Production safety risk early warning method and system based on digital twinning

Abstract

The invention relates to the field of data processing, in particular to a production safety risk early warning method and system based on digital twinning, and the method comprises the steps: obtaining a real-time smoke concentration data sequence and a real-time temperature data sequence of a production workshop, and carrying out the normalization processing; sliding window division based on the smoke concentration data sequence to obtain a plurality of window sequences, calculating a fire characteristic index of each window sequence, and taking the continuity of the abnormal degree of the smoke concentration of each window sequence as a fire trend index of the smoke concentration data sequence; taking the fire characteristic index as a weight, calculating the synchronization degree of change between the smoke concentration data and the temperature data in the window sequence based on a DTW algorithm, quantifying the fire suspected degree of each window sequence, and calculating the sum of the fire suspected degrees of the window sequences as a fire anomaly index; and based on the fire trend index and the fire abnormity index, determining a fire risk comprehensive index and performing early warning on the fire risk.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to a method and system for early warning of production safety risks based on digital twins. Background Technology

[0002] In the field of production safety in enterprise workshops, welding, grinding and other processes in metal processing workshops generate a lot of smoke and dust. If ventilation is not smooth or the workpieces burn, it may cause local fires. Fire monitoring is a key link to ensure smooth production and personnel safety.

[0003] In real-world scenarios, smoke detectors are typically used to monitor fires. However, because they rely solely on gas concentration, triggering the alarm only when the concentration reaches a certain threshold, smoke concentrations are often below this threshold in the early stages of a fire. This method is not sensitive to early fire source detection, leading to delayed fire warnings and increasing the risk of accidents. Furthermore, different work conditions during production can cause smoke concentrations to fluctuate. For example, welding processes can produce high concentrations of smoke without causing localized fires, making false alarms more likely when relying solely on smoke detectors. Summary of the Invention

[0004] To address the problems of untimely warnings and false alarms caused by the influence of workshop production when using smoke detectors to monitor smoke concentration and fire conditions, the present invention provides solutions in the following aspects.

[0005] In the first aspect, the production safety risk early warning method based on digital twins includes: acquiring real-time smoke concentration data sequences and temperature data sequences from the production workshop and normalizing them; dividing the smoke concentration data sequences into multiple window sequences based on a sliding window, calculating the trend of smoke concentration changes over time in each window sequence as a fire characteristic index for each window sequence, calculating the continuity of the abnormal degree of smoke concentration in each window sequence as a fire trend index for the smoke concentration data sequences; using the fire characteristic index as a weight, calculating the degree of synchronization between changes in smoke concentration data and temperature data within the window sequence based on the DTW algorithm, quantifying the fire suspicion degree of each window sequence, and calculating the sum of the fire suspicion degrees of each window sequence as a fire anomaly index; simultaneously judging the comprehensive fire risk index based on the fire trend index and the fire anomaly index and issuing an early warning of fire risk.

[0006] This method first divides the smoke concentration data sequence into window sequences using a sliding window, calculates the fire characteristic index of each window sequence to capture localized sustained temperature rise, and then uses the fire trend index to measure the abnormal continuity across window sequences, thereby filtering out interference from instantaneous high smoke from welding and other sources. Subsequently, using the fire characteristic index as a weight, DTW is used to measure the synchronicity of smoke and temperature changes within the same window, quantifies the fire suspicion of each window, and summarizes them into a fire anomaly index. Finally, the fire trend index and the fire anomaly index are merged to obtain a comprehensive fire risk index and provide real-time early warning. This method can detect initial smoldering within seconds and significantly reduce false alarms through multi-source synchronous verification, achieving early detection, low false alarms, and traceable intelligent prevention and control of production safety.

[0007] Preferably, the fire characteristic index is calculated as follows: the difference between each data point and the average data point within the window sequence is used as the smoke concentration change; the difference between the sampling timestamp of each data point within the window sequence and the arithmetic mean of the sampling timestamps of all data points within the window is used as the time offset; the product of the smoke concentration change of all data points within the window sequence and the time offset is calculated and summed; the smoke concentration change trend is obtained by normalization; the ratio of the minimum data point to the maximum data point within the window sequence is subtracted from 1 as the smoke concentration fluctuation coefficient; the product of the smoke concentration change trend and the smoke concentration fluctuation coefficient is calculated and normalized to obtain the fire characteristic index of the window sequence.

