High-pressure water mist fire extinguishing method and system for ship roll-on and roll-off places and special places

Abstract

The invention relates to the technical field of high-pressure water mist fire extinguishing, and discloses a high-pressure water mist fire extinguishing method and system for ship roll-on and roll-off places and special places, and the method comprises the following steps: acquiring temperature field data and ventilation airflow parameters, and integrating the temperature field data and the ventilation airflow parameters into heat source distribution mapping; performing heat source analysis according to the heat source distribution mapping to obtain multi-angle image data and heat source feature vectors; performing fire source position layout analysis to obtain a nozzle layout and a relative distance matrix; according to the nozzle layout and the relative distance matrix, nozzle direction and spraying intensity parameter optimization is executed, and fire extinguishing path configuration is obtained; when the spraying intensity parameter in the fire extinguishing path configuration is larger than a preset equipment load threshold value, feedback loop iteration is executed, and a nozzle control instruction sequence is obtained; and the nozzle control instruction sequence is transmitted to an actuator network to activate a corresponding nozzle, spraying range coverage judgment is executed, and a final nozzle control instruction sequence is obtained. According to the method, the abnormal heat source can be recognized and positioned in real time, and the spraying direction can be adaptively adjusted.

Description

Technical Field

[0001] This invention relates to the field of high-pressure fine water mist fire extinguishing technology, and particularly to high-pressure fine water mist fire extinguishing methods and systems for ship roll-on / roll-off spaces and special spaces. Background Technology

[0002] Currently, shipboard fire suppression systems primarily rely on fixed sprinkler systems or manual inspections. When faced with complex spaces such as roll-on / roll-off areas, engine rooms, and special compartments, fire response is often based on pre-set plans, lacking the ability to dynamically perceive and adjust to the state of the fire source. Due to the complex internal structure, confined spaces, and numerous heat sources on ships, traditional systems often struggle to identify the location and temperature changes of the fire source in the early stages of a fire, resulting in delayed fire suppression actions and an inability to achieve rapid and effective control.

[0003] In some existing technologies, temperature sensor networks or infrared monitoring devices are used to detect fire sources. However, due to limitations such as fixed equipment placement and susceptibility to signal interference, the accuracy of fire source identification and location remains low. Especially in roll-on / roll-off (Ro-Ro) facilities, where cargo is stacked irregularly and there is thermal radiation interference, the system is prone to misinterpreting normal high temperatures generated by equipment operation as fires, or misjudging the location of fire sources due to temperature field drift caused by ventilation airflow, making it difficult to deliver real-time and accurate fire suppression commands.

[0004] Existing technologies have the problem of being unable to monitor and accurately locate abnormal heat sources in real time in complex ship environments. Summary of the Invention

[0005] This invention provides a high-pressure fine water mist fire extinguishing method and system for ship roll-on / roll-off spaces and special spaces, so as to realize real-time identification and location of abnormal heat sources and adaptive adjustment of spray direction.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces, comprising: Acquire temperature field data and ventilation airflow parameters, and integrate the temperature field data and ventilation airflow parameters into a heat source distribution map; Based on the heat source distribution mapping, heat sources in abnormal areas are extracted and analyzed to obtain multi-angle image data and heat source feature vectors; Based on the multi-angle image data and the heat source feature vector, fire source location layout analysis is performed to obtain the nozzle layout and relative distance matrix; Based on the nozzle layout and the relative distance matrix, the nozzle direction and spray intensity parameters are optimized to obtain the fire extinguishing path configuration. When the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold, a feedback loop iteration is executed to obtain an optimized nozzle control command sequence. The nozzle control command sequence is transmitted to the actuator network to activate the corresponding nozzle, perform a spray range coverage judgment, and obtain the final nozzle control command sequence.

[0007] Secondly, the present invention provides a high-pressure fine water mist fire extinguishing system for ship roll-on / roll-off spaces and special spaces, comprising: The data acquisition module is used to acquire temperature field data and ventilation airflow parameters, and integrate the temperature field data and ventilation airflow parameters into a heat source distribution map; The anomaly analysis module is used to extract and analyze heat sources in abnormal areas based on the heat source distribution mapping, and obtain multi-angle image data and heat source feature vectors. The fire source layout module is used to perform fire source location layout analysis based on the multi-angle image data and the heat source feature vector to obtain the nozzle layout and relative distance matrix. The path configuration module is used to optimize the nozzle direction and spray intensity parameters based on the nozzle layout and the relative distance matrix to obtain the fire extinguishing path configuration; The optimization instruction module is used to perform feedback loop iteration to obtain an optimized nozzle control instruction sequence when the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold. The final instruction module is used to transmit the nozzle control instruction sequence to the actuator network to activate the corresponding nozzle, perform spray range coverage judgment, and obtain the final nozzle control instruction sequence.

[0008] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces as described in any one of the above.

[0009] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces as described in any one of the above.

[0010] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention uses temperature field and ventilation airflow as two complementary observations. First, timestamp alignment and noise reduction are performed, and then a continuous spatiotemporal heat source distribution is formed through gridded difference. On this basis, the correlation between temperature fluctuation and airflow disturbance is calculated, and the non-fire source temperature difference caused by ventilation is "subtracted" from the candidate anomalies. Then, the texture-brightness joint of multi-angle preprocessed images is used as the discrimination feature, and a support vector machine is used to establish the classification boundary in the feature space to separate normal high temperature (equipment heat dissipation, etc.) from abnormal heat sources. This derivation chain of "multi-source fusion → interference subtraction → feature discrimination" upgrades the identification of abnormal heat sources from a single threshold judgment to a statistical learning discrimination subject to physical constraints, reducing false alarms / false alarms. Therefore, in scenarios with large temperature fluctuations and complex airflow in roll-on / roll-off locations and special locations, it can lock the suspected ignition point earlier and more accurately, shorten the initial response delay and reduce invalid inspections and false triggers.

[0011] (2) This invention performs feature alignment and spatiotemporal registration on multi-view images to restore the three-dimensional coordinates of the fire source; then it embeds these coordinates into the three-dimensional model of the ship's structure to calculate the Euclidean distance between the fire source and each nozzle and construct a relative distance matrix. Using this as a constraint, a parameterized layout problem is established by combining the adjustable orientation of the nozzles and the spray intensity: when the distance exceeds the threshold or there is obstruction, coordinate system transformation and layout fine-tuning are performed to reduce geometric costs. Subsequently, the nozzle direction and intensity are optimized within a unified coordinate system, and the fire extinguishing path configuration is output. This process makes the spatial relationship between "heat source—structure—nozzle" explicit, transforming "distance, accessibility, and obstruction" into optimizable quantitative indicators, and deriving a combination of control quantities that satisfy geometric accessibility and effective energy delivery. As a result, the spray hit rate and unit fine water mist utilization rate are improved, water volume and number of actuations are reduced, suppression and cooling are more concentrated, and resource waste and secondary risks caused by spray range mismatch are avoided.

