A multi-spectral vision and AI-based early fire disaster intelligent identification, immediate blocking and multi-point linkage explosion control system for smokeless powder

Abstract

The application relates to a pyrotechnic composition early fire intelligent identification, immediate blocking and multi-point linkage explosion control system based on multispectral vision and AI. The system comprises a multi-dimensional perception fusion module, an edge AI inference engine, a double-mechanism transient blocking executor and a multi-point space damping linkage network. The multi-dimensional perception fusion module collects spectrum, heat radiation and air pressure disturbance information in a pyrotechnic composition production area through a multispectral imaging sensor, an ultraviolet pulse photoelectric detector and an ultrasonic air pressure disturbance sensor. The edge AI inference engine identifies early abnormal states of the pyrotechnic composition through a heterogeneous feature fusion algorithm and a time sequence analysis model. The double-mechanism transient blocking executor cuts off a dangerous propagation path through a pneumatic cutting mechanism, and releases a fire extinguishing and inhibiting medium through an aerosol generator. The multi-point space damping linkage network controls an explosion-proof valve and an inert gas barrier, and realizes cooperative protection. The application has the effects of improving the pyrotechnic composition early fire identification speed, realizing rapid fire risk blocking and regional linkage explosion control.

Description

Technical Field

[0001] This application relates to the field of safety in the production of fireworks and firecrackers and civilian explosives, and in particular to a multi-spectral vision and AI-based intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics. Background Technology

[0002] The production of fireworks and firecrackers and civilian explosives is a typical high-risk industry. Its production, processing, storage and transportation involve a large amount of flammable and explosive materials such as pyrotechnic powder, black powder, oxidizers, reducing agents and fuses. Due to the high energy density, fast burning speed and strong thermal sensitivity of these materials, they are prone to local thermal decomposition, smoldering, combustion or even deflagration accidents when affected by factors such as mechanical friction, electrostatic discharge, abnormal temperature rise, impact, equipment failure or operation error during the production process. Once the flame or high temperature area comes into contact with the pyrotechnic powder, the combustion process may rapidly develop from the initial combustion stage to the deflagration state in a very short time, and may even further develop into detonation propagation, posing a serious threat to production equipment, personnel safety and the surrounding environment.

[0003] Existing fireworks and firecracker manufacturers typically use traditional fire monitoring equipment to protect production areas, such as smoke detectors, infrared flame detectors, temperature sensors, and manual inspections. However, due to the characteristics of the pyrotechnics production environment, such as high dust concentration, complex lighting changes, significant electromagnetic interference, and frequent static electricity accumulation, traditional single-sensor detection methods have obvious shortcomings. Smoke sensors require a certain concentration of smoke particles to be produced during combustion before triggering an alarm, making it difficult to respond promptly to trace amounts of abnormal gases and weak thermal radiation generated during the early thermal decomposition stage of pyrotechnics. Infrared flame detectors mainly rely on the spectral characteristics of the flame after its formation, and usually require the presence of a visible flame to detect it, which cannot meet the safety requirements of "early detection, early isolation, and early control" in the pyrotechnics production process.

[0004] Meanwhile, existing detection equipment is easily affected by environmental factors in complex industrial environments. Dust particles generated during production may cause optical sensors to be falsely triggered, and dust accumulation on the equipment surface will reduce detection sensitivity. The transient signals generated by electrostatic discharge are similar to real fire signals, which can easily cause false alarms. In addition, most traditional fire alarm systems use fixed threshold judgment methods and lack the ability to comprehensively analyze the correlation between multiple physical signals, making it impossible to effectively identify the early dangerous characteristics of pyrotechnics during the process of heat accumulation to chemical reaction.

[0005] In terms of fire response, fireworks and firecrackers and civilian explosives production sites currently rely mainly on manual operation of fire extinguishers, fire water systems, or fixed fire extinguishing facilities for accident handling. When a fire occurs in dangerous processes such as mixing, loading, and transporting pyrotechnics, due to the rapid spread of combustion, high on-site temperature, and the accompanying toxic fumes and explosion risks, it is often difficult for staff to enter the accident area in time for manual handling. Traditional fire extinguishing methods that rely on personnel response have problems such as long response time, high safety risks, and inability to block the path of danger. At the same time, existing fire extinguishing systems are usually designed for the fire fighting process after a fire occurs, lacking an active intervention mechanism for the "pre-ignition stage" of pyrotechnics, making it difficult to isolate the hazard source before deflagration occurs.

[0006] In addition, existing production line safety protection systems typically adopt a single-equipment, single-area control mode, lacking effective information sharing and linkage control between production stations. When an initial combustion or deflagration accident occurs at a certain station, flames, heat, shock waves, and burning particles may spread to adjacent areas along the material conveying path or spatial channel, leading to continuous combustion or even multi-point sympathetic explosion accidents. Therefore, relying solely on single-point alarms or local fire extinguishing measures is insufficient to meet the requirements of modern pyrotechnics production processes for integrated, intelligent, and inherently safe protection.

[0007] With the development of artificial intelligence, multispectral sensing, edge computing and industrial Internet of Things technologies, the development direction of safety protection technology in the field of hazardous production is to integrate multi-type sensor data, use intelligent algorithms to predict dangerous conditions, and combine rapid actuators to achieve automatic blocking. However, existing technologies still lack an intelligent explosion control system that can identify early weak abnormal signals using multi-dimensional sensing methods based on the special combustion mechanism of pyrotechnics and automatically perform physical isolation, chemical suppression and multi-area collaborative protection in a very short time.