[0008] The fire characteristic index simultaneously locks the "slope" and "fluctuation amplitude" into a dimensionless score—the slope term amplifies the continuous temperature rise, while the fluctuation term suppresses instantaneous spikes; the product of the two is then normalized, so that the index monotonically reflects the "fire-like" growth between 0 and 1. It can identify the continuous rise in the early stage of smoldering in seconds, and significantly reduce false alarms caused by short-term high smoke from welding, etc., providing a clean, sensitive and comparable window-level feature for subsequent DTW synchronous verification and trend continuity judgment.

[0009] Preferably, the fire trend index is calculated as follows: a preset threshold for the degree of increase in smoke concentration is set, the ratio of the fire characteristic index of the window sequence to the threshold for the degree of increase in smoke concentration is used as an amplification factor, an indicator function is calculated based on the fire characteristic index of the window sequence and the threshold for the degree of increase in smoke concentration, and then multiplied by the amplification factor to obtain the smoke change effect value. The sum of the smoke change effect values ​​of each window sequence is calculated and normalized to obtain the degree of abnormality in smoke concentration change.

[0010] Preferably, the fire anomaly index is calculated as follows: based on the DTW algorithm, the sum of the distances between all data points in each window sequence and the matching points of the temperature data sequence at the same time is calculated, and after normalization, exponential decay is performed to represent the synchronization degree between smoke concentration data and temperature data. This is then multiplied by the fire characteristic index to represent the fire suspicion degree of a window sequence. The sum of the fire suspicion degrees of each window sequence is calculated as the fire anomaly index.

[0011] The Fire Anomaly Index uses DTW to quantify the "smoke-temperature synchronization" into attenuation distance, and then multiplies and accumulates it with the fire characteristic index. This preserves the weak but synchronized early temperature rise signal while eliminating single sensor drift or local disturbances, achieving a highly sensitive and low-false-alarm global fire anomaly measurement.

[0012] Preferably, the method for early warning of fire risk based on the comprehensive fire risk index is as follows: calculate the product of the fire characteristic index and the fire anomaly index as the comprehensive fire risk index, preset a fire warning threshold, issue a fire warning when the comprehensive fire risk index is greater than the fire warning threshold, and do not issue a fire warning when the comprehensive fire risk index is less than or equal to the fire warning threshold.

[0013] Preferably, the window sequence is divided as follows: half the length of the smoke concentration data sequence is selected as the sliding window length, and the sliding step is set to 1 to obtain multiple window sequences.

[0014] Preferably, the data of the smoke concentration data sequence and the temperature data sequence are aligned based on time sequence, and missing data is supplemented.

[0015] Secondly, a production safety risk early warning system based on digital twins includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the production safety risk early warning method based on digital twins described in any one of the above embodiments is implemented.

[0016] The present invention has the following effects: 1. This method first divides the smoke concentration data sequence into window sequences according to a sliding window, calculates the fire characteristic index of each window sequence to capture local continuous temperature rise, and then uses the fire trend index to measure the abnormal continuity across window sequences, thereby filtering out interference from instantaneous high smoke from welding and other processes. Subsequently, using the fire characteristic index as a weight, DTW is used to measure the synchronicity of smoke and temperature changes within the same window, quantifies the fire suspicion of each window, and summarizes it into a fire anomaly index. Finally, the fire trend index and the fire anomaly index are merged to obtain a comprehensive fire risk index and provide real-time early warning. This method can detect initial smoldering within seconds and significantly reduce false alarms through multi-source synchronous verification, achieving early detection, low false alarms, and traceable intelligent prevention and control of production safety.

[0017] 2. The fire anomaly index uses DTW to quantify the "smoke-temperature synchronization" into attenuation distance, and then multiplies and accumulates it with the fire characteristic index. This preserves the weak but synchronized early temperature rise signal while eliminating single sensor drift or local disturbances, achieving a highly sensitive and low-false-alarm global fire anomaly measurement. Attached Figure Description

[0018] Figure 1This is a flowchart of steps S1-S4 in the production safety risk early warning method based on digital twins in an embodiment of the present invention.

[0019] Figure 2 This is a flowchart illustrating the structure of the production safety risk prediction system based on digital twins, as described in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0021] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] Reference Figure 1 The production safety risk early warning method based on digital twins includes steps S1-S4, as follows: S1: Obtain real-time smoke concentration and temperature data sequences from the production workshop and normalize them.