[0012] (3) This invention utilizes historical iteration data to adaptively adjust the iteration step size and number of iterations, redistributing the intensity of each nozzle and correcting its direction without exceeding the rated envelope of the pump / pipeline network; subsequently, it performs spatial coverage determination and necessary boundary correction on the final command sequence based on the nozzle coverage database. This derivation closed loop of "threshold monitoring → adaptive iteration → coverage verification" incorporates both equipment safety and fire extinguishing effectiveness into the control law, avoiding instability caused by overload, surge, or pressure fluctuations, while ensuring that key voxel areas receive sufficient droplet flux. Thus, when high-pressure fine water mist is activated, reliability, energy efficiency, and continuous strike capability are taken into account: the fire extinguishing effect is not sacrificed, long-term stable operation is maintained, and the overall fire extinguishing success rate and lifespan are improved. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces provided in the first embodiment of the present invention; Figure 2This is a schematic diagram of the high-pressure fine water mist fire extinguishing system for ship roll-on / roll-off spaces and special spaces provided in the second embodiment of the present invention. Detailed Implementation

[0014] 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.

[0015] Reference Figure 1 The first embodiment of the present invention provides a high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces, including the following steps: S11, acquire temperature field data and ventilation airflow parameters, and integrate the temperature field data and ventilation airflow parameters into a heat source distribution map; S12, Based on the heat source distribution mapping, perform heat source extraction and analysis in abnormal areas to obtain multi-angle image data and heat source feature vectors; S13, Based on the multi-angle image data and the heat source feature vector, perform fire source location layout analysis to obtain nozzle layout and relative distance matrix; S14, Based on the nozzle layout and the relative distance matrix, optimize the nozzle direction and spray intensity parameters to obtain the fire extinguishing path configuration; S15, when the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold, a feedback loop iteration is performed to obtain an optimized nozzle control command sequence; S16, the nozzle control command sequence is transmitted to the actuator network to activate the corresponding nozzle, perform spray range coverage judgment, and obtain the final nozzle control command sequence.

[0016] In step S11, it is necessary to acquire temperature field data and ventilation airflow parameters, and integrate the temperature field data and ventilation airflow parameters into a heat source distribution map, including: The temperature field data and the ventilation airflow parameters are timestamped to obtain the original dataset with time series. The original dataset is subjected to data denoising processing to obtain a smoothed dataset; Based on the smoothed dataset, the heat source intensity and location are calculated to obtain the heat source distribution dataset; A gridded interpolation analysis is performed on the heat source distribution dataset to obtain a dynamically updated heat source distribution map.

[0017] First, a multi-point sensor array, including temperature sensors and wind speed and direction sensors, is deployed in key areas of the ship's roll-on / roll-off (Ro-Ro) compartments and special cabins to collect raw parameters such as temperature, wind speed, and wind direction in real time. Each sensor is spatially calibrated before installation and registered with a unique number and location information in the system to ensure that the collected data accurately corresponds to the actual spatial location. During data acquisition, the system automatically and synchronously records data from each channel according to a preset sampling frequency. This sampling frequency is set with reference to the "Design Code for Fire Detection and Alarm Systems in Ship Engine Rooms" and the commonly used monitoring cycles of marine fire-fighting equipment, and optimized based on temperature change response characteristics observed in multiple experiments. A sampling period of approximately 1 second is typically selected to balance real-time performance and data volume management. Within each sampling period, data from all sensors are collected with reference to the same moment to ensure that multi-source data remain synchronized on the timeline.

[0018] After data collection, the system automatically generates a timestamp for each data record and adds identification information such as sensor number, spatial coordinates, and data source, forming a structured time-series raw dataset. For example, when there are 16 temperature measuring points and 8 wind speed and direction measuring points in the cabin, the system will simultaneously collect real-time data from 24 measuring points per second, such as temperature 26.5℃, wind speed 1.2m / s, and wind direction 85°, and record the timestamp "2025-09-27 10:05:00.000". Subsequently, these data are stored in the database in chronological order of collection time, forming a raw dataset with time series characteristics.

[0019] Regarding parameter settings, key thresholds in the system are calibrated based on actual operating condition tests and industry standards. For example, the temperature fluctuation threshold is typically set at 2°C / minute, an average value derived from statistical analysis of real-ship monitoring data from the early stages of fires in the vehicle compartments of multiple roll-on / roll-off ships, effectively distinguishing between equipment operating temperature rise and abnormal heat sources. The nozzle distance threshold is usually set at 3.5 meters, derived from the typical spray radius of marine fine water mist systems and the structural constraints of the compartments, ensuring coverage while avoiding resource waste. Through this parameter setting based on standards, actual measurements, and statistical analysis, the system not only ensures the universality and portability of data acquisition and judgment logic but also guarantees its adaptability and engineering feasibility under different ship types and cargo conditions.

[0020] After the raw data collection is completed, the temperature and airflow data undergo noise reduction and correction to ensure data stability and reliability. Due to various interference factors such as equipment operation, personnel activity, and airflow disturbances within ship roll-on / roll-off areas and special cabins, sensor readings may experience instantaneous fluctuations or abnormal readings during data acquisition. Therefore, a data fluctuation detection model is first established for each sensor channel, and trend analysis and range comparison are performed on continuously sampled data. When a single-point reading deviates from the preceding and following sample values ​​by more than a preset threshold (such as ±3℃ or ±30% wind speed change), the data is identified as an anomaly. Subsequently, a local smoothing algorithm is used to correct the anomalies. Common methods include the moving average method or Gaussian smoothing, which uses several adjacent data points at the current moment as a window to calculate a weighted average to replace the abnormal readings, thereby eliminating sudden changes caused by instantaneous interference. For example, when a temperature sensor records 25.6℃, 26.0℃, 31.2℃, 26.1℃, and 25.9℃ sequentially within 3 seconds, it will identify 31.2℃ as an outlier and replace it with the average of the adjacent values, 26.0℃, so that the temperature curve remains smooth and continuous.

[0021] For missing or delayed data uploads during the acquisition process, compensation calculations will be performed based on timestamps. If the interval between adjacent sampling exceeds a set period (e.g., 1 second) and data points are missing in between, linear interpolation or spline interpolation methods will be used to estimate the value at the intermediate time based on the temperature or wind speed change trends before and after the sampling point. For example, if the temperature is 25.5℃ at 10:05:00 and 26.0℃ at 10:05:02, but data at 10:05:01 is missing, the value at the intermediate time will be automatically supplemented to 25.8℃. In extreme cases, such as multiple consecutive missing sampling points or drastic fluctuations, the confidence level of the data for that period will be marked, and its weight will be reduced in subsequent fusion analysis to avoid misleading the overall heat source assessment.

[0022] After data smoothing, the heat source identification and location phase begins. The goal of this phase is to determine the intensity and approximate spatial location of heat sources in various areas within the cabin through joint analysis of the temperature and airflow fields. First, the smoothed temperature and airflow data are projected onto a pre-defined spatial grid model, with each grid cell representing a fixed monitoring area within the cabin. Then, the difference between the temperature within each grid and the average temperature of the surrounding area is calculated. Combined with the airflow velocity and direction information for that area, it is determined whether the temperature anomaly is caused by external ventilation or by an internal heat source.