[0008] In response to the aforementioned technologies, the inventors believe it is necessary to propose an intelligent early-stage fire identification system for pyrotechnics based on multispectral vision and artificial intelligence analysis technology. By fusing and analyzing multidimensional information such as ultraviolet, visible light, infrared, and air pressure disturbances, the system can accurately identify the nascent stage of a fire. Combined with a high-speed blocking actuator and a multi-point linkage protection network, it can automatically isolate and suppress the fire before the danger spreads, thereby improving the safety and accident prevention capabilities of the pyrotechnics production process. Summary of the Invention

[0009] To address the shortcomings of the aforementioned technologies in accurately identifying early-stage fires and incorporating high-speed blocking actuators and multi-point linkage protection networks to automatically isolate and suppress hazards before they spread, this application provides a multispectral vision and AI-based intelligent identification, immediate blocking, and multi-point linkage explosion control system for early-stage pyrotechnic fires.

[0010] The technical solution of the intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics based on multispectral vision and AI provided in this application is as follows: A smart early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics based on multispectral vision and AI includes: a multi-dimensional perception fusion module, an edge AI inference engine, a dual-mechanism transient blocking actuator, and a multi-point spatial damping linkage network. The multi-dimensional sensing fusion module is set in the explosion-proof area of ​​the pyrotechnics production station and is used to collect multi-dimensional physical field information during the pyrotechnics processing, including spectral radiation information, ultraviolet pulse information and air pressure disturbance information. The edge AI inference engine is communicatively connected to the multi-dimensional perception fusion module, and is used to perform heterogeneous feature fusion processing on the multi-dimensional physical field information, and to identify the early abnormal state of pyrotechnic thermal decomposition based on the artificial intelligence inference model, and output a hazard level judgment signal. The dual-mechanism transient blocking actuator is connected to the edge AI inference engine and is used to perform physical blocking action on the propagation path of pyrotechnics and gunpowder after receiving the hazard level determination signal, and simultaneously release the fire extinguishing and suppression medium. The multi-point spatial damping linkage network is connected to multiple pyrotechnic production stations and is used to control the operation of the explosion-proof mechanism and inert gas barrier mechanism in adjacent areas according to the hazard level determination results, so as to form a multi-level spatial protection barrier.

[0011] By adopting the above technical solutions, the system achieves early perception, intelligent identification, rapid blocking, and regional collaborative explosion suppression of fire hazards during the production of pyrotechnics. The system uses a multi-dimensional perception fusion module to simultaneously acquire abnormal characteristic information such as spectral radiation, ultraviolet pulses, and air pressure disturbances, enabling it to identify early hazardous states such as thermal decomposition of pyrotechnics, electrostatic discharge, and localized abnormal temperature rise before open flame formation and deflagration occur. An edge AI inference engine is used to fuse and analyze multi-source heterogeneous data, improving the accuracy and response speed of hazardous state identification. A dual-mechanism transient blocking actuator enables physical isolation of the delivery path and simultaneous release of the extinguishing and suppression medium, rapidly reducing the risk of accident spread. Simultaneously, a multi-point spatial damping linkage network is combined to achieve explosion isolation, inerting, and coordinated protection of adjacent production areas, forming an intelligent explosion control system that transforms from single-point control to regional collaborative prevention and control, thereby improving the safety of the pyrotechnics production process and the ability to respond to accidents.

[0012] Optionally, the multi-dimensional sensing fusion module includes: a multispectral imaging sensor, an ultraviolet pulse photodetector, an ultrasonic pressure disturbance sensor, and a protective cover; The multispectral imaging sensor covers a detection band of 200nm to 14μm and is used to collect ultraviolet radiation, visible light changes and infrared thermal radiation information during the thermal decomposition of pyrotechnics. The ultraviolet pulse photodetector is used to detect ultraviolet pulse signals generated by electrostatic discharge, electric sparks, and the oxidation reaction of pyrotechnics. The ultrasonic pressure disturbance sensor is used to detect microscale pressure disturbance changes generated during the thermal decomposition of pyrotechnics.

[0013] By adopting the above technical solutions, a multi-dimensional and highly sensitive real-time perception function for the hazardous state of pyrotechnics is realized. The multispectral imaging sensor, covering the ultraviolet, visible, and infrared bands, synchronously collects abnormal radiation, temperature rise changes, and optical characteristics generated during the thermal decomposition of pyrotechnics, enabling early detection of potential ignition sources and thermal runaway states. The ultraviolet pulse photodetector can quickly capture transient ultraviolet signals generated by electrostatic discharge, electric sparks, and oxidation reactions, improving the detection capability of ignition causes. The ultrasonic pressure disturbance sensor can sense microscale pressure fluctuations, enabling dynamic monitoring of pyrotechnics decomposition, local combustion, and pre-deflagration states. Through the collaborative fusion of multiple sensors, the ability to extract hazardous features in complex production environments is improved, reducing the risk of false alarms and missed alarms from single sensors, and providing a reliable data foundation for subsequent AI intelligent identification and rapid explosion control.

[0014] Optionally, the protective cover adopts a dustproof and anti-static transparent structure, and the multispectral imaging sensor, ultraviolet pulse photodetector and ultrasonic pressure disturbance sensor are fixed inside the protective cover by sensor mounting brackets; The multispectral imaging sensor is equipped with an ultraviolet detection channel, a visible light detection channel, and an infrared detection channel to achieve synchronous acquisition across multiple frequency bands.

