and monitoring systems struggle to accurately distinguish which signal changes originate from normal transport fluctuations and which indicate that organic peroxides may have entered an unstable state. In other words, mechanical disturbances during transport directly lead to unstable monitoring results, resulting in delays or misjudgments in risk assessment—a critical technical problem that urgently needs to be addressed. Summary of the Invention

[0005] This invention provides a method for monitoring the safety of chemical production and transportation based on big data, which solves the technical problems mentioned in the background.

[0006] This invention provides a method for monitoring the safety of chemical production and transportation based on big data, including:

[0007] Based on the geometric parameters of the loading container of the chemical product to be monitored and the actual liquid filling state parameters, the intrinsic time reference of the loading container is calculated through a fluid dynamics model.

[0008] The characteristic gas concentration time series of the chemical product to be monitored is collected at a preset sampling period, and positive transition events in the concentration series are located. A waiting time series is formed based on adjacent positive transition events.

[0009] A two-segment distribution model was used to fit the waiting time series in order to calculate the time transition parameters that characterize the boundary timescale between the clustered events and the reset process.

[0010] The target monitoring parameters are calculated based on the time transition parameters and the intrinsic time base.

[0011] The monitoring parameters during the static period after loading are used as the judgment benchmark value, and the relative offset parameter of the target monitoring parameter relative to the judgment benchmark value is calculated; wherein, the monitoring parameters and the target monitoring parameters are obtained in the same way, the static period to which the monitoring parameters belong, and the transportation period to which the target monitoring parameters belong;

[0012] A preset trigger condition based on the relative offset parameter is established. When the relative offset parameter continuously meets the trigger condition for a set number of times, a preset alarm mechanism is activated.

[0013] The beneficial effects of this invention include: by introducing fluid dynamics modeling and big data analysis technology, the geometric parameters and liquid filling state of the chemical loading container are transformed into an intrinsic time reference. Combined with the dynamic release signal of characteristic gases, a two-segment distribution model is used to identify the boundaries of the burst release and reset processes, thereby obtaining target monitoring parameters that characterize the intensity of transportation disturbances. Compared to traditional solutions relying on average concentration or temperature control, this invention effectively overcomes monitoring distortions caused by random disturbances such as mechanical vibration and impact during transportation, significantly improving the accuracy and real-time performance of risk assessment. Simultaneously, by setting relative offset parameters and disposal trigger conditions, early warnings can be issued before abnormal release trends stabilize, reducing misjudgment rates and delays, and ensuring transportation safety. This enables more accurate and reliable safety monitoring and early warning support for hazardous chemicals during transportation. Attached Figure Description

[0014] Figure 1 is a flowchart of the chemical production and transportation safety monitoring method based on big data according to the present invention. Detailed Implementation

[0015] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0016] As shown in Figure 1, the big data-based method for monitoring the safety of chemical production and transportation includes:

[0017] Based on the geometric parameters of the loading container of the chemical product to be monitored and the actual liquid filling state parameters, the intrinsic time reference of the loading container is calculated through a fluid dynamics model.

[0018] The characteristic gas concentration time series of the chemical product to be monitored is collected at a preset sampling period, and positive transition events in the concentration series are located. A waiting time series is formed based on adjacent positive transition events.

[0019] A two-segment distribution model was used to fit the waiting time series in order to calculate the time transition parameters that characterize the boundary timescale between the clustered events and the reset process.

[0020] The target monitoring parameters are calculated based on the time transition parameters and the intrinsic time base.

[0021] The monitoring parameters during the static period after loading are used as the judgment benchmark value, and the relative offset parameter of the target monitoring parameter relative to the judgment benchmark value is calculated; wherein, the monitoring parameters and the target monitoring parameters are obtained in the same way, the static period to which the monitoring parameters belong, and the transportation period to which the target monitoring parameters belong;

[0022] A preset trigger condition based on the relative offset parameter is established. When the relative offset parameter continuously meets the trigger condition for a set number of times, a preset alarm mechanism is activated.

