Railway tunnel integrated fire detection intelligent lamp

Abstract

The invention discloses an integrated fire detection intelligent lamp for a railway tunnel. The integrated fire detection intelligent lamp comprises a multi-mode sensing module, an integrated thermal imaging array sensor, a multispectral flame sensor, a laser scattering type smoke sensor and an electrochemical CO sensor. An FPGA chip and a lightweight convolutional neural network model are arranged in the signal processing module, space-time fusion analysis is carried out on multi-modal sensing data in real time, and a fire judgment threshold is dynamically optimized through transfer learning; the dynamic ad hoc network communication module supports dual protocols of LoRaWAN and TSN; the adaptive power supply adopts an AC / DC power conversion module, integrates a piezoelectric vibration energy collection module and a thermoelectric power generation module, and combines a super capacitor and a lithium battery to realize quintuple redundant power supply; and the three-dimensional lighting array consists of a plurality of groups of independently controlled LED matrixes and is used for dynamically adjusting a lighting area and brightness distribution. The fire hazard detection sensitivity is improved through multi-dimensional data fusion, the illumination function is achieved, and the use requirement of the railway tunnel in the severe environment is met.

Description

Technical Field

[0001] This invention relates to the field of railway tunnel fire detection and lighting technology, and more specifically to an integrated intelligent lighting fixture for railway tunnel fire detection. Background Technology

[0002] Railway tunnels present a unique environment with relatively enclosed spaces. In the event of a fire, the flames spread rapidly, and smoke is difficult to expel, posing a significant threat to lives and railway infrastructure. Traditional fire detection systems have limitations in response speed, accuracy, and integration with existing tunnel facilities. While lighting fixtures, essential within tunnels, are widely and relatively evenly distributed, existing fixtures offer limited functionality and cannot meet real-time monitoring requirements.

[0003] Therefore, how to integrate fire detection functions into lighting fixtures to achieve comprehensive and real-time monitoring of tunnel fires, improve fire early warning capabilities, and provide strong support for the safe operation of railway tunnels has become an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0004] In view of this, the present invention provides an intelligent lighting fixture for integrated fire detection in railway tunnels, which integrates fire detection function into the lighting fixture to achieve comprehensive and real-time monitoring of tunnel fires, improve fire early warning capabilities, enhance fire detection sensitivity through multi-dimensional data fusion, and also has lighting function, meeting the usage requirements in the harsh environment of railway tunnels.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides an integrated intelligent fire detection lighting fixture for railway tunnels, comprising: The multimodal sensing module integrates a thermal imaging array sensor, a multispectral flame sensor, a laser scattering smoke sensor, and an electrochemical CO sensor. The thermal imaging array sensor constructs a tunnel temperature field model in real time through infrared radiation distribution, the multispectral flame sensor synchronously detects the ultraviolet-near-infrared spectral characteristics of the flame, the laser scattering smoke sensor inverts the smoke concentration through particle scattering intensity, and the electrochemical CO sensor detects changes in CO concentration gradient. The signal processing module, with a built-in FPGA chip and lightweight convolutional neural network model, performs spatiotemporal fusion analysis on multimodal sensor data in real time and dynamically optimizes the fire judgment threshold through transfer learning. The dynamic self-organizing network communication module supports both LoRaWAN and TSN protocols. Each lamp can autonomously build a Mesh network based on the tunnel topology and dynamically allocate communication paths to transmit data to the tunnel monitoring center. The adaptive power supply uses an AC / DC power conversion module to convert AC power into DC power required by the lamps, and integrates a piezoelectric vibration energy harvesting module and a thermoelectric power generation module, combined with a supercapacitor and a lithium battery to achieve five-fold redundant power supply. The three-dimensional lighting array consists of multiple independently controlled LED matrices. Based on the frequency of vehicle passage through the tunnel and the fire location results, the lighting area and brightness distribution are dynamically adjusted.