[0015] By adopting the above technical solutions, the safe installation and multi-band synchronous acquisition function of multi-dimensional sensing equipment in hazardous environments of pyrotechnics and gunpowder are realized. The dustproof and anti-static transparent protective cover can isolate dust, gunpowder particles and electrostatic interference generated during the production process, while ensuring the effective transmission of optical and acoustic signals, improving the reliability and service life of the sensor in high-risk environments. The multispectral imaging sensor, ultraviolet pulse photodetector and ultrasonic pressure disturbance sensor are integrated and fixed by the sensor mounting bracket, realizing the spatial consistency of the multi-source detection unit and reducing acquisition errors. The multispectral imaging sensor acquires information of different bands simultaneously through ultraviolet, visible and infrared detection channels, realizing the comprehensive identification of thermal radiation, optical changes and abnormal discharge characteristics of pyrotechnics and gunpowder, providing more complete and accurate feature data for subsequent AI fusion analysis.

[0016] Optionally, the edge AI inference engine includes a preprocessing board, an FPGA synchronous processing chip, an AI edge computing motherboard, and an AI SoC processing chip; The preprocessing board is used to filter, synchronize time, and normalize features of the data collected by the multidimensional perception fusion module. The FPGA synchronous processing chip is used to realize high-speed parallel processing of multi-sensor data; The AI ​​SoC processing chip runs a heterogeneous feature fusion module and a temporal convolutional network module to predict the risk of pyrotechnic explosion based on continuous time series data.

[0017] By adopting the above technical solutions, high-speed intelligent analysis and real-time risk prediction of the hazardous state of pyrotechnics are achieved. The edge AI inference engine filters, synchronizes time, and normalizes features of multi-dimensional sensing data through a preprocessing board, improving the consistency and effectiveness of multi-source data. The FPGA synchronous processing chip enables high-speed parallel computing of multi-sensor data, reducing data processing latency and improving the system's real-time response capability. The AI ​​edge computing motherboard and AI SoC processing chip run a heterogeneous feature fusion module and time series analysis model to deeply fuse multi-dimensional features such as spectrum, ultraviolet pulse, and air pressure disturbance, enabling intelligent prediction of pyrotechnic thermal decomposition, abnormal combustion, and explosion trends. This allows for rapid inference locally, reducing data transmission dependence, improving the accuracy of hazard identification and early warning capabilities, and providing a reliable decision-making basis for subsequent automatic blocking and coordinated explosion control.

[0018] Optionally, the heterogeneous feature fusion module is used to map multispectral image features, ultraviolet pulse features, and air pressure disturbance features to a unified feature space to form a multidimensional fused feature vector; The temporal convolutional network module is used to analyze the changing trend of the fused feature vector within a continuous sampling period and output an AI hazard level determination signal.

[0019] By adopting the above technical solutions, the system achieves deep fusion of multi-source data and intelligent dynamic risk assessment of abnormal states of pyrotechnics. The heterogeneous feature fusion module uniformly encodes and maps the thermal radiation features, ultraviolet pulse signal features, and air pressure disturbance features in multispectral images into feature spaces, eliminating scale differences between data from different sensors and improving the comprehensive expression of dangerous features in complex environments. The temporal convolutional network module further performs temporal analysis on the changing patterns of the fused feature vectors within continuous sampling periods, enabling it to identify dynamic evolution trends such as temperature rise, decomposition, discharge, and combustion precursors of pyrotechnics. Based on the degree of risk change, it outputs corresponding AI hazard level assessment signals, improving the system's accuracy and predictive ability for early fire hazards, and realizing a shift from traditional post-event alarms to an intelligent safety management model of early prediction and proactive prevention.

[0020] Optionally, the dual-mechanism transient blocking actuator includes a high-pressure pneumatic blasting and cutting device and a high-expansion microencapsulated aerosol generator; The high-pressure pneumatic blasting and cutting device is installed on the drug delivery pipeline and includes a high-pressure gas cylinder, a drive piston, an annular cutting tool, and a tool guide groove. When a blocking control signal is received, the drive piston pushes the annular cutting tool to move along the tool guide groove to cut off the drug delivery pipeline.

[0021] By adopting the above technical solution, the functions of rapid physical isolation and immediate cut-off of accident propagation paths during the delivery of pyrotechnics and explosives are realized. The dual-mechanism transient blocking actuator is linked with the control system through a high-pressure pneumatic blasting and cutting device. After receiving a danger blocking signal, it uses a high-pressure gas cylinder to release transient driving force, which drives the driving piston to move the annular cutting tool at high speed along the tool guide groove, thereby quickly cutting off the delivery pipeline and blocking the continued delivery of pyrotechnics and explosives from the source. This reduces the risk of flame, heat and deflagration energy spreading along the pipeline. The annular cutting structure can form a circumferential synchronous cut on the pipeline, improving the cutting reliability and response speed. At the same time, it avoids the danger expansion caused by the delay in closing traditional valves, providing an effective time window for subsequent fire suppression and regional linkage explosion control, and improving the inherent safety protection capability of pyrotechnics and explosives production equipment.

[0022] Optionally, the high-expansion microencapsulated aerosol generator includes a fire extinguishing medium storage chamber, a sodium bicarbonate microencapsulated dry powder layer, a fluorinated ketone fire extinguishing medium layer, a high-pressure nitrogen drive cylinder, a jet guide pipe, and a conical jet head. The high-expansion microencapsulated aerosol generator is used to generate an aerosol suppression zone covering the combustion area, which suppresses the combustion chain reaction through physical coverage and chemical free radical capture.