[0023] In one embodiment of the present invention, based on the geometric parameters and actual liquid filling state parameters of the loading container of the chemical product to be monitored, the intrinsic time reference of the loading container is calculated through a fluid dynamics model, including:

[0024] Obtain the container type and its corresponding geometric parameters. Container types include: rectangular box or cylindrical barrel.

[0025] The geometric parameters of the rectangular box include the length of its long side and its inner width and height; the geometric parameters of the cylindrical barrel include its cylinder radius and inner height.

[0026] Obtain the liquid level height inside the loading container;

[0027] Calculate the first-order wavenumber based on the container type, including:

[0028] When the container type is a rectangular box, the first wave number is equal to the circumference constant divided by the length of the longer side;

[0029] When the container type is a cylindrical barrel, the first wave number is equal to 1.841 divided by the radius of the cylinder;

[0030] The first-order angular frequency is calculated based on the gravitational acceleration, the first-order wave number, and the liquid level. The formula for calculating the first-order angular frequency includes a hyperbolic tangent function, whose independent variable is the product of the first-order wave number and the liquid level. The intrinsic time base is calculated based on the first-order angular frequency, which is equal to twice the circumferential constant divided by the first-order angular frequency.

[0031] In detail, the intrinsic time reference is a time scale parameter that reflects the inherent period of liquid surface sloshing under the actual liquid filling state and the geometric parameters of the loading container itself. It is an inherent time characteristic index of the container itself.

[0032] In detail, the first-order wavenumber is a key spatial characteristic parameter describing the sloshing motion of the liquid surface within the loading container, used to quantify the spatial periodicity of the sloshing motion.

[0033] In detail, the first-order angular frequency is a vibration frequency parameter that characterizes the first-order inherent sloshing mode of the liquid surface in the container. It is used to quantify the time periodicity of the liquid surface sloshing and reflect the speed at which the liquid surface naturally sloshes in the container.

[0034] In one embodiment of the present invention, a time series of characteristic gas concentrations of the chemical product to be monitored is collected at a preset sampling period, positive transition events in the concentration series are located, and a waiting time series is formed based on adjacent positive transition events, including:

[0035] The concentration of characteristic gases is collected at a preset sampling period to form a concentration time series;

[0036] Calculate the differential rate of change of the concentration time series. The differential rate of change is equal to the concentration at the i-th sampling time minus the concentration at the (i-1)-th sampling time, and then divided by the preset sampling period to obtain the rate of change sequence.

[0037] First, calculate the median absolute deviation of the rate of change sequence. The median absolute deviation is the median of the absolute differences between each value in the rate of change sequence and the median of the sequence. Then, multiply the median absolute deviation by the preset threshold coefficient to obtain the event threshold.

[0038] Determine positive transition events and form an event time sequence. The conditions for determining positive transition events include:

[0039] If the rate of change at the i-th sampling time is greater than the event threshold, the rate of change at the i-th sampling time is greater than the rate of change at the (i-1)-th sampling time, and the rate of change at the i-th sampling time is not less than the rate of change at the (i+1)-th sampling time, then it is determined that there is a positive transition event at the i-th sampling time.

[0040] Calculate the time interval between positive transition events at adjacent sampling times. The time interval of the p-th positive transition event is the sampling time corresponding to the p-th positive transition event minus the sampling time corresponding to the (p-1)-th positive transition event. Combine the time intervals of all positive transition events to form a waiting time sequence.

[0041] In detail, the concentration time series is a data sequence formed by continuously collecting the concentration of characteristic gases released by the chemical product to be monitored through a fixed preset sampling period, and arranging the concentration value corresponding to each sampling moment in chronological order.

[0042] In detail, the differential rate of change is a parameter that quantifies the speed and direction of change of the concentration of a characteristic gas between adjacent sampling times. It is calculated by the ratio of the concentration difference between adjacent times to the sampling period.

[0043] In detail, the median absolute deviation is a statistic used to measure the dispersion of a rate of change series. It is calculated in two steps: First, calculate the absolute difference between each value in the rate of change series and the median of the series; second, take the median of these absolute differences, which is the median absolute deviation. Compared to the mean and standard deviation, the median absolute deviation is more robust to outliers in the rate of change series (such as occasional concentration fluctuations) and more accurately reflects the overall dispersion of the rate of change series.