[0006] Furthermore, the signal processing module determines the fire status in the following ways: Temperature field gradient features are extracted from thermal imaging data and combined with the temporal variation of laser-scattered smoke concentration to generate a fire probability matrix; A CNN model was used to extract frequency domain features from multispectral flame sensor data, and then the features were weighted and fused with the probability matrix. When the fusion value exceeds the dynamic threshold and the CO concentration gradient conforms to the preset combustion curve, a multi-level alarm is triggered.

[0007] Furthermore, in the absence of a central node within the tunnel, the dynamic self-organizing network communication module achieves low-latency topology reconstruction based on the RPL routing protocol, and fire alarm information is transmitted to the tunnel monitoring center via a time-sensitive network (TSN) with deterministic latency.

[0008] Furthermore, the piezoelectric vibration energy harvesting module adopts a MEMS cantilever beam structure, has 8 sets of PZT-5H piezoelectric sheets, the resonant frequency matches the 15-25Hz vibration spectrum when a train passes by, and the maximum output power is 280mW.

[0009] Furthermore, the thermoelectric power generation module generates additional electricity by utilizing the temperature difference between the inside and outside of the tunnel. It is composed of a Bi2Te3 thermocouple and a lead telluride composite structure. When the temperature difference between the inside and outside of the tunnel ΔT ≥ 10℃, the output power is ≥ 1.5W.

[0010] Furthermore, the LED matrix of the three-dimensional lighting array achieves zoned control through a PWM dimming algorithm. When a fire occurs, the lighting in the fire source area is extinguished and the lighting is gradually increased to 150% brightness along the escape direction to form a guide light strip.

[0011] Furthermore, the lamp housing is made of graphene-reinforced composite material, with an embedded electromagnetic shielding layer and a self-cleaning photocatalytic coating on the outer surface to reduce the impact of dust and electromagnetic interference on the sensor.

[0012] Furthermore, the signal processing module also integrates a device health prediction model, which analyzes the sensor signal noise spectrum and power supply fluctuation characteristics to provide early warning of device aging failures.

[0013] Furthermore, the lamp also contains a six-axis gyroscope to monitor displacement vibration in real time and feed it back to the signal processing module to compensate for illumination deviation caused by mechanical deformation.

[0014] Furthermore, the luminaire has a three-layer architecture: the upper layer is the installation space for the multimodal sensing module, the middle layer is the installation space for the signal processing module, the dynamic self-organizing network communication module and the adaptive power supply, and the lower layer is the installation space for the three-dimensional lighting array. A 1mm thick aluminum shielding partition is installed between adjacent layers; The LED matrix package is fitted with a 650nm low-pass filter, and the multispectral flame sensor is fitted with a 700nm high-pass filter to avoid interfering with the near-infrared detection of the multispectral flame sensor.

[0015] As can be seen from the above technical solution, compared with the prior art, the present invention has the following technical advantages: 1. Multimodal sensor fusion: By combining multispectral, thermal imaging and smoke / CO detection, the accuracy of fire identification is significantly improved; 2. Low-latency decision-making: FPGA-accelerated CNN model inference enables low-latency fire response; 3. Energy self-sufficiency: Multiple redundant power supply features effectively ensure long-term functional operation even after a mains power outage; 4. Lighting strategy: Dynamic lighting strategy enhances personnel escape guidance capabilities and improves emergency response efficiency.

[0016] In summary, this invention, through the deep integration of multiple key technologies such as multimodal sensor fusion, intelligent signal processing, adaptive communication and energy management, and intelligent lighting control, constructs a highly integrated, reliable, and intelligent railway tunnel fire detection and emergency lighting device. It not only improves the accuracy and response speed of fire early warning but also enables intelligent lighting guidance and equipment health management during a fire, significantly enhancing the safety and intelligence level of railway tunnel operations. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 The structural block diagram of the intelligent lighting fixture for integrated fire detection in railway tunnels provided by the present invention.

