Real-time monitoring system for safety production based on artificial intelligence

Abstract

The invention relates to the technical field of safety monitoring, in particular to a real-time monitoring system for safety production based on artificial intelligence. The system comprises a processor, a memory, a smoke sensor and a camera device, the memory stores a computer program instruction, and when the computer program instruction is executed by the processor, the following steps are realized: inputting an acquired workshop image into a semantic segmentation model, and obtaining a change sequence of smoke concentration and smoke position; inputting the change sequence of the smoke concentration and the smoke position into a trained LSTM model, and predicting future smoke concentration and smoke position; an escape route is calculated in the digital twin workshop model through superposition of the gravitational field function and the repulsive field function; whether the difference between the smoke concentration collected in real time and the future smoke concentration exceeds a preset value or not is judged, and if yes, the escape route is determined again. The safety of the escape route generated by the system is higher.

Description

Technical Field

[0001] The present invention relates to the field of safety monitoring technology, and more specifically, to a real-time monitoring system for safe production based on artificial intelligence. Background Art

[0002] As people's living standards improve, the demand for various products is also increasing. Generally, products are produced and processed in workshops. However, when producing and processing products in such a complex environment as a workshop, safety accidents caused by equipment failures are prone to occur. For example, fire. If the product happens to be flammable, the workshop personnel fail to escape in time or evacuate improperly, and no corresponding safety measures are taken, it is easy to cause casualties and damage to equipment.

[0003] With the development of artificial intelligence and big data technology, the existing technology has begun to use artificial intelligence algorithms to monitor the safety of the workshop production process in real time to avoid casualties or equipment damage caused by accidents. For example, the Chinese patent application document with publication number CN110275473A discloses an intelligent chemical production safety monitoring system, which achieves safe production in the factory by real-time monitoring of the video situation and gas concentration changes in the chemical production plant, and helps monitoring personnel take safety measures in the event of an accident by predicting future gas change trends, such as issuing alarm prompts to remind production workers to evacuate as soon as possible.

[0004] However, in the face of sudden fires, some escape routes may have fires that cause smoke concentrations to be too high. If production personnel still evacuate according to the originally planned evacuation routes, production personnel in workshops evacuating along evacuation routes with high smoke concentrations will not be able to evacuate smoothly. In this regard, how to optimize escape routes based on real-time monitoring of the safety status of the workshop is a technical problem that needs to be solved urgently. Summary of the invention

[0005] In order to solve the technical problem of low safety of the above-mentioned existing escape routes, the present invention proposes a real-time monitoring system for safe production based on artificial intelligence, including: a processor, a memory, a smoke sensor and a camera device, wherein the smoke sensor is used to collect the smoke concentration in the workshop, the camera device is used to collect workshop videos, and the memory is used to store computer program instructions. When the computer program instructions are executed by the processor, the following method is implemented: establishing a digital twin workshop model based on the BIM model; collecting workshop videos, converting the workshop videos into images and inputting them into a trained semantic segmentation model to obtain the change sequence of smoke concentration and smoke position; inputting the change sequence of smoke concentration and smoke position into a trained LSTM model to predict future smoke concentration and smoke position; calculating the escape route in the digital twin workshop model by superimposing the gravitational field function with the repulsive field function, the repulsive field function is as follows: ; In the formula, represents the repulsive field function, k represents the positive proportionality coefficient, Represents a vector whose size is the Euclidean distance from the position of the workshop personnel to the future smoke position, and whose direction is from the position of the workshop personnel to the future smoke position. Indicates the relationship between the impact range of smoke and smoke concentration; determines the real-time collected smoke concentration and the future smoke concentration Whether the difference between them exceeds the preset value, if so, the escape route is re-determined.

[0006] The safety production monitoring system of the present invention can monitor the safety of the workshop in real time and can predict the future smoke concentration when a fire occurs. Joint modeling with the repulsion field function model can provide workshop personnel with a more accurate and reliable escape route. Moreover, by judging the difference between the collected smoke concentration and the predicted smoke concentration, the generated escape route is more reliable.

[0007] Furthermore, obtaining a variation sequence of smoke concentration and smoke position includes: obtaining the variation sequence through the highest concentration and the highest concentration position of the smoke area of ​​a plurality of semantically segmented images.