[0023] When the temperature in a grid area is consistently higher than the surrounding area, and the airflow velocity is significantly lower or the airflow direction is turbulent, it will be marked as a potential heat source area. To improve the reliability of the identification, data from multiple time points will be compared to confirm that the temperature anomaly is persistent rather than a transient disturbance. For example, in grid monitoring of a vehicle compartment, if the temperature in a certain area is consistently more than 3°C higher than the surrounding area for 30 consecutive seconds, and the wind speed at that location is less than 0.5 m / s, it can be determined that there is a potential heat source there.

[0024] After identifying suspected heat source areas, the degree of heat concentration within the area is calculated based on the magnitude of temperature difference, spatial distribution gradient, and airflow attenuation characteristics, thereby estimating the heat source intensity. Higher heat source intensity indicates more concentrated energy release in the area. Each monitoring grid is marked with color or numerical levels, with areas of high heat concentration marked as red high-intensity zones and areas with stable temperatures marked as green background zones. This method provides a clear visual representation of the heat source distribution within the cabin.

[0025] After generating the heat source distribution dataset, spatial interpolation and visualization processing are performed to construct a complete and continuous heat source distribution map. Due to the complex structure of ship roll-on / roll-off (Ro-Ro) compartments or special cabins, sensor deployment is limited, resulting in some areas lacking direct monitoring. Therefore, based on the cabin's geometry and sensor placement, a two-dimensional or three-dimensional coordinate grid model corresponding to the actual space is first established, dividing the cabin space into multiple regular grid units, each representing a fixed-volume spatial region. The temperature values, airflow velocity, and direction data collected and processed by each sensor are then mapped to the corresponding grid nodes, forming a preliminary heat source distribution.

[0026] For areas without sensors, spatial interpolation algorithms are used for data extrapolation. The interpolation process comprehensively considers the temperature gradient of adjacent monitoring points, airflow direction, and distance weights to estimate the temperature and heat source intensity of the blank areas. For example, if the temperature at a monitoring point is 30℃, and the temperatures at its two adjacent monitoring points are 28℃ and 27℃ respectively, the estimated temperature of the intermediate area will be calculated based on the distance and airflow direction, ensuring a continuous and smooth transition trend in the overall temperature field. Simultaneously, in channel areas with strong airflow interference, the influence weights of wind speed and direction are increased to avoid the interpolation results being misled by a single temperature trend, thus ensuring that the mapping results more accurately reflect the actual physical conditions inside the cabin.

[0027] After interpolation calculations are completed, the temperature and airflow data from all grid nodes are combined to create a heat source distribution map. This map visually displays temperature changes and heat source distribution within the cabin using color gradients; areas with higher temperatures are highlighted with darker colors, emphasizing areas of high temperature concentration. To reflect the dynamic changes in the cabin environment, a fixed refresh cycle is set (which can be set to every second or every minute depending on the application scenario). Once new data is acquired, the interpolation results are automatically recalculated and the map is updated. In this way, the heat source distribution map is continuously refreshed over time, forming a continuous and dynamic trend of heat source changes.

[0028] Through this dynamic heat source distribution mapping, ship monitoring can grasp the overall situation of the internal temperature and airflow fields in real time, and promptly detect areas with abnormal temperature rises or heat accumulation. When the temperature in a certain area continues to rise and the airflow disturbance weakens, it will be automatically marked as a key monitoring area or potential fire source location, providing accurate basis for subsequent fire source location, abnormal area analysis, and sprinkler control strategy optimization. This process realizes intelligent calculation from discrete monitoring points to the complete spatial distribution, effectively compensating for monitoring blind spots caused by the limited number of sensors, and improving the comprehensiveness and accuracy of early warning and response to fires inside ship cabins.

[0029] In step S12, based on the heat source distribution mapping, it is necessary to extract and analyze heat sources in abnormal areas to obtain multi-angle image data and heat source feature vectors, including: Based on the temperature field data and ventilation airflow parameters in the heat source distribution mapping, the correlation between temperature fluctuations and ventilation airflow is calculated to obtain abnormal temperature fluctuations; When the abnormal temperature fluctuation exceeds a preset fluctuation threshold, the abnormal temperature fluctuation is divided into network interpolation regions to obtain the abnormal heat source region. Extract multi-angle image data corresponding to the abnormal heat source region; The multi-angle image data are subjected to data denoising and enhancement processing respectively to obtain preprocessed image data; Texture and brightness features are extracted from the preprocessed image data, and a heat source feature vector is generated by combining the support vector machine algorithm.

[0030] After mapping the heat source distribution, the temperature field data and ventilation airflow parameters are first extracted, and a temporal and spatial correlation analysis is performed on them. To ensure data alignment, the temperature and airflow change curves of each monitoring point are synchronized according to the timestamp, ensuring comparisons are made on the same time scale. Subsequently, a short time window (e.g., 30 seconds or 1 minute) is established in each monitoring grid to analyze the trend of temperature rise, fall, or fluctuation within this time period, and compare it with the corresponding changes in airflow velocity and wind direction. When the temperature change is basically synchronized with the increase in airflow or the change in wind direction, it is determined that the fluctuation is caused by ventilation disturbance; conversely, when the temperature continues to rise but the airflow change is not obvious, or the temperature rise trend does not match the wind direction, it is considered that there is an independent heat source in the area, exhibiting characteristics of abnormal temperature rise.

[0031] Based on this, the coupling degree between temperature changes and airflow changes is further calculated to reflect their correlation. Specifically, the temperature change amplitude, rate of change, and wind speed change trend of each grid within a time window are recorded, and then combined with spatial location relationships to analyze whether there are temperature propagation characteristics along the airflow direction. If the upstream grid first experiences temperature fluctuations, and the downstream grid responds with a temperature rise in a short period of time, the fluctuation is considered to be related to airflow conduction; if the temperature rise is limited to a local area and does not spread along the airflow direction, it indicates a high degree of independence and is more likely to be an abnormal heat source. Based on the comprehensive comparison results of multiple points, a "ventilation correlation index" is calculated for each monitoring grid, and grids with low correlation and significant temperature rise amplitude are selected as suspected anomalies according to a set threshold.

[0032] Finally, these suspected anomalies were compiled into an "Abnormal Temperature Fluctuation" record table. Each record includes the time period, spatial location, temperature rise rate, duration, and ventilation-related index values. When the index of a certain area is lower than the preset value and the temperature change exceeds the normal operating range, that point is marked as an abnormal temperature fluctuation, providing crucial input for subsequent heat source area delineation and multi-angle imaging. Through this process, it is possible to accurately screen temperature rise points with potential fire source characteristics in ship environments with strong ventilation interference, improving the accuracy of subsequent analysis and response.