[0019] Figure 2The multispectral flame detection flowchart provided by this invention; Figure 3 The fire determination flowchart of the signal processing module provided by the present invention is shown. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] This invention discloses an integrated intelligent fire detection lighting fixture for railway tunnels. The outer shell is made of graphene-reinforced composite material with a tensile strength >500MPa and electromagnetic shielding effectiveness >60dB (1GHz band). The outer surface is coated with a self-cleaning photocatalytic coating (e.g., a TiO2 nano-coating, which decomposes organic pollutants under ultraviolet light, with a transmittance attenuation of <5% / year), reducing the impact of dust interference on the sensor. Internally, it has a three-layer structure with independent chambers: a detection chamber at the top, a processing chamber in the middle, and an illumination chamber at the bottom. Adjacent chambers are separated by 1mm thick aluminum shielding partitions. Embedded electromagnetic shielding partitions ensure physical isolation or electromagnetic shielding between the sensor and the illumination components, preventing signal interference. The top of the outer shell also has a guide rail mounting assembly or magnetic block for fixing to the tunnel sidewall or top.

[0022] 1. Reference Figure 1 As shown, the detection chamber at the top: The system is equipped with a multimodal sensing module, consisting of a thermal imaging array sensor, a multispectral flame sensor, a laser scattering smoke sensor, and an electrochemical CO sensor; among which: The thermal imaging array sensor uses an uncooled microbolometer with a resolution of 1280×720 and a frame rate of 60Hz. It covers the tunnel cross section through a wide-angle lens with a field of view of 120° and generates three-dimensional point cloud data of the temperature field in real time. The multispectral flame sensor uses an InGaAs focal plane array to cover the wavelength range of 250-1800nm. It has a built-in narrowband filter wheel that switches the filter every 100ms to capture the ultraviolet (280-320nm), visible (430-670nm), and near-infrared (900-1700nm) characteristic spectra of the flame. The laser scattering smoke sensor uses a 785nm VCSEL laser diode with an emission power of 5mW. It receives the scattered light signal through a four-quadrant photodetector and uses Mie scattering theory to invert the smoke particle concentration, achieving a detection sensitivity of 0.1%obs / m. The electrochemical CO sensor employs a three-electrode solid electrolyte structure, with a detection range of 0-2000ppm and a resolution of 1ppm.

[0023] 2. Reference Figure 1 As shown, the processing chamber in the middle: Install signal processing module, dynamic self-organizing network communication module and adaptive power supply; 2.1 The signal processing module receives data from the multimodal sensing module and performs spatiotemporal fusion analysis on the multimodal sensing data in real time through the built-in FPGA chip and lightweight convolutional neural network model. The fire judgment threshold is dynamically optimized through transfer learning.

[0024] Reference Figure 2 The following is a flowchart of the multispectral flame detection process: 1) Filter wheel switching: The narrowband filter (center wavelength 290nm, 530nm, 880nm, 1550nm) is switched every 100ms, and spectral data of ultraviolet, visible and near-infrared bands are collected in sequence.

[0025] 2) Spectral data acquisition: The InGaAs focal plane array synchronously captures the light intensity distribution of each band and generates a digital spectral matrix (resolution 250-1800nm) through analog-to-digital conversion.

[0026] 3) Feature extraction: Ultraviolet band analysis: Detecting a sudden increase in intensity at 290nm wavelength (threshold > 5000 counts / ms) to determine whether there are initial combustion characteristics of a flame; Near-infrared emissivity calculation: Calculate the ratio of radiant intensity in the 1350nm and 880nm bands (R=1350 / 880). If R>2.5, it is determined that the flame is continuously burning. Spectral entropy calculation: Calculating the energy distribution entropy value from multi-channel spectral data ( , where p i (This refers to the energy percentage of each wavelength channel), and the entropy value in a flame scenario is typically <2.5.

[0027] 4) Comprehensive judgment: When the conditions of sudden increase in ultraviolet intensity, exceeding the limit of near-infrared emissivity and low entropy are met simultaneously, flame detection is triggered and verified in conjunction with temperature field and smoke concentration data.

[0028] Reference Figure 3 As shown, the specific signal processing module implements fire detection in the following ways: S1. Bilateral filtering is used to reduce noise in the thermal imaging data, and smoke data is smoothed using a Savitzky-Golay filter; then, temperature field gradient features are extracted from the thermal imaging data.