[0008] Furthermore, the change sequence is obtained through the highest concentration and the highest concentration position of the smoke area in multiple semantically segmented images, including: grayscale processing of the smoke area in the semantically segmented image; and obtaining the highest concentration of the smoke area and the position of the highest concentration based on the mapping relationship between the grayscale value of the smoke area and the smoke concentration.

[0009] The present invention maps smoke concentration by gray value, and can conveniently, quickly and intuitively display the maximum value of smoke concentration and the changes in smoke concentration and smoke position.

[0010] Furthermore, the method for acquiring the mapping relationship between the grayscale value of the smoke area and the smoke concentration includes: establishing the mapping relationship between the smoke concentration value collected by the smoke sensor and the grayscale value of the smoke area.

[0011] Furthermore, the loss function of the LSTM model is a mean square error loss function, and is trained using a gradient descent method.

[0012] The LSTM model is used to predict smoke concentration and smoke location. Compared with the RNN model, the LSTM model has better and more accurate prediction effect.

[0013] Furthermore, the future smoke concentration The smoke concentration and smoke position are the smoke concentration and smoke position 20s later.

[0014] Taking into account the time that workshop personnel spend on the escape route, the smoke concentration and smoke location can be predicted after a period of time to provide early warning and evacuation for workshop personnel, ensuring that workshop personnel can evacuate safely and quickly.

[0015] Furthermore, the gravitational field function is: ; In the formula, represents the gravitational field function, Represents a vector whose size is the Euclidean distance from the position of the workshop personnel to the workshop safety exit, and whose direction is from the position of the workshop personnel to the workshop safety exit. is the proportional gain coefficient.

[0016] Furthermore, the digital twin workshop model includes the locations of workshop personnel, the locations of workshop safety exits and workshop personnel passages.

[0017] Through the modeling of the digital twin model, the escape route can be visualized. When an unsafe condition occurs and causes a fire, workshop personnel can evacuate according to the escape route more intuitively, accurately and quickly.

[0018] The technical effect of the present invention is that the real-time monitoring system for safe production of the present invention can monitor the safe production of the workshop in real time, and in case of fire, it can monitor the safety of the workshop according to the future smoke concentration. and smoke location to optimize the escape route, generating a safe and reliable escape route for each workshop personnel. Furthermore, due to the future smoke concentration In order to predict the maximum value of smoke concentration in the future, when the maximum value of smoke concentration in the area reaches a point that seriously affects the escape of workshop personnel, the artificial potential field method can avoid these dangerous areas, thereby ensuring that the generated escape route is safe and reliable. In addition, by judging the difference between the collected smoke concentration and the predicted smoke concentration, the generated escape route is more reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 is a structural diagram schematically showing a real-time monitoring system for safe production based on artificial intelligence according to an embodiment of the present invention; Figure 2 It is a flow chart schematically illustrating a method for implementing a real-time monitoring system for safe production based on artificial intelligence according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0021] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0022] Figure 1 It is a structural diagram schematically showing a real-time monitoring system for safe production based on artificial intelligence according to an embodiment of the present invention.

[0023] like Figure 1 As shown, the real-time monitoring system for safe production based on artificial intelligence of the present invention comprises: a processor, a memory, a smoke sensor and a camera device, wherein the smoke sensor is used to collect the smoke concentration of the workshop, the camera device is used to collect the workshop video, and the memory is used to store computer program instructions. When the computer program instructions are executed by the processor, the following is achieved: Figure 2 The method shown.

[0024] In one embodiment, the memory can be connected to the processor via an address bus, the smoke sensor can be connected to the processor via an RS485 serial port, and the camera device can be connected to the processor via a USB interface. In other optional embodiments, technicians in this field can select the communication method between the processor and other devices according to actual conditions. For example, both the smoke sensor and the camera device can communicate with the processor via Bluetooth.

[0025] Figure 2 It is a flow chart schematically illustrating a method for implementing a real-time monitoring system for safe production based on artificial intelligence according to an embodiment of the present invention.