[0023] By adopting the above technical solution, rapid fire suppression and local space flame-retardant protection functions are achieved during the early combustion process of pyrotechnics. The high-expansion microencapsulated aerosol generator provides transient injection power through a high-pressure nitrogen drive cylinder, which rapidly releases the sodium bicarbonate microencapsulated dry powder layer and fluorinated ketone fire extinguishing medium layer in the fire extinguishing medium storage chamber. The aerosol is then dispersed into a uniformly diffused aerosol suppression area through the injection guide pipe and conical injection head. The released microencapsulated fire extinguishing medium can cover the combustion area, reduce the degree of oxygen participation, and inhibit the combustion chain reaction through the capture of chemical free radicals, rapidly reducing the flame spread rate and heat release intensity. This improves the fire extinguishing response speed and medium coverage efficiency in hazardous spaces, and can achieve active suppression before the pyrotechnics undergo violent combustion and explosion, providing effective protection for subsequent explosion-proof linkage and accident control.

[0024] Optionally, the multi-point spatial damping linkage network includes an explosion-proof valve, an inert gas storage tank, an inert gas delivery pipeline, and an air curtain forming structure. The explosion-proof valve is installed between adjacent production stations to block the propagation of explosion shock waves; The inert gas storage tank delivers inert gas to the air curtain injection position through an inert gas delivery pipeline to form an inert gas isolation air curtain. The multi-point spatial damping linkage network uses industrial communication links to connect multiple production stations. When any production station generates a danger signal, explosion-proof control commands are simultaneously sent to adjacent production stations via hard-wired emergency stop lines and network control.

[0025] By adopting the above technical solutions, the multi-station collaborative explosion-proof and spatial cascade protection functions of the pyrotechnics production area are realized. The multi-point spatial damping linkage network, by setting explosion-proof valves between adjacent production stations, achieves rapid blocking of the propagation path of explosion shock waves, flames, and high-temperature airflows. By utilizing inert gas storage tanks, delivery pipelines, and gas curtain structures, an inert isolation barrier is formed around the hazardous area, reducing the concentration of flammable media and suppressing combustion and deflagration. At the same time, through industrial communication links combined with hard-wired emergency stop control, rapid information exchange and synchronous response between multiple production stations are realized. When any station detects a dangerous state, the relevant explosion-proof mechanism can be shut down and the gas curtain protection can be activated in a coordinated manner to prevent a single-point accident from spreading to the surrounding area, thereby improving the overall regional explosion-proof capability and accident control level of the production line.

[0026] Optionally, the system further includes a timing control module for recording and controlling the time relationship between fire identification, physical cutting, aerosol release, and area explosion-proof actions; These include: fire detection signals, AI hazard level determination signals, cutting execution trigger signals, aerosol spray trigger signals, explosion-proof linkage control signals, and full-area explosion suppression control signals; Through the timing coordination of the above signals, the system can complete identification, blocking, and area protection in the early stages of a fire.

[0027] By adopting the above technical solutions, rapid response, precise timing control, and multi-stage collaborative protection functions are achieved in the handling of pyrotechnic fire accidents. The timing control module uniformly schedules and controls key actions such as fire identification, AI risk assessment, physical cutting, aerosol release, and area explosion isolation, enabling each actuator to respond quickly and orderly according to preset logic, avoiding the expansion of danger due to delays or asynchrony. By recording and coordinating fire identification signals, hazard level judgment signals, and trigger signals at all levels in real time, a continuous control process can be achieved from early anomaly detection, transport path cutoff, local fire suppression to full-area linkage protection, improving the system's emergency response efficiency and reliability, and realizing the transformation from passive alarm to active prevention, rapid interruption, and collaborative explosion control in the pyrotechnic production environment.

[0028] Optionally, the edge AI inference engine establishes a danger index model based on the continuous sampling results of the multi-dimensional perception fusion module, and outputs the following signals according to the danger index level: Level 1 warning signal, Level 2 blocking signal, and Level 3 full-domain linkage explosion suppression signal; Among them: the Level 1 early warning signal is used to activate the monitoring enhancement mode; The secondary blocking signal is used to activate the dual-mechanism transient blocking actuator; The Level 3 global linkage explosion suppression signal is used to activate the multi-point spatial damping linkage network. The system adopts a modular explosion-proof structure, with each module connected via an industrial communication interface. It is installed in hazardous production areas such as pyrotechnics mixing, filling, conveying, and packaging areas to achieve automatic sensing, intelligent judgment, rapid interruption, and regional coordinated explosion control during the pyrotechnics production process. By adopting the above technical solutions, intelligent graded risk assessment and automated safety control functions are realized in the pyrotechnics production process. The edge AI inference engine establishes a hazard index model based on multi-source data such as spectrum, ultraviolet light, and air pressure continuously collected by the multi-dimensional perception fusion module. It can dynamically quantify and analyze the risks of thermal decomposition, abnormal heat release, and combustion and explosion of pyrotechnics, and output control signals of different levels according to the degree of danger. The first-level early warning signal can activate the enhanced monitoring mode to realize continuous tracking of potential risks; the second-level blocking signal can quickly trigger the dual-mechanism transient blocking actuator to realize the isolation and suppression of hazard sources; the third-level full-domain linkage explosion suppression signal can drive the multi-point spatial damping linkage network to realize the overall protection of the production area. Through modular explosion-proof structure and industrial communication interface connection, the system can be flexibly deployed, rapidly expanded, and multi-area collaborative control can be realized, improving the automatic perception, intelligent decision-making, rapid response, and overall safety protection capabilities in the pyrotechnics production process.