[0029] and These represent the rates of change of the temperature field in the horizontal (x-axis) and vertical (y-axis) directions, respectively, reflecting the severity of local temperature changes. In an FPGA, this is achieved through difference operations between adjacent pixels. For example, for a pixel with coordinates (i,j):

[0030]

[0031] By utilizing an FPGA parallel pipeline architecture, partial derivatives of multiple pixels can be calculated simultaneously, thereby improving throughput.

[0032] This represents the magnitude of the temperature gradient, combining temperature changes in both the x and y directions to characterize the overall intensity of temperature change at a given point. A larger value indicates a higher probability of temperature anomalies in that region. As shown in the formula above, it is obtained by summing the squares of the partial derivatives and then taking the square root.

[0033] Furthermore, by combining the temporal changes in the concentration of laser-scattered smoke, a fire probability matrix is ​​generated; S2. Using a CNN model to process multispectral flame sensor data: , where p i This represents the energy percentage of each wavelength channel. Frequency domain features are extracted and then weighted and fused with the probability matrix; S3. When the fusion value exceeds the dynamic threshold and the CO concentration gradient conforms to the preset combustion curve, a multi-level alarm is triggered.

[0034] A lightweight CNN model (input layer: temperature gradient, spectral entropy, smoke concentration, CO concentration; hidden layer: 3 fully connected layers; output layer: fire probability 0-1) is used to fuse multi-source data, and an alarm is triggered when the probability value is ≥0.98.

[0035] The aforementioned CNN model is optimized through transfer learning: in the pre-training stage, the base network is trained using a public fire dataset (such as FDDS); in the fine-tuning stage, real-world tunnel data (temperature, spectrum, smoke, CO time series data) is loaded, and the TripletLoss function is used to optimize the feature space distance, thereby improving the generalization ability in small sample scenarios.

[0036] In this embodiment, a thermal imaging array constructs a temperature field model, combines laser scattering smoke detection with multispectral flame recognition, and achieves feature-level fusion of multi-source data through CNN; the transfer learning model dynamically optimizes the alarm threshold according to regional climate characteristics, which can reduce the false alarm rate to below 0.5%.

[0037] In addition, the signal processing module also integrates a device health prediction model, which can provide early warning of device aging failures by analyzing the noise spectrum of sensor signals and power supply fluctuation characteristics. For example, it can use an LSTM network to predict the remaining lifetime, with input dimensions such as noise spectrum entropy, voltage fluctuation variance, and temperature gradient.

[0038] 2.2 Dynamic self-organizing network communication module, supporting LoRaWAN and TSN dual protocols, each lamp autonomously builds a Mesh network based on the tunnel topology, dynamically allocates communication paths to transmit data to the tunnel monitoring center; For example, TSN is dedicated to data transmission with high real-time requirements, such as fire alarms, while LoRaWAN is used for device health monitoring in non-emergency situations, dynamically adjusting communication paths through priority scheduling.

[0039] For example, in the absence of a central node within the tunnel, low-latency topology reconfiguration is achieved based on the RPL routing protocol, and fire alarm information is transmitted to the tunnel monitoring center via a Time-Sensitive Network (TSN) with deterministic latency. The TSN-based clock synchronization mechanism can achieve microsecond-level time synchronization, ensuring alarm information transmission latency is less than 10ms.

[0040] 2.3 Adaptive power supply: It adopts an AC / DC power conversion module to convert AC power into DC power required by the lamps, and integrates a piezoelectric vibration energy harvesting module and a thermoelectric power generation module. Combined with a supercapacitor (capacity 100F) and a lithium battery (capacity 5000mAh), it achieves five-fold redundant power supply; it can ensure full-function operation for up to 72 hours after the mains power is interrupted.

[0041] The piezoelectric vibration energy harvesting module adopts a MEMS cantilever beam structure and has 8 sets of PZT-5H piezoelectric sheets (size 10×3×0.2mm). The resonant frequency matches the 15-25Hz vibration spectrum when the train passes by, and the maximum output power is 280mW.