[0026] In one embodiment, Figure 2 As shown, when the computer program instructions are executed by the processor, the method implemented includes: S101. Establish a digital twin workshop model based on the BIM model.

[0027] BIM model, or building information model, is a digital model that integrates various relevant information of a construction project. The digital twin platform is a visual management platform that can build a highly simulated and intelligently simulated digital world through BIM models, three-dimensional models, tilted shooting and other technologies, and can also display simulation data and measured data through numerical values, three-dimensional views and other methods.

[0028] In one embodiment, in order to establish a digital twin workshop model, at least one smoke sensor and a camera device are installed in different workshops. In this embodiment, the camera device is a camera. The smoke sensor is used to collect smoke concentrations at different locations in the workshop, and the camera is used to record videos of the workshop in real time. Furthermore, the collected smoke concentration and video data are displayed on the digital twin platform through the BIM model. By collecting smoke concentrations and videos at different locations in the workshop in real time, the accuracy and reliability of the monitoring system are improved.

[0029] Specifically, the BIM model first restores the three-dimensional posture of the workshop through the position of the camera and the shooting angle of the camera, and establishes a virtual three-dimensional model of the workshop based on the three-dimensional posture of the workshop. Then, the digital twin platform establishes a visual three-dimensional model of the workshop through the BIM model, and finally displays the input smoke concentration and video data to construct a digital twin workshop model.

[0030] In one embodiment, the digital twin workshop model may include the location of workshop personnel, the location of workshop safety exits, and workshop personnel passages. By establishing a digital twin workshop model, a real-time situation of a real workshop can be simulated, so that the situation of the real workshop can be monitored more intuitively, comprehensively, and conveniently. Furthermore, the digital twin workshop model can also visualize the escape route, allowing workshop personnel to evacuate more intuitively, clearly, and quickly according to the escape route.

[0031] S102, collecting workshop videos, converting the workshop videos into images and inputting them into a trained semantic segmentation model to obtain a change sequence of smoke concentration and smoke position.

[0032] The semantic segmentation model can segment the input image into regional blocks with certain semantic meanings. The collected workshop video is converted into an image so as to obtain the subsequent change sequence of smoke concentration and smoke position. In this embodiment, the method of using the semantic segmentation model is as follows: the collected workshop image is input into the semantic segmentation model, and the semantic segmentation model automatically segments it, and the output result is: the input image is segmented into smoke area and personnel area. Furthermore, the position of the workshop personnel can be determined through the personnel area, and the change sequence of smoke concentration and smoke position can be obtained through the smoke area.

[0033] In one embodiment, the change sequence of smoke density and smoke position can be obtained through the highest density and highest density position of the smoke area after semantic segmentation of the above multiple images. Since the highest density and the highest density position of the image in the video will change with time, the change of the highest density and the highest density position of the multiple images over time constitutes the change sequence of smoke density and smoke position.

[0034] In one embodiment, the change sequence of smoke concentration and smoke position can be expressed as (t1, N1, W1), (t2, N2, W2). Among them, t1 and t2 represent time, N1 and N2 represent the smoke concentration values ​​at time t1 and t2 respectively, and W1 and W2 represent the smoke positions at time t1 and t2 respectively. It should be noted that this embodiment only schematically lists the representation methods of two moments, and the specific number is determined by the captured video. In other optional embodiments, those skilled in the art can choose other methods of representing the change sequence of smoke concentration and smoke position.

[0035] In one embodiment, in order to obtain the highest concentration and the highest concentration position of the smoke area in the semantically segmented image, the smoke area needs to be grayed first to obtain the gray value of the smoke area; wherein, the greater the smoke concentration in the smoke area, the greater the gray value. Secondly, a mapping relationship between the gray value and the smoke concentration is established based on the gray value of the smoke area and the smoke concentration value collected by the smoke sensor; finally, the highest concentration and the position of the highest concentration in the smoke area are obtained based on the mapping relationship between the gray value and the smoke concentration. In this embodiment, since the grayscale value of the smoke area is an image processing parameter and the smoke concentration is a physical quantity, the two cannot be directly operated to calculate the maximum concentration of the smoke area. Therefore, it is necessary to obtain the maximum concentration of the smoke area by establishing a mapping between the grayscale value and the smoke concentration.