[0029] In summary, this application includes at least one of the following beneficial technical effects: 1. By setting up a multi-dimensional perception fusion module, using multispectral imaging sensors, ultraviolet pulse photodetectors and ultrasonic pressure disturbance sensors, multi-dimensional information such as spectral radiation, electrostatic discharge, thermal anomalies and microscale pressure disturbances in the pyrotechnics production process is collected simultaneously, so as to achieve high sensitivity to early dangerous states such as thermal decomposition and abnormal combustion of pyrotechnics, and improve the timeliness and accuracy of fire hazard detection. 2. By setting up an edge AI inference engine, and using a heterogeneous feature fusion module and a temporal convolutional network module to perform fusion analysis and trend prediction on multi-source sensor data, it can identify the risk of pyrotechnic explosion based on continuous time series changes, realize the transformation from traditional passive alarm to active prediction and early warning, and improve the ability to identify dangerous conditions and the level of intelligence. 3. By setting up a dual-mechanism transient blocking actuator, the high-pressure pneumatic blasting and cutting device is combined with the high-expansion microencapsulated aerosol generator. In the event of a dangerous situation, the propagation path of the pyrotechnics can be quickly cut off, and the extinguishing and suppressing medium can be released simultaneously to achieve the synergistic effect of physical isolation and chemical suppression, thereby reducing the risk of flame spread and deflagration escalation. 4. By setting up a multi-point spatial damping linkage network, using explosion-proof valves, inert gas isolation curtains and industrial communication links, information sharing and linkage protection between multiple production stations can be realized. When a dangerous situation occurs in a local area, explosion-proof and inerting measures in the adjacent area can be activated simultaneously to prevent the accident from spreading to the surrounding area and improve the overall explosion-proof capability of the production line. 5. By setting up a timing control module, fire identification, hazard level judgment, blocking execution, aerosol release, and area explosion suppression actions are uniformly coordinated and controlled, enabling rapid and orderly response among various safety execution mechanisms, shortening accident handling time, and improving the system's emergency control efficiency; 6. By adopting a modular explosion-proof structure design, the multi-dimensional sensing, AI analysis, transient interruption, and linkage explosion control modules can be flexibly configured and expanded according to different hazardous processes such as pyrotechnic mixing, filling, conveying, and packaging, thereby improving the system's adaptability and engineering application value. This provides an integrated safety protection solution for the pyrotechnic production process, featuring automatic sensing, intelligent judgment, rapid interruption, and regional collaborative explosion control. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the overall system architecture of the present invention.

[0031] Figure 2 This is a hardware connection block diagram for the multidimensional perception and edge AI inference of the present invention.

[0032] Figure 3 This is a cross-sectional view of the dual-mechanism transient blocking actuator structure of the present invention.

[0033] Figure 4 This is a waveform diagram of the multi-point linkage timing operation of the present invention.

[0034] Figure labeling: 1. Multi-dimensional perception fusion module; 2. Edge AI inference engine; 3. Dual-mechanism transient interruption actuator; 4. Multi-point spatial damping linkage network; 5. Explosion-proof wall; 6. Mixing station; 7. Loading station; 8. Packaging station; 9. Production station isolation area; 10. Drug delivery pipeline; 11. Mixing device; 12. Conveying mechanism; 13. Explosion-proof control cabinet; 14. Multispectral imaging sensor; 15. Ultraviolet pulse photodetector; 16. Ultrasonic pressure disturbance sensor; 17. Protective cover; 18. Sensor mounting bracket; 19. Ultraviolet detection channel; 20. Visible light detection channel; 21. Infrared detection channel; 22. Preprocessing board; 23. FPGA synchronous processing chip; 24. High-speed data cable; 25. AI edge computing motherboard; 26. AI SoC processing chip; 27. Heterogeneous feature fusion module; 28. Temporal convolutional network module; 29. ​​Communication interface; 30. Blocking control output terminal; 31. Multi-point linkage network interface; 32. High-pressure pneumatic blasting and cutting device; 33. High-pressure gas cylinder; 34. Drive piston; 35. Annular cutting tool; 36. Tool guide groove; 37. Pipe isolation cavity; 38. High-expansion microencapsulated aerosol generator; 39. Fire extinguishing medium storage cavity; 40. Sodium bicarbonate microencapsulated dry powder layer; 4 1. Fluorinated ketone extinguishing medium layer; 42. High-pressure nitrogen driving cylinder; 43. Spray guide pipe; 44. Conical spray head; 45. Aerosol suppression zone; 46. Explosion-proof valve; 48. Inert gas storage tank; 50. Inert gas isolation curtain; 51. Industrial communication link; 53. Fire identification signal; 54. AI hazard level judgment signal; 55. Cutting execution trigger signal; 56. Aerosol spray trigger signal; 57. Explosion-proof linkage control signal; 58. Global explosion suppression control signal. Detailed Implementation

[0035] The following is in conjunction with the appendix Figure 1-4 This application will be described in further detail.

[0036] This application discloses a multispectral vision and AI-based intelligent early-stage fire identification, immediate containment, and multi-point coordinated explosion control system for pyrotechnics. For example... Figures 1-4 As shown in the embodiment of this application, a system for intelligent identification, immediate blocking, and multi-point linkage explosion control of early fires in pyrotechnics based on multispectral vision and AI is provided. This system is applied to high-risk process areas such as mixing, filling, conveying, and packaging in the production process of pyrotechnics, and is used to perceive, intelligently identify, and actively control the risks of thermal decomposition, electrostatic discharge, local combustion, and deflagration that may occur during the production process of pyrotechnics.

[0037] The system in this embodiment includes a multi-dimensional perception fusion module 1, an edge AI inference engine 2, a dual-mechanism transient interruption actuator 3, and a multi-point spatial damping linkage network 4. The multi-dimensional perception fusion module 1 is located inside the explosion-proof area of ​​the pyrotechnics production station and is used to collect multi-dimensional physical field change information in the production environment. The multi-dimensional perception fusion module 1 includes a multispectral imaging sensor 14, an ultraviolet pulse photodetector 15, an ultrasonic pressure disturbance sensor 16, and a protective cover 17.