[0042] The thermoelectric power generation module generates additional electricity by utilizing the temperature difference between the inside and outside of the tunnel. It consists of a Bi2Te3 thermocouple and a lead telluride composite structure. When the temperature difference between the inside and outside of the tunnel ΔT≥10℃, the output power is ≥1.5W.

[0043] Power supply priority is defined as follows: AC / DC main power supply > lithium battery > supercapacitor > piezoelectric vibration > thermoelectric generator. Seamless switching and load balancing of multiple input sources can be achieved through a power management chip (such as TI BQ25703A).

[0044] 3. Reference Figure 1 As shown, the bottom lighting cavity: The three-dimensional lighting array consists of multiple independently controlled LED matrices. Based on the frequency of vehicle passage through the tunnel and the fire location results, the lighting area and brightness distribution are dynamically adjusted.

[0045] The three-dimensional lighting array consists of a 16×24 matrix of 384 high-brightness LEDs, each LED is driven independently, and brightness control from 0 to 100% is achieved through PWM dimming. After a fire is confirmed, the LEDs within a 5m radius of the fire source are extinguished (response time < 50ms), and the lighting is gradually increased to 150% brightness along the escape direction to form a guide light strip.

[0046] The LED matrix package incorporates a 650nm low-pass filter, while the multispectral flame sensor incorporates a 700nm high-pass filter to eliminate cross-interference and avoid interfering with the near-infrared detection of the multispectral flame sensor.

[0047] Specifically, for example: fire source location: based on the coordinates of the maximum gradient of the thermal imaging temperature field (x0, y0); shut-off radius: r = min(5m, 0.2v), where v is the current wind speed (obtained through a tunnel wind pressure sensor); escape route lighting: along the longitudinal direction of the tunnel, the brightness is adjusted by distance weighting with respect to the fire source (L(d) = 150% × e). -0.1∣d∣ (where d is the lateral distance from the escape route).

[0048] Furthermore, the lamp also contains a six-axis gyroscope to monitor the lamp tilt angle changes caused by displacement vibration in real time and feed them back to the signal processing module. Then, the module sends the control fine signal to the LED driver circuit and adjusts the LED drive current through the coordinate transformation matrix to compensate for the light offset.

[0049] The dynamic lighting control algorithm prioritizes the following logic: fire mode > vibration compensation > vehicle passage, ensuring priority response to emergency events.

[0050] In one embodiment, an alarm module may also be included, installed within the lighting cavity at the bottom and connected to the signal processing module. When the signal processing module determines that a fire has occurred, it triggers the alarm module. The alarm module employs an audible and visual alarm method, emitting a high-decibel alarm sound and flashing bright light to attract the attention of personnel inside the tunnel. Simultaneously, the alarm signal can be transmitted to the tunnel monitoring center via a dynamic self-organizing network communication module.

[0051] The intelligent fusion architecture of the multimodal sensing and signal processing module designed in this invention improves the detection accuracy to the ppm level; and it achieves full tunnel coverage communication based on dynamic self-organizing network. Multiple power supply methods break through the limitations of traditional power supply and adapt to extreme environments.

[0052] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0053] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart lighting fixture for integrated fire detection in railway tunnels, characterized in that, include: The multimodal sensing module integrates a thermal imaging array sensor, a multispectral flame sensor, a laser scattering smoke sensor, and an electrochemical CO sensor. The thermal imaging array sensor constructs a tunnel temperature field model in real time through infrared radiation distribution, the multispectral flame sensor synchronously detects the ultraviolet-near infrared spectral characteristics of the flame, the laser scattering smoke sensor inverts the smoke concentration through particle scattering intensity, and the electrochemical CO sensor detects the change in CO concentration gradient. The signal processing module, with a built-in FPGA chip and lightweight convolutional neural network model, performs spatiotemporal fusion analysis on multimodal sensor data in real time and dynamically optimizes the fire judgment threshold through transfer learning. The dynamic self-organizing network communication module supports both LoRaWAN and TSN protocols. Each lamp can autonomously build a Mesh network based on the tunnel topology and dynamically allocate communication paths to transmit data to the tunnel monitoring center. The adaptive power supply uses an AC / DC power conversion module to convert AC power into DC power required by the lamps, and integrates a piezoelectric vibration energy harvesting module and a thermoelectric power generation module, combined with a supercapacitor and a lithium battery to achieve five-fold redundant power supply. The three-dimensional lighting array consists of multiple independently controlled LED matrices. Based on the frequency of vehicle passage through the tunnel and the fire location results, the lighting area and brightness distribution are dynamically adjusted.