[0023] Smoke concentration data is acquired through smoke detectors, and temperature data is acquired through temperature sensors. Smoke detectors and temperature sensors are installed in key areas of the production workshop that are prone to smoke generation, such as the welding equipment area. Smoke concentration and temperature data are collected in real time at fixed intervals during the production process, forming smoke concentration data sequences and temperature data sequences. The smoke concentration data sequences and temperature data sequences are time-aligned and normalized. When data is missing, the average of the adjacent data is taken as supplementary data to ensure the integrity and correspondence of the smoke concentration data sequences and temperature data sequences.

[0024] In this embodiment, smoke concentration and temperature data can be collected every 10 seconds or 30 seconds. The smoke concentration and temperature data sequences can be sequences formed from smoke concentration and temperature data within one hour. In other embodiments, the data collection interval and the length of the smoke concentration and temperature data sequences can be flexibly set as needed.

[0025] S2: Based on the sliding window division of the smoke concentration data sequence, multiple window sequences are obtained. The trend of smoke concentration change over time in each window sequence is calculated as the fire characteristic index of each window sequence. The continuity of the abnormality of smoke concentration in each window sequence is calculated as the fire trend index of the smoke concentration data sequence.

[0026] Half the length of the smoke concentration data sequence is selected as the sliding window length, and the sliding step is set to 1, resulting in multiple window sequences. The difference between each data point and the average data point within the window sequence is calculated as the smoke concentration change. The difference between the sampling timestamp of each data point within the window sequence and the arithmetic mean of the sampling timestamps of all data points within the window is calculated as the time offset. The product of the smoke concentration change and the time offset of all data points within the window sequence is calculated and summed, and normalized to obtain the smoke concentration change trend. The ratio of the minimum data point to the maximum data point within the window sequence is subtracted from 1 to obtain the smoke concentration fluctuation coefficient. The product of the smoke concentration change trend and the smoke concentration fluctuation coefficient is calculated and normalized to obtain the fire characteristic index of the window sequence.

[0027] The formula for calculating the fire characteristic index is: ; in, This represents the first [number]th [item] in the smoke concentration data sequence. Fire characteristic index of each window, The larger the value, the greater the degree of abnormality in the smoke concentration within that window. For normalization function, Indicates the The first window The smoke concentration values ​​of each data point. This is the smoke concentration data sequence number. Each window contains the mean of the smoke concentration values ​​for all data. For the first The first window The data and the first The difference between the average smoke concentration data of all windows represents the change in smoke concentration. This represents the first data sequence of smoke concentration. The first window The timestamp of each data item The data sequence representing smoke concentration is shown in the figure. The arithmetic mean of the timestamps of all data in each window. For the first The first window The timestamp of the data and the first The difference between the arithmetic mean of the timestamps of all data within a window represents the time offset. Represents the number of smog concentration sequences. The length of each window Indicates the first Sum the products of smoke concentration changes and time offsets for all data within each window. Indicates to Normalization processing Used to reflect the The relationship between changes in smoke concentration in a window and time, i.e. whether the smoke concentration continues to increase over time.

[0028] Indicates the first The minimum smoke concentration within each window. Indicates the first The maximum smoke concentration within each window. Indicates the first The smoke concentration fluctuation coefficient within each window.

[0029] This embodiment considers the relationship between smoke concentration change trends and time when calculating the degree of smoke concentration anomaly within a window sequence, which can effectively improve the early fire identification rate and reduce false alarms. In actual production processes, different working conditions may cause changes in smoke concentration. For example, welding processes may generate high-concentration smoke, and judging only the change in smoke concentration may lead to the identification of welding smoke as a fire. Compared with the smoke generated by a fire, the smoke generated by welding is instantaneous. The smoke concentration of a fire shows a continuous upward trend over time, while the smoke generated by welding increases instantaneously in a short period of time and disappears quickly after the welding operation is completed.

[0030] To make it easier to understand, a simple example will be used below.

[0031] Select a window of length 8, and the timestamps corresponding to each data point. Represented by 0, 1, 2, ..., 7 seconds respectively. =3.5s, the smoke concentration data of the window sequence in the fire scenario is In fire scenarios, data values ​​increase over time; in welding scenarios, smoke concentration data for the window sequence are... In welding scenarios, the increase in data values ​​is not continuous. The two sets of window data above are only examples to reflect the changing trends of smoke concentration under different scenarios.