[0038] In the specific implementation process, the multispectral imaging sensor 14 is installed at the monitoring position in the hazardous area for continuous imaging detection of the pyrotechnics processing area. The multispectral imaging sensor 14 has an ultraviolet detection channel 19, a visible light detection channel 20, and an infrared detection channel 21, covering a detection band of 200nm to 14μm. When the pyrotechnics are in normal production, the multispectral imaging sensor 14 collects background spectral information within the production area. When the pyrotechnics experience abnormal conditions due to friction, static electricity, mechanical impact, or environmental temperature rise, its ultraviolet radiation, visible light changes, and infrared thermal radiation characteristics will change. The multispectral imaging sensor 14 can acquire the above-mentioned abnormal change information in real time and form spectral characteristic data. Simultaneously, the ultraviolet pulse photodetector 15 is used to detect transient ultraviolet radiation within the hazardous area. During the pyrotechnics production process, when electrostatic discharge, electric sparks, or localized oxidation reactions occur, short-duration, high-intensity ultraviolet pulse signals are generated. The ultraviolet pulse photodetector 15 can quickly capture these signals and convert them into electrical signals for output.

[0039] The ultrasonic pressure disturbance sensor 16 is used to detect microscale pressure changes generated during the thermal decomposition of pyrotechnics. When thermal decomposition or enhanced local reactions occur inside the pyrotechnics, weak pressure fluctuations are generated. These pressure fluctuations are transmitted to the ultrasonic pressure disturbance sensor 16 through the air medium, thereby enabling auxiliary detection of the pre-ignition and detonation state. To improve the stable operation of the sensors in the pyrotechnics production environment, the multispectral imaging sensor 14, the ultraviolet pulse photodetector 15, and the ultrasonic pressure disturbance sensor 16 are all housed inside the protective cover 17 and fixed by the sensor mounting bracket 18. The protective cover 17 is made of dustproof and antistatic transparent material. While ensuring the normal transmission of optical signals, it can reduce the impact of dust, chemical particles and static environment on the detection element, and improve the long-term reliability of the equipment. The data collected by the multi-dimensional perception fusion module 1 is transmitted to the edge AI inference engine 2 through the industrial communication interface.

[0040] like Figure 2As shown, the edge AI inference engine 2 includes a preprocessing board 22, an FPGA synchronous processing chip 23, an AI edge computing motherboard 25, and an AI SoC processing chip 26. During implementation, the preprocessing board 22 first performs preliminary processing on the data output by the multi-dimensional perception fusion module 1, including noise filtering, time synchronization, data calibration, and feature normalization. Since multispectral image data, ultraviolet pulse data, and air pressure disturbance data have different data formats and sampling frequencies, the preprocessing board 22 unifies the time reference of data from different sources, so that the data from each sensor can correspond to the same time window, thereby improving the accuracy of subsequent fusion analysis. The preprocessed data enters the FPGA synchronous processing chip 23. The FPGA synchronous processing chip 23 utilizes hardware parallel computing capabilities to perform high-speed synchronous processing of multi-channel sensor data, realizing low-latency computation of large amounts of real-time data. The AI ​​SoC processing chip 26 internally runs a heterogeneous feature fusion module 27 and a temporal convolutional network module 28. The heterogeneous feature fusion module 27 is used to uniformly map multispectral image features, ultraviolet pulse features, and air pressure disturbance features to form a fused feature vector.

[0041] Multispectral image features include temperature distribution changes, abnormal brightness areas, and radiation intensity changes; ultraviolet pulse features include pulse amplitude, duration, and frequency of change; pressure disturbance features include pressure change amplitude and trend. These features are fused together to form a comprehensive hazardous state feature. The temporal convolutional network module 28 further performs time series analysis on the fused feature vector within the continuous sampling period. By analyzing the change law of the dangerous features over time, it determines whether the pyrotechnics are in a state of thermal decomposition, abnormal combustion, or critical combustion and explosion. The edge AI inference engine 2 establishes a danger index model based on the analysis results: when the danger index is lower than the first set threshold, the system maintains normal monitoring status. When the danger index reaches the first threshold but not the second threshold, a level one warning signal is output, the monitoring enhancement mode is activated, the sampling frequency is increased, and the monitoring of dangerous areas is strengthened. When the danger index reaches the second threshold, a secondary blocking signal is output, and the dual-mechanism transient blocking actuator 3 is activated. When the danger index exceeds the third threshold, a level 3 global linkage explosion suppression signal is output, and the multi-point spatial damping linkage network 4 is activated.

[0042] like Figure 3As shown, the dual-mechanism transient blocking actuator 3 includes a high-pressure pneumatic blasting and cutting device 32 and a high-expansion microencapsulated aerosol generator 38. The high-pressure pneumatic blasting and cutting device 32 is installed on the drug delivery pipeline 10 and includes a high-pressure gas cylinder 33, a drive piston 34, an annular cutting tool 35, and a tool guide groove 36. When the edge AI inference engine 2 outputs a secondary blocking signal, the control module immediately opens the high-pressure gas cylinder 33, allowing high-pressure gas to enter the drive chamber and drive the drive piston 34 to move at high speed.