2. The intelligent fire detection lighting fixture for railway tunnels according to claim 1, characterized in that, The signal processing module determines the fire status in the following ways: Temperature field gradient features are extracted from thermal imaging data and combined with the temporal variation of laser-scattered smoke concentration to generate a fire probability matrix; A CNN model was used to extract frequency domain features from multispectral flame sensor data, and then the features were weighted and fused with the probability matrix. When the fusion value exceeds the dynamic threshold and the CO concentration gradient conforms to the preset combustion curve, a multi-level alarm is triggered.

3. The intelligent fire detection lighting fixture for railway tunnels according to claim 1, characterized in that, The dynamic self-organizing network communication module, in the absence of a central node in the tunnel, achieves low-latency topology reconstruction based on the RPL routing protocol, and fire alarm information is transmitted to the tunnel monitoring center through a time-sensitive network (TSN) with deterministic latency.

4. The intelligent fire detection lighting fixture for railway tunnels according to claim 1, characterized in that, The piezoelectric vibration energy harvesting module adopts a MEMS cantilever beam structure, has 8 sets of PZT-5H piezoelectric sheets, and its resonant frequency matches the 15-25Hz vibration spectrum when a train passes by, with a maximum output power of 280mW.

5. The intelligent fire detection lighting fixture for railway tunnels according to claim 1, characterized in that, The thermoelectric power generation module generates additional electricity by utilizing the temperature difference between the inside and outside of the tunnel. It is composed of a Bi2Te3 thermocouple and a lead telluride composite structure. When the temperature difference between the inside and outside of the tunnel ΔT≥10℃, the output power is ≥1.5W.

6. The intelligent fire detection lighting fixture for railway tunnels according to claim 1, characterized in that, The LED matrix of the three-dimensional lighting array achieves zoned control through a PWM dimming algorithm. When a fire occurs, the lighting in the fire source area is extinguished and the lighting is gradually increased to 150% brightness along the escape direction to form a guide light strip.

7. The intelligent fire detection lighting fixture for railway tunnels according to claim 1, characterized in that, The lamp housing is made of graphene-reinforced composite material with an embedded electromagnetic shielding layer and a self-cleaning photocatalytic coating on the outer surface to reduce the impact of dust and electromagnetic interference on the sensor.

8. The intelligent fire detection lighting fixture for railway tunnels according to claim 1, characterized in that, The signal processing module also integrates a device health prediction model, which analyzes the sensor signal noise spectrum and power supply fluctuation characteristics to provide early warning of device aging failures.

9. A smart lighting fixture for integrated fire detection in railway tunnels according to claim 1, characterized in that, The lamp also contains a six-axis gyroscope, which monitors displacement vibration in real time and feeds it back to the signal processing module to compensate for illumination deviation caused by mechanical deformation.

10. The intelligent fire detection lighting fixture for railway tunnels according to claim 1, characterized in that, The luminaire has a three-layer architecture: the upper layer is the installation space for the multimodal sensing module, the middle layer is the installation space for the signal processing module, the dynamic self-organizing network communication module and the adaptive power supply, and the lower layer is the installation space for the three-dimensional lighting array. A 1mm thick aluminum shielding partition is installed between adjacent layers; The LED matrix package is fitted with a 650nm low-pass filter, and the multispectral flame sensor is fitted with a 700nm high-pass filter to avoid interfering with the near-infrared detection of the multispectral flame sensor.

Images