[0032] based on Calculate the trend of smoke concentration over time in fire and welding scenarios. In the fire scenario... The value is 15.75. In welding scenarios, The value is 13.125. Based on the pattern of smoke concentration changing over time in a fire scenario, By weighting the smoke concentration rise near the end of the window, a rapid fire response and early warning system can be established, while avoiding false alarms.

[0033] During the production process, if the smoke concentration increases over a certain period of time, it indicates that there are fire characteristics in that period. By observing the trend of smoke concentration changes over time within a single window sequence, it is possible to quickly determine whether there are fire characteristics in the corresponding time period of that window sequence. To further ensure the accuracy of fire early warning, it is also necessary to determine whether the occurrence of fire characteristics is continuous, that is, it is necessary to determine whether there is a fire trend at the same time.

[0034] Based on the fire characteristic index of each window sequence, calculate whether the fire characteristic index of each window is continuous, and quantify the fire trend index of a smoke concentration data sequence.

[0035] The formula for calculating the fire trend index is: ; in, A fire trend index representing a smoke concentration data series, used to reflect the continuity and persistence of smoke concentration changes within the data series. The larger the value, the more continuous the upward trend of smoke concentration is within a smoke concentration data series, and the greater the possibility of a fire. This represents the first [number]th [item] in the smoke concentration data sequence. Fire characteristic index of each window, This indicates the preset threshold for the degree of increase in smoke concentration. The value is used to determine whether the increasing trend of smoke concentration within the window is significantly abnormal. It can be selected based on experience. In this embodiment... The value is 0.5. For indicator functions, when The value is 1 when the time is right, and 0 otherwise. Ratio This is used to normalize and amplify the trend intensity of abnormal windows, reflecting the strength of local trends. Indicates the first The smoke change effect value for each window, This represents the total number of window sequences.

[0036] S3: Using the fire characteristic index as the weight, the degree of synchronization between the changes in smoke concentration data and temperature data within the window sequence is calculated based on the DTW (Dynamic Time Warping) algorithm. The fire suspicion of each window sequence is quantified, and the sum of the fire suspicion of each window sequence is calculated as the fire anomaly index.

[0037] While abnormal smoke concentration data may indicate a fire risk, it is not sufficient to determine whether a fire has occurred. To further ensure the accuracy of fire warnings, it is necessary to combine temperature data to determine whether a fire has occurred. If the smoke concentration rises abnormally along with a corresponding increase in temperature data, the likelihood of a fire will increase significantly. Therefore, the correlation between temperature data sequences and smoke concentration data sequences can be used for judgment.

[0038] Based on the DTW algorithm, the sum of distances between all data points in each window sequence of the smoke concentration data sequence and their matching points in the temperature data sequence is calculated. After normalization, exponential decay is performed to represent the degree of synchronization between the smoke concentration data and the temperature data. This is multiplied by the degree of smoke concentration anomaly to represent the fire suspicion of a window sequence. The sum of the fire suspicion of each window sequence is calculated as the fire anomaly index.

[0039] The formula for calculating the fire anomaly index is: ; in, This indicates the fire anomaly index. This represents the sum of distances between all data points within each window sequence and their matching points in the temperature data sequence, where the matching degree between the data points in the window sequence and the temperature data sequence is time-aligned. This represents the shortest path distance between the smoke concentration data sequence and the temperature data sequence calculated based on DTW, used for normalization. , This indicates that an exponentiation operation is being performed. This represents the first [number]th [item] in the smoke concentration data sequence. Fire characteristic index of each window, This represents the total number of window sequences.

[0040] S4: Simultaneously determine the comprehensive fire risk index and issue early warnings of fire risks based on the fire trend index and fire anomaly index.

[0041] The formula for calculating the comprehensive fire risk index is: ; A fire trend index representing a smoke concentration data series. The larger the value, the more continuous the fire characteristics are within a certain period of time, and the greater the possibility of a fire occurring. The fire anomaly index represents the correlation value between the smoke concentration data series and the temperature data series. The higher the value, the more significant the temperature change with increasing smoke concentration within an hour, and the greater the likelihood of a fire.