[0033] Upon detecting abnormal temperature fluctuations, spatial extent identification and regional division are performed. First, all monitoring points marked as exhibiting abnormal fluctuations are extracted from the heat source distribution map, and a local grid is established based on their coordinates within the cabin space. Then, a proximity-based network interpolation method is used to expand these discrete abnormal points into continuous regions. During interpolation, the temperature change trend in unmonitored areas is estimated based on the temperature fluctuation amplitude, duration, and distance weights to adjacent points. If adjacent monitoring points exhibit consistent fluctuation directions and similar amplitudes, they are grouped into the same region; if the fluctuation trends are significantly different or airflow interruptions exist, they are considered independent regions. This combination of spatial interpolation and cluster analysis automatically delineates the boundary range of abnormal heat sources, generating an abnormal heat source region map containing the region center, boundary coordinates, and heat intensity distribution. This result provides accurate target region input for subsequent infrared imaging and three-dimensional fire source localization.

[0034] In a preferred embodiment, the "preset fluctuation threshold" is dynamically set based on the cabin operating conditions and equipment characteristics. Under normal circumstances, historical operating data and environmental characteristics are referenced to select an anomaly detection threshold where the temperature fluctuation exceeds 2°C / minute or the instantaneous temperature rise rate is more than 30% higher than the surrounding average. In areas with strong ventilation or high-temperature equipment, the threshold can be appropriately increased to 3°C / minute to avoid misjudging normal airflow disturbances as anomalies. In some high-sensitivity scenarios (such as hazardous materials warehouses and enclosed vehicle compartments), an adaptive threshold mode can also be used: real-time analysis of the overall standard deviation of the current temperature distribution is performed, and an anomaly detection is triggered when the temperature rise at a certain point exceeds twice the standard deviation. Through this multi-layered setting mechanism, the threshold can adapt to different operating conditions while maintaining high sensitivity in critical areas, ensuring that abnormal heat sources can be accurately identified without generating false alarms.

[0035] Once the abnormal heat source area is identified, the imaging module is automatically triggered to acquire images of that area from multiple angles. To avoid interference from obstruction or reflections in a single viewpoint, infrared cameras or visible light-assisted cameras deployed on the cabin roof, bulkheads, and key nodes are scheduled to simultaneously capture image data of the target area from different angles. Each camera carries a timestamp and spatial location information during acquisition, ensuring a one-to-one correspondence between the images and previous temperature fluctuation analysis results. When multiple abnormal areas are detected, images are scheduled to be captured sequentially according to risk level and spatial priority, prioritizing the capture of areas with larger temperature increases. To improve image coverage, the overlap range of the camera fields of view is automatically calculated, selecting the combination of angles that maximizes coverage of the target area. After acquisition, the multi-angle images are stored as a unified task batch for convenient subsequent data fusion and feature analysis.

[0036] After image acquisition, the multi-angle image data undergoes denoising and enhancement processing to improve the accuracy of subsequent recognition and feature extraction. First, in the denoising stage, median filtering or adaptive smoothing algorithms are used to remove isolated bright spots and artifacts caused by ambient light reflection, sensor thermal noise, or weak vibrations, thus maintaining the continuity of image texture and edge clarity. Then, in the enhancement stage, techniques such as contrast stretching and histogram equalization are employed to enhance the brightness difference between the heat source area and the background, making local high-temperature points more visually prominent. For images with uneven lighting or occlusion, a region-adaptive enhancement algorithm can be used to compensate for local contrast, ensuring complete presentation of details. After this processing, the output preprocessed image data reduces noise interference and highlights heat source features, providing clear and stable image input for subsequent texture and brightness analysis.

[0037] After obtaining stable preprocessed image data, the feature extraction and model recognition stages begin. First, texture and brightness features are extracted for each image. Texture features are primarily used to depict surface changes and temperature distribution details in the heat source region. By using a sliding window to statistically analyze local grayscale differences and directional changes, the texture contrast, uniformity, and directionality of the region are obtained. Brightness features reflect the overall temperature level and energy concentration of the heat source, extracting statistics such as the image's average brightness, maximum brightness, and brightness distribution skewness. All extracted features undergo normalization to eliminate biases caused by different camera angles or lighting conditions, forming a feature set at a uniform scale.

[0038] Subsequently, the extracted texture and brightness features are input into a Support Vector Machine (SVM) model for classification and feature encoding. The model employs a Radial Basis Function (RBF) kernel to balance the discriminative power of linear and nonlinear features. Key parameters include the penalty coefficient C and the kernel parameter γ. C balances the classification margin and error tolerance, set to 1.0 in this embodiment to maintain moderate sensitivity to outliers. γ controls the curvature of the decision boundary, determined through cross-validation based on sample distribution within a typical range, typically between 0.01 and 0.1. Training samples are drawn from historical labeled datasets, covering three scenarios: normal ventilation fluctuations, high-temperature equipment backgrounds, and real fire sources. Samples are divided at a 70% training and 30% validation ratio to ensure the model's generalization performance under different environments. During training, feature samples are input batch by batch, and the support vector positions and hyperplane parameters are adjusted by minimizing the classification loss function. The validation set accuracy and loss change are calculated after each iteration. When the decrease in validation loss is less than a preset threshold (e.g., 0.001) for several consecutive rounds or when the maximum number of iterations (e.g., 500 rounds) is reached, the training process will automatically terminate, which is considered as model convergence.

[0039] After model training is complete, the model weights and support vector set are fixed, and the system enters the online inference phase. At this stage, each newly generated image feature vector is input into the SVM model for classification and encoding. The output includes a heat source type label and a confidence score. If the classification result indicates an abnormal heat source with a confidence score higher than 0.9, it is determined that a real, suspected heat source exists in the area, and a final heat source feature vector is generated. This vector includes texture parameters, brightness indices, model confidence, and the corresponding image number and timestamp. This feature vector will serve as input data for subsequent fire source localization and nozzle optimization, achieving automatic conversion from image information to structured features and providing a reliable basis for intelligent decision-making.

[0040] In step S13, it is necessary to perform fire source location layout analysis based on the multi-angle image data and the heat source feature vector to obtain the nozzle layout and relative distance matrix, including: Feature point matching and alignment are performed on the multi-angle image data to obtain registered image data; Based on the registered image data, multi-view image data integration is performed to obtain unified image data; Based on the unified image data and the heat source feature vector, spatial geometric calculations are performed to obtain the three-dimensional coordinates of the fire source. Based on the three-dimensional coordinates of the fire source, and combined with a preset three-dimensional model of the ship's internal structure, the nozzle layout is extracted. Perform Euclidean distance calculations on the three-dimensional coordinates of the fire source and the nozzle layout to obtain a relative distance matrix.

[0041] Before spatially locating the fire source, feature point matching and alignment are first performed on multi-angle image data to achieve accurate registration of images from different perspectives within the same spatial coordinate system. Specifically, key feature points are extracted from thermal or visible light images taken from various angles. These feature points include the bright center points of the heat source region, edge contour intersections, and pixel nodes with prominent temperature gradient changes. To ensure the stability of the matching, feature description algorithms based on local image structure (such as corner detection or region gradient methods) are used to generate a unique description vector for each feature point, which is then used for subsequent cross-view comparison.