[0044] In detail, the event threshold is calculated by the median absolute deviation of the rate of change sequence and a preset coefficient. The median absolute deviation is characterized by its strong resistance to accidental noise and minor fluctuations in the sequence, accurately reflecting the normal dispersion level of the rate of change. If this condition is not met, a small increase in the rate of change may be due to environmental interference (such as minor sensor drift) or random fluctuations in the natural diffusion of gas, rather than a rapid release caused by mechanical disturbances (vibration, sloshing). Due to the rapid release of characteristic gases caused by mechanical disturbances, the rate of change in concentration exhibits a single-peak trend of first increasing and then decreasing (from a slow increase to a rapid increase, and then a decrease). If the rate of change at this moment is not greater than that at moment i-1, it indicates that the rate of change has entered the decreasing phase. The increase at this time may be an accidental minor rebound during the decreasing process (such as a brief reverse movement of liquid surface sloshing), rather than an effective increase during the release process. This condition ensures that the positioning moment is in the increasing phase of the rate of change, which is a key node approaching the release peak, avoiding misjudging the interference point in the decreasing phase as a release event. The peak point of the rapid release concentration change rate is the moment when the concentration rises the fastest during a single release, and it is also the core time node that best represents the release triggered by mechanical disturbance. Specifically, the rate of change increases before the peak (enhanced release), and decreases after the peak (weakened release). If this condition is not met, it means that the moment is still in the increasing phase of the rate of change, and the faster rise point is at time i+1. In this case, time i is not the true release peak, which can lead to the same release process being misjudged as different events multiple times. This condition can accurately pinpoint the peak moment of a single release, ensuring that each positive transition event corresponds to a unique and most representative release node.

[0045] In detail, the waiting time series is a data sequence formed by arranging the sampling time differences corresponding to two adjacent positive transition events in chronological order, and is used to reflect the temporal distribution pattern of positive transition events.

[0046] In one embodiment of the present invention, a two-segment distribution model is used to fit the waiting time series to calculate the time transition parameters characterizing the boundary timescale between the clustered event and the reset process, including:

[0047] Set the initial turning time τ 0 Convergence threshold ε and maximum number of iterations K, τ 0 The median of the waiting time series is taken, ε takes the value of twice the preset sampling period, and K is a positive integer greater than or equal to 3; where τ 0 ε is the initial value for iterative calculation, and ε is the error threshold for determining iterative convergence.

[0048] Set the cutoff time , The value range is three times the preset sampling period; the waiting time sequence is determined based on the turning point τ of the current iteration. Divided into short sample sets With long sample sets The division relationship is as follows and ;

[0049] Based on short sample sets Calculating the power law exponent using the maximum likelihood method The calculation formula is: ,in, Let be the shape parameter of the short-segment power-law distribution in the k-th iteration. The number of samples in the short sample set. It is the natural logarithm function. A single sample within a short sample set;

[0050] Based on power law index Calculate the short-segment risk rate The calculation formula is:

[0051] ;in, The risk rate of any waiting time t within the short sample set in the kth iteration;

[0052] Based on long sample sets arithmetic mean Calculating the exponential rate using the maximum likelihood method The calculation formula is ;in, Let be the long-segment exponential distribution rate parameter for the k-th iteration. for The average of all samples in the sample;

[0053] exponential rate As a long-term risk rate ;in, The risk rate of any waiting time t within the long sample set during the k-th iteration;

[0054] Establish a boundary condition where two segments have equal hazard rates at the transition time, i.e. ;in, This indicates the turning point time of the (k+1)th iteration;

[0055] Derive the explicit update formula for the turning point time based on the boundary conditions. ,like If the number of iterations reaches K, the iteration terminates, and the turning point at the termination of the iteration is recorded as the time turning point parameter.