[0042] Set fire early warning thresholds , judge when There is a high probability of fire at this time, so it is necessary to sound an alarm to alert staff to check for and handle any potential sources of ignition; when In this case, no fire will occur, no alarm is needed, and production can proceed normally. In this embodiment, a threshold is set. In actual production, fire warning thresholds can be flexibly set according to the actual situation. The value of .

[0043] Reference Figure 2 The system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the production safety risk early warning method based on digital twin according to the first aspect of the present invention.

[0044] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0045] The production safety risk early warning method and system based on digital twins provided by this invention simultaneously judges fire risk based on the continuity of smoke changes over time and the correlation between smoke and temperature, thereby improving the efficiency of fire early warning and significantly reducing the possibility of false fire alarms.

[0046] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A production safety risk early warning method based on digital twins, characterized in that, include: Acquire real-time smoke concentration and temperature data sequences from the production workshop and normalize them; Multiple window sequences are obtained by dividing the smoke concentration data sequence into sliding windows. The trend of smoke concentration change over time in each window sequence is calculated as the fire characteristic index of each window sequence. The continuity of the degree of smoke concentration anomaly in each window sequence is calculated as the fire trend index of the smoke concentration data sequence. Using the fire characteristic index as the weight, the DTW algorithm is used to calculate the degree of synchronization between changes in smoke concentration data and temperature data within a window sequence, quantify the fire suspicion of each window sequence, and calculate the sum of fire suspicion of each window sequence as the fire anomaly index. Simultaneously, the comprehensive fire risk index is determined based on the fire trend index and the fire anomaly index, and fire risk is warned.

2. The production safety risk early warning method based on digital twins according to claim 1, characterized in that, The fire characteristic index is calculated as follows: The difference between each data point and the average data point within the window sequence is calculated as the smoke concentration change. The difference between the sampling timestamp of each data point within the window sequence and the arithmetic mean of the sampling timestamps of all data points within the window is calculated as the time offset. The product of the smoke concentration change of all data points within the window sequence and the time offset is calculated and summed. After normalization, the smoke concentration change trend is obtained. The ratio of the minimum data point to the maximum data point within the window sequence is subtracted from 1 as the smoke concentration fluctuation coefficient. The product of the smoke concentration change trend and the smoke concentration fluctuation coefficient is calculated and normalized to obtain the fire characteristic index of the window sequence.

3. The production safety risk early warning method based on digital twins according to claim 1, characterized in that, The fire trend index is calculated as follows: A preset threshold for the degree of smoke concentration increase is set. The ratio of the fire characteristic index of the window sequence to the threshold for the degree of smoke concentration increase is calculated as an amplification factor. The smoke change effect value is obtained by multiplying the indicator function calculated based on the fire characteristic index of the window sequence and the threshold for the degree of smoke concentration increase with the amplification factor. The sum of the smoke change effect values ​​of each window sequence is calculated and normalized to obtain the degree of abnormality in smoke concentration change.

4. The production safety risk early warning method based on digital twins according to claim 1, characterized in that, The calculation method for the fire anomaly index is as follows: Based on the DTW algorithm, the sum of distances between all data points in each window sequence and the matching points of the temperature data sequence at the same time is calculated. After normalization, exponential decay is performed to represent the synchronization degree between smoke concentration data and temperature data. This is then multiplied by the fire characteristic index to represent the fire suspicion degree of a window sequence. The sum of the fire suspicion degrees of each window sequence is calculated as the fire anomaly index.

5. The production safety risk early warning method based on digital twins according to claim 1, characterized in that, The method for issuing early warnings of fire risk based on a comprehensive fire risk index is as follows: The product of the fire characteristic index and the fire anomaly index is calculated as the comprehensive fire risk index. A fire warning threshold is preset. A fire warning is issued when the comprehensive fire risk index is greater than the fire warning threshold, and no fire warning is issued when the comprehensive fire risk index is less than or equal to the fire warning threshold.

6. The production safety risk early warning method based on digital twins according to claim 1, characterized in that, The window sequence is divided as follows: Half the length of the smoke concentration data sequence is selected as the sliding window length, and the sliding step is set to 1 to obtain multiple window sequences.

7. The production safety risk early warning method based on digital twins according to claim 1, characterized in that, The data from the smoke concentration and temperature data sequences were aligned based on time series, and missing data was supplemented.

8. A production safety risk early warning system based on digital twins, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the production safety risk early warning method based on digital twins according to any one of claims 1-7.

Images