[0042] Subsequently, feature matching is performed between images from different viewpoints. The similarity of the descriptor vectors of each feature point is compared, and true corresponding points are selected through nearest neighbor search and mismatch elimination mechanisms. For the matching results, a random sampling consensus algorithm is used for model estimation to determine the geometric transformation relationship between the two images. This transformation model describes the differences in rotation, translation, and perspective projection between different viewpoints, enabling any image to be accurately mapped to the coordinate system of the reference viewpoint.

[0043] After obtaining the transformation relationship, the system performs spatial alignment processing on all viewpoint images. First, to ensure registration accuracy, the optical distortion of each camera is calibrated and corrected before performing coordinate transformation. Specifically, during the installation phase, the system pre-acquires standard checkerboard or calibration board images and generates lens intrinsic parameters and distortion coefficient models based on the calibration results. During registration, the system uses this model to correct radial and tangential distortion in the original images, thereby eliminating edge stretching, barrel or pincushion distortion caused by the lens and ensuring the linear consistency of the spatial geometry.

[0044] After distortion correction, the central region of the heat source is selected as a reference. Coordinate transformation and pixel resampling are then performed on images from other viewpoints to ensure that the position, shape, and boundary height of the same heat source region are consistent across different images. For images with significant viewpoint differences, the system uses a combination of bilinear interpolation and local affine transformation to correct spatial projection errors in distorted regions, ensuring that the overall contour is pixel-level aligned before fusion.

[0045] To address the occlusion issue, the system introduces an occlusion detection and compensation mechanism based on feature confidence. In multi-view data, if the feature confidence of a heat source region in a certain view is significantly lower than a threshold (e.g., below 0.7), it is determined that the region may be occluded by equipment or goods. In this case, the system reduces the weight of the local region in the data of that view and performs spatial interpolation or optical flow reconstruction using pixel information from other unoccluded viewpoints, thereby repairing the outline of the occluded heat source. The final output registered image data is the result after distortion correction, occlusion compensation, and spatial alignment, providing an accurate and robust input foundation for subsequent multi-view fusion and 3D coordinate calculation.

[0046] After feature point matching and registration of multi-view images, all aligned images are fused and integrated to generate unified image data that comprehensively reflects the spatial characteristics of the fire source. Specifically, spatial overlap detection is first performed on the registered images from each viewpoint to identify common viewing areas and independently visible areas among the images. For common viewing areas, a pixel-level weighted fusion method is used to weight and average the pixel brightness, temperature, or infrared radiation values ​​at the same location from multiple views; the weights are allocated based on the imaging sharpness, illumination uniformity, and occlusion degree of each viewpoint to ensure that high-confidence views contribute more to the results. For areas visible only from a single viewpoint, their pixel data is directly retained to avoid local information gaps caused by missing data.

[0047] During the fusion process, regional consistency constraints are employed to ensure the continuity of temperature gradients and contour changes across different viewpoints. For areas with slight edge shifts, optical flow correction and pixel interpolation techniques are used to repair seams, resulting in a structurally coherent image with a smooth transition in brightness distribution. Furthermore, local brightness compensation is performed on overexposed or underexposed areas to avoid temperature misjudgments caused by differences in lighting conditions across different viewpoints. By integrating multi-view data, the visual information from each camera can be utilized to the maximum extent, eliminating occlusion and bias issues from a single viewpoint.

[0048] The resulting unified image data uses thermal intensity as the primary information layer, supplemented by contour, texture, and gradient features, to fully present the projected shape of the fire source in three-dimensional space. This image not only clearly shows the boundaries and high-temperature concentration areas of the fire source but also preserves detailed features from different perspectives, providing accurate and continuous visual input for subsequent 3D coordinate reconstruction and spatial geometric calculations of the fire source. This integration step effectively realizes the transformation from multi-view distributed imaging to a unified view representation, laying a data foundation for precise fire source localization.

[0049] After obtaining unified image data, the spatial geometry calculation stage begins to transform the two-dimensional imaging results into three-dimensional spatial coordinates. First, a pre-defined camera geometric model is invoked, which includes the installation position, orientation angle, focal length parameters, and spatial coordinate information relative to the ship's cabin coordinate system for each camera. Based on these parameters, an imaging projection model is established for each camera, defining the correspondence between image pixels and actual spatial rays. By reading the calibrated feature points (such as heat source centers and highlighted boundary points) in the unified image data, the pixel coordinates of these features at each viewpoint are determined and projected into space to form a line-of-sight vector.

[0050] Subsequently, the intersection points of multiple lines of sight from different perspectives in three-dimensional space are calculated. Ideally, multiple lines of sight should intersect at the same spatial point; however, due to imaging errors and equipment jitter, the intersection points may deviate slightly. Therefore, a minimum error fitting method is used to optimize the intersection points of multiple lines of sight, and the common intersection area with the smallest error is selected as the center point of the fire source. When multiple high-temperature feature points are detected, the three-dimensional coordinates of each feature point are calculated separately, and the overall centroid is determined with thermal intensity as the weight, serving as the final representative coordinates of the fire source.

[0051] After positioning, the three-dimensional coordinates of the fire source are aligned with the ship's cabin coordinate system to correct spatial offsets caused by camera installation errors or projection distortion. During the correction process, known coordinates of fixed reference points within the cabin (such as cabin wall corners, vents, or calibration points) are used to fine-tune the fire source position, ensuring the result matches the actual spatial layout. The final output of the fire source's three-dimensional coordinates is recorded in (X, Y, Z) format, in meters, with an accuracy down to the centimeter level. These coordinates are not only used for subsequent nozzle distance calculations and direction optimization but can also be displayed in real-time on the ship's three-dimensional monitoring interface, helping to visually present the fire source location and enabling rapid positioning and response.

[0052] After obtaining the 3D coordinates of the fire source, they are matched with a pre-defined 3D model of the ship's internal structure to determine the specific compartment location of the fire source and extract the sprinkler layout information for the corresponding area. First, the 3D model stored in the ship's structural database is accessed. This model is built based on the actual compartment structure and includes geometric information such as decks, bulkheads, compartments, ventilation ducts, equipment areas, and vehicle storage areas. The 3D coordinates of the fire source are projected onto the global coordinate system of this model to determine its spatial unit and adjacent areas. Once the location is determined, the corresponding fire sprinkler nodes for that compartment are extracted from the model, including data such as the number of sprinklers, their coordinate positions, installation height, spray direction, and coverage angle.

[0053] When extracting the sprinkler layout, valid sprinklers are selected based on the structural characteristics of the area where the fire source is located. For example, if the fire source is located inside a vehicle compartment, sprinklers with unobstructed visibility and a spray direction that can cover the center of the fire source are prioritized; if the fire source is close to the bulkhead or equipment area, sprinkler nodes that cannot be directly reached by spray due to obstacles will be filtered out. The final generated sprinkler layout data includes not only the spatial coordinates of the sprinklers, but also the topological relationships and numbering information between the sprinklers, which are used for subsequent calculations and control command allocation.