[0043] The drive piston 34 moves the annular cutting tool 35 along the tool guide groove 36, causing the annular cutting tool 35 to quickly cut off the delivery pipe 10, thereby blocking the continued delivery of pyrotechnic materials and preventing the dangerous situation from spreading along the delivery path. At the same time, the high-expansion microencapsulated aerosol generator 38 operates synchronously. The high-expansion microencapsulated aerosol generator 38 includes an extinguishing medium storage chamber 39, a sodium bicarbonate microencapsulated dry powder layer 40, a fluorinated ketone extinguishing medium layer 41, a high-pressure nitrogen drive cylinder 42, a jet guide pipe 43, and a conical jet head 44. Upon receiving the activation signal, the high-pressure nitrogen drive cylinder 42 releases its driving force, causing the extinguishing material in the extinguishing medium storage chamber 39 to be released rapidly. After the sodium bicarbonate microcapsule dry powder layer 40 is released, it forms a physical covering effect, reducing the oxygen participation in the combustion zone. After the fluorinated ketone extinguishing medium layer 41 is released, it inhibits the combustion chain reaction by capturing combustion free radicals. The sprayed extinguishing medium enters the conical spray head 44 through the spray guide pipe 43 and forms an aerosol suppression zone 45 covering the combustion zone, achieving rapid fire extinguishing and thermal reaction suppression.

[0044] like Figure 4 As shown, the multi-point spatial damping linkage network 4 is used to achieve coordinated protection between multiple production stations. The multi-point spatial damping linkage network 4 includes an explosion-proof valve 46, an inert gas storage tank 48, an inert gas delivery pipeline, and an air curtain forming structure. During normal production, each production station maintains a data connection through an industrial communication link 51. When a production station detects a dangerous state and generates a level 3 global linkage explosion suppression signal, the signal is simultaneously sent to adjacent production stations through a hard-wired emergency stop line and the industrial communication network. After receiving the control command, the adjacent area immediately closes the explosion-proof valve 46 to block the possible propagation of shock waves and flames. At the same time, the inert gas storage tank 48 opens and releases inert gas to a designated location through the delivery pipeline, forming an inert gas isolation air curtain 50 between production areas. Through the combined action of the explosion-proof structure and the inert gas curtain, the energy propagation efficiency between the accident area and the surrounding area can be reduced, thereby improving the overall safety level of the production line.

[0045] This embodiment also includes a timing control module to control the execution order of different safety actions. Specifically, it includes a fire identification signal 53, an AI hazard level judgment signal 54, a cutting execution trigger signal 55, an aerosol spray trigger signal 56, an explosion-proof linkage control signal 57, and a full-area explosion suppression control signal 58. During operation, the system first obtains abnormal information on site through the multi-dimensional perception fusion module 1; then, the edge AI inference engine 2 completes the identification of the dangerous state; and triggers physical cutting, fire suppression, and area explosion-proof actions in sequence according to the hazard level. Through the above timing control method, fire identification, hazard judgment, blocking execution, and area protection form a continuous closed-loop control, improving the system response speed.

[0046] In this embodiment, each module adopts a modular explosion-proof design and can be combined and installed according to different pyrotechnic production processes. For example, multispectral sensing modules and aerosol suppression modules can be prioritized in the mixing process area; delivery pipeline cutting modules can be prioritized in the conveying area; and multi-point linkage explosion-proof modules can be prioritized in the encapsulation area.

[0047] In summary, the embodiments of this application, through multi-dimensional perception fusion, AI intelligent reasoning, transient blocking execution, and multi-point spatial linkage protection, realize the early identification, rapid isolation, and regional collaborative control of dangerous states in the production process of pyrotechnics and gunpowder, thereby improving the safety protection level in high-risk production environments.

[0048] The implementation principle of the intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics based on multispectral vision and AI in this application embodiment is as follows: A multi-dimensional perception fusion module collects spectral, ultraviolet, and pressure disturbance information during the pyrotechnics production process in real time. This information is then fused using an edge AI inference engine for multi-source data fusion and temporal risk analysis to identify thermal decomposition and explosion risks in advance. When the hazard level reaches a set threshold, the system automatically triggers a dual-mechanism transient blocking actuator, which simultaneously cuts off the delivery path and suppresses aerosol fire extinguishing to reduce the speed of accident development. Simultaneously, a multi-point spatial damping linkage network is controlled according to the risk level to achieve explosion isolation and inerting protection in adjacent areas, thus forming a closed-loop safety control system of "early perception - intelligent judgment - rapid blocking - regional explosion control".

[0049] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A smart early-stage fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI, characterized in that, include: Multidimensional perception fusion module (1), edge AI inference engine (2), dual-mechanism transient blocking actuator (3) and multi-point spatial damping linkage network (4); The multidimensional sensing fusion module (1) is set in the explosion-proof area of ​​the pyrotechnic production station and is used to collect multidimensional physical field information during the pyrotechnic processing, including spectral radiation information, ultraviolet pulse information and air pressure disturbance information. The edge AI inference engine (2) is connected to the multidimensional perception fusion module (1) for heterogeneous feature fusion processing of the multidimensional physical field information, and for identifying early abnormal states of pyrotechnic thermal decomposition based on the artificial intelligence inference model, and outputting a hazard level judgment signal. The dual-mechanism transient blocking actuator (3) is connected to the edge AI inference engine (2) and is used to perform physical blocking action on the propagation path of pyrotechnics after receiving the hazard level determination signal, and simultaneously release the fire extinguishing and suppression medium. The multi-point spatial damping linkage network (4) is connected to multiple pyrotechnic production stations respectively, and is used to control the operation of the explosion-proof mechanism and inert gas barrier mechanism in adjacent areas according to the hazard level determination result, so as to form a multi-level spatial protection barrier.