[0036] In one embodiment, the mapping relationship between the gray value and the smoke concentration can be established by designing a mapping function to map the smoke concentration to the gray value. Specifically, the mapping relationship between the gray value and the smoke concentration mainly depends on the environment of the workshop, and those skilled in the art can determine the mapping relationship between the gray value and the smoke concentration through experiments. For example, in a similar experimental environment (mainly the lighting conditions are similar to the workshop to be simulated), combustibles are burned, and a video is taken by a camera, and the smoke concentration at each position is detected by a smoke sensor. Based on the above process, a large amount of data on the gray value and the smoke concentration in the video can be obtained. Based on these data, the mapping relationship between the gray value and the smoke concentration can be established by a related method (such as fitting, etc.). In this embodiment, the gray value and the smoke concentration are nonlinear, so they can be mapped by a polynomial function. In other optional embodiments, the mapping function can also be an exponential function or a logarithmic function. Further, in different application environments, the gray value and the smoke concentration may present a linear mapping or a segmented mapping, and those skilled in the art can select a suitable mapping function to determine the mapping relationship between the gray value and the smoke concentration according to different needs.

[0037] S103: Input the smoke density and smoke position change sequence into a trained LSTM model to predict the future smoke density. and smoke location.

[0038] The LSTM model is a long short-term memory network model, which is used to process and learn time series data, such as speech recognition, time series prediction, etc. LSTM solves the problems of explosion and gradient disappearance that may be encountered when training traditional RNN (recurrent neural network), so the LSTM model has higher prediction accuracy and better effect. In this embodiment, the LSTM model is used to predict changes in smoke concentration and smoke position.

[0039] The training process of the LSTM model in this embodiment is: a large number of smoke concentration and smoke position change sequences are obtained and input into the LSTM model, at which time the LSTM model continuously adjusts parameters according to the input data, and when the preset maximum number of training times is reached or the loss is less than the preset threshold, the LSTM model ends the training. Among them, the loss function of the LSTM model is the mean square error loss, and the training method is the descent gradient method.

[0040] In one embodiment, the future smoke density is obtained The method for calculating the smoke concentration and smoke position includes: firstly, inputting the smoke concentration and smoke position of the previous moments in the change sequence of smoke concentration and smoke position into the LSTM model, the LSTM model predicts them, outputs the smoke concentration and smoke position of the next moment, then inputting the predicted smoke concentration and smoke position of the next moment into the LSTM model for rolling prediction, and finally obtaining the future smoke concentration. and smoke location. In this example, the future smoke density and smoke position are the smoke concentration and smoke position after 20 seconds. In other optional embodiments, the future smoke concentration The smoke concentration and smoke position may also be the smoke concentration and smoke position after 30 seconds, which can be selected by those skilled in the art according to actual needs.

[0041] It should be noted that in order to ensure the safe evacuation of personnel in the event of an accident, the smoke concentration in the future To predict the maximum value of the future smoke concentration. Among them, select the future smoke concentration The reason for predicting the maximum value of smoke concentration in the future is that when the maximum value of smoke concentration in the area reaches a level that seriously affects the escape of workshop personnel (for example, visibility is less than 4m, etc.), these dangerous areas can be avoided by calculating the repulsive field function and the gravitational field function, thereby ensuring that the generated escape route is safe and reliable.

[0042] By predicting future smog concentrations The smoke location can be detected and early warning can be given to workshop personnel, allowing them to obtain escape routes before they are in danger and evacuate quickly and safely according to the escape routes, thus ensuring the safety of workshop personnel. In addition, by timely warning before a fire occurs and before damage is caused, equipment can be shut down in time, thus avoiding serious damage and losses caused by fire in the workshop.

[0043] S104. Calculate an escape route in the digital twin workshop model by superimposing a gravitational field function on a repulsive field function.

[0044] In this embodiment, the search algorithm used to generate the escape route is the artificial potential field method, which is a local path planning algorithm. Specifically, the artificial potential field method first constructs a repulsive potential field around the workshop where the unsafe state is generated, and constructs a gravitational potential field around the safe exit of the workshop. At this time, the workshop personnel are in a composite field composed of these two potential fields and are affected by repulsion and gravitation. Then, under the guidance of the combined force of repulsion and gravitation, they evacuate according to the searched safe and obstacle-avoiding escape route.