[0054] In a preferred embodiment, the pre-defined 3D model of the ship's internal structure consists of multi-layered partitions and functional modules. The model employs a layered modeling approach, dividing the entire ship into multiple functional sections, including roll-on / roll-off vehicle compartments, engine room, living quarters, and special equipment compartments. Each section is modeled as a 3D geometric unit. Fire sprinkler nodes, ventilation duct paths, and equipment installation areas are pre-defined within each section. Node information is stored in coordinate form and associated with sprinkler type, flow parameters, and spray angle. The model is uniformly calibrated using a global coordinate system, with the origin set at the ship's center or the main deck reference plane, ensuring accurate mapping between fire source coordinates and model data within the same reference system. This 3D structural model enables rapid location of sprinkler distribution corresponding to fire sources in complex cabin environments, providing accurate structural basis for subsequent distance calculations and spray strategy optimization.

[0055] After determining the three-dimensional coordinates of the fire source and the corresponding nozzle layout, Euclidean distance calculations are performed to quantify the spatial distance relationship between each nozzle and the fire source. First, the spatial coordinate information of all nozzles, including their X, Y, and Z position parameters in the ship's coordinate system, is extracted from the nozzle layout data. Simultaneously, the three-dimensional coordinates of the fire source are used as a reference point to establish a coordinate comparison relationship. Then, the straight-line distance from each nozzle to the center point of the fire source is calculated one by one, and the calculation results are recorded as a numerical matrix, with the rows and columns corresponding to the nozzle number and fire source identifier, respectively. To avoid slight deviations caused by modeling errors, the coordinate data undergoes coordinate system correction and unit consistency processing before calculation to ensure the accuracy of the calculation results is within the centimeter range.

[0056] During distance calculation, not only is the spatial distance between the nozzle and the fire source output, but information on bulkheads and obstacles in the ship's internal structural model is also considered to determine if there are any obstructions in the nozzle's spray path. When a structural barrier (such as a bulkhead, pipe, or large equipment) is detected between the nozzle and the fire source, the nozzle is marked as "not for direct fire," and an accessibility weight is added to the relative distance matrix so that subsequent nozzle selection algorithms prioritize nozzle nodes with unobstructed paths. For cases with multiple fire sources or heat source centers, the distances from each nozzle to different fire sources are calculated to form a multi-target distance matrix, which is used to comprehensively optimize the spray range and resource allocation.

[0057] The resulting relative distance matrix fully reflects the spatial distribution relationship between the nozzles and the fire source in the current compartment. Each item in the matrix represents the straight-line distance between a specific nozzle and the center point of the fire source, along with auxiliary parameters such as path accessibility, angular deviation, and spray coverage angle. This matrix will serve as the core input for subsequent fire suppression strategy optimization, providing a quantitative basis for nozzle direction adjustment, spray intensity allocation, and fire suppression path planning. Through this process, a closed-loop transformation from spatial positioning to geometric calculation is achieved, ensuring that in complex compartment environments, nozzle scheduling can achieve precise coverage of the fire source with the shortest distance and optimal angle.

[0058] In step S14, based on the nozzle layout and the relative distance matrix, it is necessary to optimize the nozzle direction and spray intensity parameters to obtain the fire extinguishing path configuration, including: When the relative distance in the relative distance matrix is ​​greater than a preset distance threshold, coordinate transformation is performed to adjust the three-dimensional coordinates of the fire source and the nozzle layout to obtain an adjustment distance matrix. The nozzle direction and spray intensity parameters are optimized on the adjusted distance matrix to obtain the fire extinguishing path configuration.

[0059] In one specific embodiment, each "nozzle-fire source" distance in the relative distance matrix is ​​first checked item by item. When the spatial straight-line distance between a nozzle and the fire source exceeds a preset distance threshold, the system initiates a coordinate transformation and geometric correction process. Taking a ship's engine room as an example, assuming that the three-dimensional coordinate points of the heat source and the nozzle placement points are mapped to a unified cabin coordinate system, when the distance between a nozzle in a certain area and the center point of the fire source is detected to exceed 3.5 meters, the system first determines whether the nozzle is still within the effective spraying range based on the distribution density, spray angle, and coverage overlap of the surrounding nozzles. If it exceeds the range, the system performs a local coordinate transformation operation on the nozzle coordinates and the fire source coordinates.

[0060] This coordinate transformation mainly consists of two parts: rotation adjustment and spatial translation. First, based on the relative position of the fire source and the nozzle, the system adjusts the nozzle direction vector. By controlling the angle between the nozzle's spray axis and the line connecting the fire source, the spray centerline is aligned as closely as possible with the center of the fire source. For example, when the nozzle's original spray direction deviates from the fire source by a certain angle, the system calculates the required rotation angle and decomposes it into pitch and yaw angle adjustment commands, thereby correcting the nozzle's direction. Subsequently, based on the adjusted spray direction and the movable travel of the nozzle support, the system performs a slight translational correction on the nozzle's installation position to compensate for spatial deviations caused by equipment installation errors or coordinate offsets.

[0061] After completing the rotation and translation adjustments, the system recalculates the minimum reachable distance between the nozzle and the fire source and updates the relative distance matrix. Through this combined "rotation + translation" transformation, the nozzle's coverage area can be more accurately targeted at the fire source area, while ensuring that the distance meets the preset fire extinguishing requirements, thereby improving the spray hit rate and water mist utilization efficiency.

[0062] Specifically, the coordinate transformation includes two main steps: First, coordinate translation is performed, shifting the three-dimensional coordinates of the fire source along the shortest vector direction to the boundary position of the nozzle spray coverage cone, ensuring that the fire source is within the effective influence range of the jet flow; second, attitude correction is performed, calculating the adjustable optimal direction vector based on the pitch and yaw angles allowed by the nozzle support mechanism (e.g., ±25° range), ensuring that the nozzle's outflow direction is aligned only with the heat center of the fire source. After correction, the spatial distance and angular deviation between the nozzle and the fire source are recalculated, generating an updated adjustment distance matrix. If the distance after adjustment still exceeds the acceptable range, the nozzle is marked as "unreachable," and its weight is automatically reduced to the minimum or temporarily removed in subsequent optimization steps to avoid energy waste caused by ineffective spraying.

[0063] In this embodiment, the "preset distance threshold" is set differently based on the cabin type, nozzle model, and characteristics of the extinguishing medium. For example, for a high-temperature, enclosed space like a ship's engine room, a maximum effective extinguishing distance threshold of 2.5 to 3.5 meters is selected as the maximum effective extinguishing distance threshold between the nozzle and the fire source. For more open cargo holds or crew cabins, considering the larger spray diffusion radius, the threshold can be adjusted to 4.0 to 4.5 meters. The threshold setting is determined based on the nozzle spray velocity, atomization angle, and particle size characteristics of the extinguishing medium (such as water mist or dry powder). For example, when using a fine water mist nozzle with a flow velocity of 15 m / s and a spray angle of 60°, its effective extinguishing coverage radius is approximately 3 meters; therefore, a default threshold of 3.0 meters is used as a standard reference. This threshold can be adjusted through calibration tests during the deployment phase to balance extinguishing efficiency and energy consumption.