[0056] In detail, the initial turning point time is the starting reference value for the iterative calculation of the two-segment distribution model, used to initially divide the short-segment and long-segment sample sets. The initial turning point time is taken as the median of the waiting time series; the median is chosen because it is robust to outliers in the waiting time series (such as extremely short or extremely long time intervals), and can provide a reasonable starting point for iteration that is close to the center of the data distribution, avoiding excessive deviation of the initial value that would lead to difficulties in iteration convergence.

[0057] In detail, the convergence threshold is an error standard used to determine whether the iterative process has reached a stable state. It controls the accuracy of the iteration results and its value ranges from twice the preset sampling period. This value matches the time scale of data acquisition, avoiding both excessive iterations and inefficiency due to an excessively small threshold, and insufficient accuracy of the iteration results due to an excessively large threshold, which fails to accurately reflect the boundary between the two distributions. When the difference in the turning point time between two adjacent iterations is less than the convergence threshold, it indicates that the iteration results have tended to stabilize.

[0058] In detail, the maximum number of iterations is an upper limit set to avoid infinite iteration cycles, serving as a constraint to balance computational accuracy and efficiency. Even if the iteration does not reach the required accuracy for convergence, the iteration is forcibly terminated when the maximum number of iterations is reached, thus preventing infinite iterations caused by data fluctuations.

[0059] In detail, the lower cutoff time is the lower limit criterion for filtering waiting time samples, used to eliminate abnormally short intervals in the waiting time series. The lower cutoff time is used to purify the data source of short sample sets, avoid abnormally short intervals interfering with the fitting accuracy of the power-law distribution, and ensure that short sample sets can truly reflect the temporal distribution pattern of clustered events.

[0060] In detail, the waiting time series is a data sequence formed by arranging the sampling time differences corresponding to two adjacent positive transition events in chronological order.

[0061] In detail, the short sample set is a subset of all samples in the waiting time series that satisfy the condition that the waiting time is greater than or equal to the next cutoff time and less than or equal to the current iteration transition time. The short sample set corresponds to the time interval of cluster events in the transport of organic peroxides, that is, the time interval between adjacent positive transition events when mechanical disturbances (vibration, swaying) cause the rapid and dense release of characteristic gases, and its distribution follows a power law.

[0062] In detail, the long sample set is a subset of all samples in the waiting time series that satisfy the condition that the waiting time is greater than the current iteration transition time. The long sample set corresponds to the time interval of the reset process in the transportation of organic peroxides, that is, the time interval between adjacent positive transition events after the ventilation device is activated and the gas concentration in the container returns to stability, and its distribution follows an exponential law.

[0063] In detail, the maximum likelihood method calculates the power law exponent by maximizing the probability of occurrence of a short sample set, thus solving for the shape parameter (i.e., the power law exponent) of the short-segment power law distribution. The calculation process is as follows: the power law exponent is calculated by adding the number of samples in the short sample set to the sum of the natural logarithms of the ratios of each sample in the short sample set to the lower cutoff time; thus, the calculated power law exponent maximizes the probability of occurrence of the current short sample set, thereby best reflecting the actual distribution of the short sample set.

[0064] In detail, the long-segment exponential distribution rate parameter is a rate parameter that characterizes the exponential distribution of a long-segment sample set, obtained through the maximum likelihood method. The formula for calculating the exponential rate parameter is: 1 divided by the difference between the arithmetic mean of the long-segment sample set and the current iteration transition time. This logic is based on the statistical characteristics of the exponential distribution, ensuring that the calculated rate parameter maximizes the probability of the long-segment sample set's occurrence, closely reflecting the time distribution of the reset process. The long-segment exponential distribution rate parameter quantifies the decay rate of the exponential distribution; that is, the larger the exponential rate parameter, the shorter the time interval of the long-segment samples, and the faster the reset process; conversely, the smaller the parameter, the slower the reset process, reflecting the efficiency of container ventilation and gas mixing.

[0065] In detail, the long-segment hazard rate is the probability of a positive transition event occurring per unit time in a long-segment sample set. The long-segment hazard rate is equal to the long-segment exponential distribution rate parameter. In an exponential distribution, the hazard rate is constant, meaning that the probability of a positive transition event occurring at any time during the reset process is stable, which is consistent with the physical characteristic that gas release tends to stabilize during the reset process.