[0045] Specifically, in this embodiment, firstly according to the future smoke density The gravitational field function and repulsive field function in the artificial potential field method are determined based on the smoke position and the digital twin workshop model. Then the gravitational field function is superimposed on the repulsive field function to obtain the artificial potential field function. Finally, a safe and reliable escape route is generated for each workshop personnel through the artificial potential field function. Specifically: First, calculate the gravitational field function, which is expressed as follows: ; in, represents the gravitational field function, Represents a vector whose size is the Euclidean distance from the position of the workshop personnel to the workshop safety exit, and whose direction is from the position of the workshop personnel to the workshop safety exit. is the proportional gain coefficient.

[0046] From the above expression of the gravitational potential field function, it can be seen that the gravitational potential field function is mainly related to the distance between the workshop personnel and the workshop safety exit. Specifically, the farther the workshop personnel are from the safety exit, the greater the gravitational force they are subject to; conversely, the closer the workshop personnel are to the safety exit, the smaller the gravitational force they are subject to.

[0047] Then, the repulsive field function is calculated, and its expression is as follows: ; in, represents the repulsive field function, k represents the positive proportionality coefficient, is a vector whose size is the Euclidean distance from the position of the workshop personnel to the predicted future smoke position, and whose direction is from the position of the workshop personnel to the predicted future smoke position. Indicates the relationship between the impact range of smoke and smoke concentration. is the predicted future smoke density.

[0048] about The greater the smoke concentration, the greater the impact range of the smoke. When the smoke concentration is below a certain level, it has no effect on the human body in a short period of time, and the impact range of the smoke is zero. In this embodiment, in order to facilitate simulation, the PM2.5 of the smoke concentration is set to 150 (critical concentration), and the impact range of the smoke is 0. When the PM2.5 concentration of the smoke reaches 150, the impact range of the smoke is The distance between the critical concentration of smoke and the maximum concentration of smoke is 0.01 meters. When the PM2.5 concentration of smoke reaches the maximum (the smoke concentration value corresponding to the maximum gray value), the influence range of smoke is 1.2 meters. From the critical concentration of smoke to the maximum concentration of smoke, The corresponding smoke influence range is set to linear change.

[0049] In one embodiment, information such as the materials and combustibles of the actual building can also be considered to assist in determining For example, if the fire is dangerous and the damage caused by the burning of combustibles is great, then The value of should be set larger, otherwise it can be set smaller. The value can be set according to actual situation.

[0050] In one embodiment, it is also possible to design You can even set the size according to different sizes. , get different escape routes. For example Corresponding to escape route 1, Corresponding to escape route 2, if Greater than , then obviously The corresponding escape route 1 is safer, so different colors or safety factors can be marked on these different escape routes (in this sense, In fact, it is also a safety factor) for users to choose freely.

[0051] Furthermore, according to the expression of the above repulsive field function, it can be known that the workshop personnel are not always affected by the smoke concentration. Specifically, on the one hand, when the distance between the workshop personnel and the smoke position is greater than the influence range of the smoke, the smoke has no effect on the escape route of the workshop personnel, so the repulsive field function at this time is 0. On the other hand, when the relative distance between the workshop personnel and the smoke position is less than the influence range of the smoke, the smoke begins to affect the escape route of the workshop personnel. Specifically, the smaller the relative distance between the workshop personnel and the smoke position, the greater the influence of the repulsive force, that is, the greater the influence of the smoke on the escape route of the workshop personnel.

[0052] In one embodiment, when the smoke concentration in an area is too high and the visibility in the workshop is less than 4m, the area is no longer suitable for the evacuation of workshop personnel, so the area should not be considered in the generation of the escape route. Therefore, the repulsive force of the area can be set specifically so that the escape route generated by the artificial potential field method can avoid the area, thereby making the generated escape route safer.

[0053] Furthermore, an artificial potential field function can be obtained by adding the above-mentioned gravitational field function and repulsive field function, and a safe and reliable escape route can be found for workshop personnel by using a gradient descent algorithm on the artificial potential field function.