[0064] After obtaining the adjusted distance matrix, an improved genetic algorithm is used to jointly optimize the nozzle direction and spray intensity parameters to generate the optimal fire extinguishing path configuration. The algorithm primarily optimizes fire source coverage and minimizes equipment energy consumption, while also considering spray hit rate, single and cumulative water consumption, nozzle attitude adjustment costs, and pump and pipeline load constraints. During the algorithm initialization phase, a set of candidate solutions is generated, including nozzle on / off states, nozzle angles, and spray intensities. Each candidate solution corresponds to a spraying scheme. The population size is set to 50 individuals, the maximum number of iterations is 100 generations, the crossover rate is 0.8, the mutation rate is 0.05, and two optimal individuals are retained for elite inheritance.

[0065] In each iteration, the simulation module is invoked to adaptively evaluate all candidate solutions. The simulation calculates the spatial coverage of the fire source by adjusting the distance matrix and the nozzle coverage model, and evaluates the equipment load by combining data such as pump power, pipeline flow rate, and nozzle pressure. If a solution causes the pump pressure or flow rate to exceed the safety limit, a high penalty score is automatically assigned to reduce its probability of being selected. The algorithm iterates continuously to balance coverage effect and energy consumption cost, and terminates optimization when the fire source is sufficiently covered and the equipment load is within the safe range.

[0066] The final fire suppression path configuration includes the activated nozzle numbers, pitch and yaw angles, spray intensity values, activation sequence, and duration. The results undergo two verification steps before output: first, simulation coverage verification to confirm that the fire source area is within the coverage range of the spray cone; second, equipment safety verification to verify that pump power, flow rate, and valve operation meet rated operating conditions. If the tests fail, local parameter corrections or re-optimization will be triggered to ensure that the generated control scheme meets requirements in both fire suppression effectiveness and safety.

[0067] In step S15, when the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold, a feedback loop iteration is performed to obtain an optimized nozzle control command sequence. When the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold, the number of iterations is adjusted based on the preset historical iteration data to obtain the updated number of iterations; Based on the number of update iterations, the spray intensity parameters are optimized to obtain an optimized nozzle control command sequence.

[0068] In this embodiment, when the spray intensity parameter in the fire extinguishing path configuration exceeds a preset equipment load threshold, a feedback loop is initiated to adaptively adjust the spray intensity. The preset equipment load threshold is a dynamic limit determined based on the rated operating capacity and safety margin of fine water mist fire extinguishing, used to prevent abnormalities such as overpressure, flow overload, or sudden increases in energy consumption in the pump set or pipeline network. Specifically, during the design phase, multiple threshold levels are set based on equipment specifications and operating experience. For example, the maximum spray pressure of a single nozzle can be set to 12-14 MPa, the total pump set power to be 90% of the rated value, and the flow rate of a single branch not exceeding 95% of the design value. Simultaneously, to cope with transient fluctuations under complex operating conditions, a time smoothing window is also set. Only when the spray intensity of a certain nozzle exceeds the threshold for three consecutive sampling periods is it considered a valid trigger condition to avoid malfunctions.

[0069] Upon detecting that the spray intensity parameters exceed the equipment load threshold, the system first retrieves a pre-set historical iteration database to guide subsequent optimization strategies. This database is automatically generated and updated by the system during long-term operation and multiple shipboard tests. The collected data includes information from each firefighting mission, such as fire source type, heat source size, compartment geometry parameters, nozzle arrangement, pump response curves, flow and pressure changes, spray intensity adjustment records, and final convergence efficiency. After the mission is completed, this data is automatically organized, labeled, and stored in a local database or central management platform by the control system. Each record includes a timestamp, mission number, and scene feature vector to ensure rapid retrieval and access later.

[0070] During feedback iteration, the system uses similarity-based matching algorithms (such as nearest neighbor search based on Euclidean distance of feature vectors or similar scene matching based on KNN) to search for the record closest to the current operating condition in the historical database, based on the currently detected heat distribution of the fire source, nozzle layout, pump load, and equipment operating status. The system extracts statistical features such as the average convergence algebra, parameter adjustment rate, and system stabilization delay from these matched scenarios to dynamically adjust the initial step size and maximum number of iterations for the current optimization. For example, when the current operating condition is detected to have a similarity of more than 85% with a typical historical scenario of "high heat flux density + medium wind speed interference," the system will directly use the average number of convergence iterations of that scenario (e.g., 45) as the initial setting for the current optimization process.

[0071] During the iterative optimization process, the system employs a parameter optimization mechanism based on a genetic algorithm, gradually adjusting the spray intensity value and allocation strategy through population search and multi-generation evolution. The main parameter settings of the algorithm include: initial population size (50 candidate solutions), maximum number of iterations (100 generations), crossover probability (0.8), mutation probability (0.05), and elite retention ratio (5%). If the pump pressure change rate is detected to be too rapid or the nozzle load curve fluctuation exceeds the threshold during iteration, the system will automatically reduce the step size and increase the number of iterations to improve search accuracy; if the convergence trend is rapid, the evolution can be terminated early to shorten the response time. Finally, when the iteration error is below the preset threshold for several consecutive generations, or the fire extinguishing coverage reaches the target range, the system will stop optimization and output an updated nozzle control command sequence.

[0072] When adjusting the number of iterations, the system employs adaptive control logic to dynamically adjust the optimization process. When a large deviation in spray intensity is detected and the overload duration is prolonged, the number of iterations is automatically increased to allow for finer-grained parameter correction. Conversely, when the deviation is small and the trend is relatively stable, the number of iterations is reduced to improve response speed. For example, within one iteration cycle, if the system determines that the average intensity of multiple nozzles exceeds the limit by more than 20%, the number of iterations will be increased from the default 10 to 15, and the time interval between each iteration will be shortened accordingly to converge to the safe operating range more quickly. Conversely, when the exceedance ratio is below 5% and the pressure fluctuation curve remains within a stable range for several consecutive cycles, the number of iterations will be reduced to 5 to avoid unnecessary redundant calculations.

[0073] To prevent the iteration process from continuing indefinitely, the system incorporates multiple convergence conditions as stopping criteria. First, a spray intensity error threshold is set as the core standard: when the deviation between the calculated sprinkler spray intensity and the equipment's safety rating is less than ±3% for three consecutive iterations, the system is considered to have reached convergence. Second, a coverage rate criterion is introduced: when the spray coverage rate of the fire source area is higher than 95% in two consecutive assessments, the system considers the parameter adjustment complete. Furthermore, an iteration limit constraint is set; for example, regardless of the optimization progress, the total number of iterations cannot exceed 50. Once any convergence condition is met, the system immediately terminates the iteration process, outputs the current optimal parameter set, and updates the sprinkler control command sequence.