[0066] In detail, the boundary condition refers to ensuring that the risk rates of the short segment and the long segment are equal at the new iteration transition time. That is, the probability of the short segment's event occurring at that time point is consistent with the probability of the long segment's event occurring at that time point. The time transition parameter is the boundary timescale between the clustered event and the reset process. At this timescale, the event probabilities of the two processes should be continuous; otherwise, contradictory event probabilities will occur at the same time point. Therefore, the constraint of equal risk rates is needed to ensure the rationality of the transition time.

[0067] In detail, it is known that the short-segment hazard rate is equal to the power law exponent minus one divided by the inflection time, and the long-segment hazard rate is equal to the exponential rate parameter. Based on the boundary condition that the two hazard rates are equal, the two can be combined to solve for the new inflection time, that is, the new inflection time is equal to the power law exponent minus one divided by the exponential rate parameter. This formula is the explicit update formula.

[0068] In detail, there are two scenarios for iteration termination: First, the difference between two adjacent transition times is less than the convergence threshold. In this case, the two hazard rates have been stably matched, and the transition time accurately reflects the boundary between clustering and resetting. Second, the maximum number of iterations is reached. After multiple parameter adjustments, the transition time has essentially eliminated the initial value deviation. Although not fully converged, it meets the accuracy requirements for engineering applications, and the results are reliable. The time transition parameter is defined as the boundary timescale characterizing the clustering event and the resetting process. The transition time at the end of the iteration is a boundary timescale obtained by fitting a two-segment distribution (power law characterizing clustering, exponential characterizing resetting) and constraining equal hazard rates. This transition time can quantitatively distinguish between dense releases caused by mechanical disturbances and stable releases caused by ventilation recovery.

[0069] In one embodiment of the present invention, the ratio of the time transition parameter to the intrinsic time base is used as the target monitoring parameter.

[0070] In detail, the target monitoring parameter is a dimensionless parameter used to quantify the correlation between the characteristic gas release rhythm and the inherent characteristics of the container, obtained by comparing the time transition parameter with the intrinsic time base. The target monitoring parameter reflects the coupling effect between the intensity of mechanical disturbance and the container's response characteristics. If the target monitoring parameter deviates from the baseline, it indicates that the change in the gas release rhythm caused by the mechanical disturbance has exceeded the stable range corresponding to the inherent characteristics of the container, suggesting a potential safety risk.

[0071] In one embodiment of the present invention, the difference between the target monitoring parameter and the judgment benchmark value is calculated to obtain the relative offset parameter.

[0072] In detail, the relative offset parameter is a quantitative indicator obtained by calculating the difference between the target monitoring parameter and the judgment benchmark value. It is used to characterize the degree of deviation of the characteristic gas release dynamics from the stable benchmark state during the transportation period. During the stationary period, there are no mechanical disturbances such as vehicle vibration or impact, and the release of characteristic gases in the container is in a relatively stable benchmark state. During the transportation period, the target monitoring parameter is affected by mechanical disturbances, and the difference between it and the judgment benchmark value (relative offset parameter) can reflect whether the changes in the gas release rhythm caused by mechanical disturbances during transportation deviate from the stable benchmark. If the difference is negative and the absolute value is large, it indicates that the cluster events during transportation are more frequent, the reset process is faster, and the gas release dynamics deviate from the stable state more significantly. The relative offset parameter corresponds to the change in the intensity of transportation mechanical disturbances. When the relative offset parameter continues to deviate from the benchmark, it can conversely confirm that the mechanical disturbances such as vibration and impact in the current transportation environment may exceed the reasonable range, and timely intervention is required to ensure transportation safety.

[0073] In one embodiment of the present invention, a pre-set processing trigger condition based on a relative offset parameter is provided. When the relative offset parameter continuously meets the processing trigger condition a set number of times, a pre-set alarm mechanism is activated, including:

[0074] Set a treatment threshold, which is a constant less than zero;

[0075] Set a set number of attempts, where the set number is a positive integer not less than two;

[0076] For each sampling time, the relative offset parameter is determined. If the relative offset parameter is less than or equal to the processing threshold, the corresponding sampling time is marked as meeting the processing triggering condition; otherwise, the corresponding sampling time is marked as not meeting the processing triggering condition.