2. The intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI according to claim 1, characterized in that: The multidimensional sensing fusion module (1) includes: a multispectral imaging sensor (14), an ultraviolet pulse photodetector (15), an ultrasonic pressure disturbance sensor (16), and a protective cover (17). Among them: the detection band of the multispectral imaging sensor (14) covers 200nm~14μm, and is used to collect ultraviolet radiation, visible light changes and infrared thermal radiation information during the thermal decomposition of pyrotechnics; The ultraviolet pulse photodetector (15) is used to detect ultraviolet pulse signals generated by electrostatic discharge, electric sparks and pyrotechnic oxidation reactions; The ultrasonic pressure disturbance sensor (16) is used to detect microscale pressure disturbance changes generated during the thermal decomposition of pyrotechnics.

3. The intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI according to claim 2, characterized in that: The protective cover (17) adopts a dustproof and anti-static transparent structure. The multispectral imaging sensor (14), ultraviolet pulse photodetector (15) and ultrasonic air pressure disturbance sensor (16) are fixed inside the protective cover (17) through sensor mounting bracket (18). The multispectral imaging sensor (14) is equipped with an ultraviolet detection channel (19), a visible light detection channel (20) and an infrared detection channel (21) to achieve synchronous acquisition of multiple frequency bands.

4. The intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI according to claim 1, characterized in that: The edge AI inference engine (2) includes a preprocessing board (22), an FPGA synchronous processing chip (23), an AI edge computing motherboard (25), and an AI SoC processing chip (26). The preprocessing board (22) is used to filter, synchronize time and normalize features of the data collected by the multidimensional perception fusion module (1); The FPGA synchronous processing chip (23) is used to realize high-speed parallel processing of multi-sensor data; The AI ​​SoC processing chip (26) runs a heterogeneous feature fusion module (27) and a temporal convolutional network module (28) to predict the risk of pyrotechnic explosion based on continuous time series data.

5. The intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI according to claim 4, characterized in that: The heterogeneous feature fusion module (27) is used to map multispectral image features, ultraviolet pulse features and air pressure disturbance features to a unified feature space to form a multidimensional fused feature vector; The temporal convolutional network module (28) is used to analyze the changing trend of the fused feature vector within a continuous sampling period and output an AI hazard level determination signal.

6. The intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI according to claim 1, characterized in that: The dual-mechanism transient blocking actuator (3) includes a high-pressure pneumatic blasting and cutting device (32) and a high-expansion microencapsulated aerosol generator (38). The high-pressure pneumatic blasting and cutting device (32) is installed on the drug delivery pipeline (10) and includes a high-pressure gas cylinder (33), a drive piston (34), an annular cutting tool (35) and a tool guide groove (36). When a blocking control signal is received, the drive piston (34) pushes the annular cutting tool (35) to move along the tool guide groove (36) to cut off the drug delivery pipeline (10).

7. The intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI according to claim 6, characterized in that: The high-expansion microencapsulated aerosol generator (38) includes a fire extinguishing medium storage chamber (39), a sodium bicarbonate microencapsulated dry powder layer (40), a fluorinated ketone fire extinguishing medium layer (41), a high-pressure nitrogen drive cylinder (42), a jet guide pipe (43), and a conical jet head (44). The high-expansion microencapsulated aerosol generator (38) is used to generate an aerosol suppression zone (45) covering the combustion zone, which suppresses the combustion chain reaction through physical coverage and chemical free radical capture.

8. The intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI according to claim 1, characterized in that: The multi-point spatial damping linkage network (4) includes an explosion-proof valve (46), an inert gas storage tank (48), an inert gas delivery pipeline, and an air curtain forming structure; The explosion-proof valve (46) is installed between adjacent production stations to block the propagation of explosion shock waves; The inert gas storage tank (48) delivers inert gas to the air curtain injection position through the inert gas delivery pipeline to form an inert gas isolation air curtain (50). The multi-point spatial damping linkage network (4) uses an industrial communication link (51) to connect multiple production stations; When any production station generates a danger signal, explosion-proof control commands are simultaneously sent to adjacent production stations via hard-wired emergency stop lines and network control.

9. The intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives based on multispectral vision and AI according to claim 1, characterized in that: The system also includes a timing control module for recording and controlling the time relationship between fire detection, physical cutting, aerosol release and area explosion-proof actions; These include: fire identification signal (53), AI hazard level judgment signal (54), cutting execution trigger signal (55), aerosol spray trigger signal (56), explosion-proof linkage control signal (57), and full-area explosion suppression control signal (58); Through the timing coordination of the above signals, the system can complete identification, blocking, and area protection in the early stages of a fire.

10. A multispectral vision and AI-based intelligent early fire identification, immediate blocking, and multi-point linkage explosion control system for pyrotechnics and explosives according to any one of claims 1 to 9, characterized in that: The edge AI inference engine (2) establishes a danger index model based on the continuous sampling results of the multi-dimensional perception fusion module (1), and outputs the following signals according to the danger index level: Level 1 warning signal, Level 2 blocking signal and Level 3 full-domain linkage explosion suppression signal; Among them: the Level 1 early warning signal is used to activate the monitoring enhancement mode; The secondary blocking signal is used to activate the dual-mechanism transient blocking actuator (3); The Level 3 global linkage explosion suppression signal is used to activate the multi-point spatial damping linkage network (4). The system adopts a modular explosion-proof structure, with each module connected through an industrial communication interface. It is installed in hazardous production areas such as pyrotechnics mixing, filling, conveying, and packaging to achieve automatic sensing, intelligent judgment, rapid blocking, and regional collaborative explosion control during the pyrotechnics production process.

Images