[0054] S105: Determine the smoke concentration collected in real time and the future smoke concentration Whether the difference between them exceeds the preset value, if so, the escape route is re-determined.

[0055] In one embodiment, the preset value is set to 10. It is understandable that if the smoke concentration collected by the smoke sensor in real time is different from the predicted future smoke concentration If the difference between the predicted smoke concentration and the actual smoke concentration exceeds 10, it means that the predicted smoke concentration is significantly different from the actual smoke concentration, which will result in the predicted smoke concentration in the future being The generated escape route may not be accurate, so it is necessary to return to S102 to ensure that the escape route obtained by the workshop personnel is accurate and reliable.

[0056] By judging the real-time smoke concentration and the future smoke concentration Whether the difference between them exceeds a preset value ensures that the escape route obtained by the workshop personnel is accurate and reliable, thereby improving the accuracy and safety of the escape route obtained by the monitoring system of the present invention.

[0057] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0058] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A real-time monitoring system for safe production based on artificial intelligence, characterized in that: include: A processor, a memory, a smoke sensor and a camera device, wherein the smoke sensor is used to collect smoke concentration in a workshop, the camera device is used to collect workshop video, and the memory is used to store computer program instructions. When the computer program instructions are executed by the processor, the following method is implemented: Establish a digital twin workshop model based on the BIM model; Collect workshop videos, convert them into images and input them into a trained semantic segmentation model to obtain a change sequence of smoke concentration and smoke position; The change sequence of smoke concentration and smoke position is input into a trained LSTM model to predict the future smoke concentration and smoke location; The escape route is calculated in the digital twin workshop model by superimposing the gravitational field function with the repulsive field function, and the repulsive field function is as follows: ； In the formula, represents the repulsive field function, k represents the positive proportionality coefficient, represents a vector, the size of which is the Euclidean distance from the position of the workshop personnel to the future smoke position, and the direction of which is from the position of the workshop personnel to the future smoke position. Indicates the relationship between the impact range of smoke and smoke concentration; Determine the real-time smoke concentration and the future smoke concentration Whether the difference between them exceeds the preset value, if so, the escape route is re-determined.

2. In the real-time monitoring system for safe production based on artificial intelligence as claimed in claim 1, it is characterized in that: Get the change sequence of smoke density and smoke position, including: The change sequence is obtained through the highest density and the highest density position of the smoke area in the multiple semantically segmented images.

3. In the real-time monitoring system for safe production based on artificial intelligence as claimed in claim 2, it is characterized in that: The variation sequence is obtained through the highest density and the highest density position of the smoke region of the multiple semantically segmented images, including: graying the smoke region of the semantically segmented image; The highest concentration of the smoke area and the position of the highest concentration are obtained according to the mapping relationship between the gray value of the smoke area and the smoke concentration.

4. In the real-time monitoring system for safe production based on artificial intelligence as claimed in claim 3, it is characterized in that: The method for obtaining the mapping relationship between the gray value of the smoke area and the smoke concentration includes: The mapping relationship is established between the smoke concentration value collected by the smoke sensor and the grayscale value of the smoke area.

5. In the real-time monitoring system for safe production based on artificial intelligence as claimed in claim 1, it is characterized in that: The loss function of the LSTM model is the mean square error loss function, and the gradient descent method is used for training.

6. The real-time monitoring system for safe production based on artificial intelligence as claimed in claim 1, characterized in that: Future smog concentrations The smoke concentration and smoke position are the smoke concentration and smoke position 20s later.

7. The real-time monitoring system for safe production based on artificial intelligence as claimed in claim 1, characterized in that: The gravitational field function is: ； In the formula, represents the gravitational field function, Represents a vector whose size is the Euclidean distance from the position of the workshop personnel to the workshop safety exit, and whose direction is from the position of the workshop personnel to the workshop safety exit. is the proportional gain coefficient.

8. The real-time monitoring system for safe production based on artificial intelligence as claimed in claim 1, characterized in that: The digital twin workshop model includes the location of workshop personnel, the location of workshop safety exits and workshop personnel passages.

Images