[0077] When the relative offset parameter is present for a set number of consecutive times and the trigger condition is met, the preset alarm mechanism is activated.

[0078] In detail, the handling threshold is a numerical standard for determining whether the relative offset parameter reaches an abnormal state requiring attention; it is the boundary between normal fluctuations and abnormal deviations. The handling threshold is a constant less than zero. Because the relative offset parameter represents the difference between the monitoring results during transportation and the static stability benchmark, a negative difference indicates that the gas release dynamics during transportation (more concentrated clustering, faster reset) deviate from the static stability state. The larger the absolute value of the negative difference, the more significant the deviation and the higher the risk. Therefore, the threshold is set as a negative constant, and monitoring is only conducted for risky negative deviations.

[0079] In detail, the set number of times is the lower limit of the number of consecutive sampling moments that must meet the triggering conditions to activate the alarm mechanism. It is a constraint to filter out random fluctuations and avoid false alarms. The set number of times must be a positive integer not less than two. If the relative offset parameter of a single sampling moment meets the threshold, it may be due to random factors such as detection noise or instantaneous disturbances, and does not represent a stable abnormal state. If it meets the threshold twice or more consecutively, it indicates that the abnormal deviation has formed a stable trend, eliminating the possibility of random interference and ensuring that the alarm targets real risks.

[0080] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A method for monitoring the safety of chemical production and transportation based on big data, characterized in that, include: Based on the geometric parameters and actual liquid filling parameters of the loading container of the chemical product to be monitored, the intrinsic time reference of the loading container is calculated using a fluid dynamics model. The characteristic gas concentration time series of the chemical product to be monitored is collected at a preset sampling period, and positive transition events in the concentration series are located. A waiting time series is formed based on adjacent positive transition events. A two-segment distribution model is used to fit the waiting time series to calculate the time transition parameters characterizing the boundary between the clustering event and the reset process, including: setting the initial transition time τ. 0 Convergence threshold ε and maximum number of iterations K, τ 0 The median of the waiting time series is taken, ε takes the value of twice the preset sampling period, and K is a positive integer greater than or equal to 3; where τ 0 The initial value for the iterative calculation is ε, which is the error threshold for determining iterative convergence; the lower cutoff time is set. ， The value range is three times the preset sampling period; the waiting time sequence is determined based on the turning point τ of the current iteration. Divided into short sample sets With long sample sets The division relationship is as follows and Based on short sample sets Calculating the power law exponent using the maximum likelihood method The calculation formula is: ,in, Let be the shape parameter of the short-segment power-law distribution in the k-th iteration. The number of samples in the short sample set. It is the natural logarithm function. For a single sample in a short sample set; based on the power law exponent Calculate the short-segment risk rate The calculation formula is: ;in, The risk rate of any waiting time t within the short sample set in the kth iteration; based on the long sample set. arithmetic mean Calculating the exponential rate using the maximum likelihood method The calculation formula is ;in, Let be the long-segment exponential distribution rate parameter for the k-th iteration. for The average of all samples in the sample; the exponential rate As a long-term risk rate ;in, Let be the risk rate for any waiting time t within the long sample set during the k-th iteration; establish the boundary condition where the risk rates of two segments are equal at the inflection time, i.e. ;in, This represents the turning point time of the (k+1)th iteration; the explicit update formula for the turning point time is derived based on the boundary conditions. ,like If the number of iterations reaches K, the iteration terminates, and the turning point at the termination of the iteration is recorded as the time turning point parameter. The target monitoring parameter is calculated based on the time turning point parameter and the intrinsic time base. The monitoring parameter during the static period after loading is used as the judgment base value, and the relative offset parameter of the target monitoring parameter relative to the judgment base value is calculated. The monitoring parameters and the target monitoring parameters are obtained in the same way. The static period to which the monitoring parameters belong and the transportation period to which the target monitoring parameters belong are defined. A pre-set handling trigger condition based on the relative offset parameter is set. When the relative offset parameter continuously meets the handling trigger condition for a set number of times, the pre-set alarm mechanism is activated.

2. The method for monitoring the safety of chemical production and transportation based on big data according to claim 1, characterized in that, Based on the geometric parameters and actual liquid filling parameters of the loading container of the chemical product to be monitored, the intrinsic time reference of the loading container is calculated through a fluid dynamics model. This includes: obtaining the container type and corresponding geometric parameters, where the container type includes a rectangular box or a cylindrical barrel; the geometric parameters of a rectangular box include the length of the long side and the inner width and height; the geometric parameters of a cylindrical barrel include the radius of the cylinder and the inner height; obtaining the liquid level height inside the loading container; calculating the first-order wavenumber according to the container type, where: for a rectangular box, the first-order wavenumber equals the circumference constant divided by the length of the long side; for a cylindrical barrel, the first-order wavenumber equals 1.841 divided by the radius of the cylinder; calculating the first-order angular frequency based on gravitational acceleration, the first-order wavenumber, and the liquid level height, where the formula for the first-order angular frequency includes a hyperbolic tangent function, the independent variable of which is the product of the first-order wavenumber and the liquid level height; and calculating the intrinsic time reference based on the first-order angular frequency, where the intrinsic time reference equals twice the circumference constant divided by the first-order angular frequency.

3. The method for monitoring the safety of chemical production and transportation based on big data according to claim 2, characterized in that, The process involves: collecting time series of characteristic gas concentrations of the monitored chemical product at a preset sampling period; locating positive transition events in the concentration series; and forming waiting time series based on adjacent positive transition events. This includes: collecting characteristic gas concentrations at a preset sampling period to form a concentration time series; calculating the differential rate of change of the concentration time series, where the differential rate of change equals the concentration at the i-th sampling time minus the concentration at the (i-1)-th sampling time, divided by the preset sampling period to obtain the rate of change sequence; first calculating the median absolute deviation of the rate of change sequence, where the median absolute deviation is the median of the absolute differences between each value in the rate of change sequence and the sequence median, and then multiplying the median absolute deviation by a preset threshold coefficient. Obtain the event threshold; determine positive transition events and form an event time sequence. The conditions for determining a positive transition event include: the rate of change at the i-th sampling time is greater than the event threshold, the rate of change at the i-th sampling time is greater than the rate of change at the (i-1)-th sampling time, and the rate of change at the i-th sampling time is not less than the rate of change at the (i+1)-th sampling time. If a positive transition event exists at the i-th sampling time, then the time interval between positive transition events at adjacent sampling times is calculated. The time interval between the p-th positive transition event is the sampling time corresponding to the p-th positive transition event minus the sampling time corresponding to the (p-1)-th positive transition event. The time intervals of all positive transition events are combined to form a waiting time sequence.

4. The method for monitoring the safety of chemical production and transportation based on big data according to claim 3, characterized in that, The ratio of the time transition parameter to the intrinsic time base is used as the target monitoring parameter.

5. The method for monitoring the safety of chemical production and transportation based on big data according to claim 4, characterized in that, The relative offset parameter is obtained by calculating the difference between the target monitoring parameter and the judgment benchmark value.

6. The method for monitoring the safety of chemical production and transportation based on big data according to claim 5, characterized in that, A preset trigger condition based on a relative offset parameter is established. When the relative offset parameter continuously meets the trigger condition for a set number of times, a preset alarm mechanism is activated. This includes: setting a trigger threshold, which is a constant less than zero; setting a set number of times, which is a positive integer not less than two; judging the relative offset parameter at each sampling time; if the relative offset parameter is less than or equal to the trigger threshold, the corresponding sampling time is marked as meeting the trigger condition; otherwise, the corresponding sampling time is marked as not meeting the trigger condition; when the relative offset parameter continuously meets the trigger condition for a set number of times, the preset alarm mechanism is activated.

Images