[0074] Through this multi-level termination mechanism based on deviation trend, error threshold, coverage index and iteration upper limit, the system can ensure that the parameters converge quickly within a reasonable range, and avoid infinite loop problems caused by external interference or abnormal data, thereby maintaining the safety and controllability of the fire extinguishing system in dynamic optimization.

[0075] After a number of iterations, the spray intensity parameters are optimized to achieve a dynamic balance between safe load and fire extinguishing effect in the sprinkler control strategy. The optimization process is automatically executed by the central control module. Its core objective is to adjust the spray intensity of each sprinkler to below the equipment load threshold while ensuring sufficient spray coverage density in critical fire source areas. First, based on the adjusted distance matrix and the current spatial distribution of the sprinklers, the effective range and overlap area of ​​each sprinkler are determined. Then, combining the temperature gradient and heat flux information of the three-dimensional coordinates of the fire source, the relative spray weight that each sprinkler should bear in the next iteration is calculated. For areas with high heat source intensity and close proximity to the sprinkler, the spray intensity weight is appropriately increased; conversely, for sprinklers that are far away or have high coverage overlap, the spray intensity is automatically reduced to decrease the total flow load.

[0076] In each iteration, parameters are updated based on the previous spraying results and equipment load feedback. The control module evaluates the intensity parameters of all nozzles in parallel, calculating a new flow distribution scheme and pressure setpoint. If, after a certain iteration, the total flow rate is still found to be 5% higher than the threshold limit, a proportional scaling adjustment mechanism will be automatically triggered, reducing the intensity of all nozzles by 2% to 3%, and the effect will be re-evaluated in the next cycle. Conversely, if the overall spraying intensity is too low, resulting in insufficient fire extinguishing coverage, the intensity of some nozzles can be locally increased, but the overall intensity is still constrained by the total load. This process continues iterating until the spraying intensity meets the load safety range and the fire extinguishing coverage reaches the preset standard, or the maximum number of iterations is reached.

[0077] Once the optimization process converges, the system automatically generates the final sprinkler control command sequence. This sequence includes detailed control parameters such as the unique number of each sprinkler, pitch and yaw angle settings, target spray intensity, on / off time points, continuous spray duration, power budget, and execution priority. It also scores the stability and safety of the current scheme. For example, in a typical fire scenario, the output control sequence includes the coordinated actions of 24 sprinklers. 18 sprinklers are activated synchronously in the initial stage, with the spray intensity set to 80% of the maximum value. Subsequently, the output is gradually reduced according to the heat source decay trend. The remaining 6 sprinklers maintain the uniformity of atomization coverage through low-intensity intermittent supplementary spraying.

[0078] Considering the issues of communication latency, data packet congestion, or differences in response speed among local nodes in the actuator network within a real-world marine environment, the system automatically initiates a communication latency compensation and synchronization scheduling mechanism after command generation. Specifically, the system first measures the average transmission latency and response time difference of each actuator node in real time on the control bus, and adds a timestamp and execution offset parameter to each control command to ensure that all nozzle actions are aligned on the global timeline. When the network latency of a nozzle is detected to exceed the allowable range, the system automatically sends commands in advance or adjusts the startup sequence to synchronize its spraying time with that of other nozzles.

[0079] Furthermore, after the command is issued, the system monitors the response status and action time of the nozzles in real time through the field feedback link and compares them with the expected control sequence. If it is found that some nozzles start late or fail to reach the preset spray intensity, the control system will immediately trigger a secondary correction process to dynamically adjust the spray sequence or compensate the nozzle output, thereby avoiding the imbalance of the overall fire extinguishing rhythm caused by network latency. Through this three-layer control strategy of "delay perception - advance scheduling - feedback correction", the system ensures the global synchronization of nozzle actions and the consistency of the timing of the fire extinguishing strategy, enabling the final control command to be executed efficiently and reliably even in complex ship communication environments.

[0080] In step S16, the nozzle control command sequence needs to be transmitted to the actuator network to activate the corresponding nozzle, perform a spray range coverage determination, and obtain the final nozzle control command sequence, including: Based on a preset nozzle coverage database, determine whether the spray coverage area includes the three-dimensional coordinates of the fire source, and obtain the coverage determination result; Based on the coverage determination result, the spray intensity parameter of the nozzle control command sequence is adjusted to obtain the final nozzle control command sequence.

[0081] The preset sprinkler coverage database is a key supporting module for the system's spray coverage determination and sprinkler parameter optimization. Its construction process, undertaken during the design phase, involves a collaborative approach using experimental testing, numerical simulation, and scenario modeling to ensure the database's parameter accuracy and applicability. Firstly, the sprinkler model parameter table is compiled from technical manuals, product testing reports, and factory calibration data provided by the equipment manufacturer. It undergoes secondary verification through laboratory testing before installation, such as measuring key indicators like nozzle diameter error and deviation between rated pressure and actual output using a standard water mist testing platform, ensuring the accuracy of the basic parameters.

[0082] Secondly, the spraying performance dataset was constructed through a dual approach of "experimental measurement + numerical simulation." In the experimental phase, the system systematically measured the coverage radius, droplet distribution, and cone angle changes of different nozzle models under various pressure, flow rate, spray angle, and ambient temperature and humidity conditions within a standard chamber test platform. In the simulation phase, a CFD fluid dynamics model was used to simulate the coupling behavior of water mist diffusion and airflow, supplementing data from extreme conditions or special environments that are difficult to cover experimentally. The experimental and simulation data were fused, smoothed, and fitted to form multidimensional performance curves, which were then used in the performance mapping module of the database.

[0083] The construction of the environmental correction coefficient set is based on a large amount of measured data from inside the cabin and comparative simulation analysis. The system installs wind speed, temperature, and humidity sensors in different types of ship cabins to collect data on airflow fields, thermal convection distribution, and the influence coefficients of obstacles on the water mist diffusion trajectory. After comparative analysis with simulation models, these influences are transformed into quantifiable correction coefficients, and a lookup table model is established according to cabin type, equipment layout density, ventilation mode, and other conditions for subsequent real-time compensation.

[0084] Finally, the three-dimensional spray cone distribution model is constructed driven by both experimental and simulation data. The system discretizes the spray coverage area into three-dimensional voxel units and calibrates the voxel space based on the spray cone morphology under different pressure, angle, and environmental conditions to form a standardized spatial distribution model. After the database is established, this model will be verified through actual shipboard spraying experiments. For example, by deploying humidity sensor arrays at different heights and locations, the deviation between the atomized coverage boundary and the model prediction results will be detected to ensure that the error is controlled within an acceptable range (e.g., within ±5%).

[0085] The database establishment process includes three types of data acquisition: first, experimental measurement, using infrared ranging and high-speed photography to record spray trajectories and boundary morphologies under different pressures and angles; second, numerical simulation, using computational fluid dynamics models to simulate droplet movement paths and obtain effective spray radius and density distribution under different operating conditions; and third, on-site correction, adjusting parameters based on the actual environment of the ship's cabins (such as ventilation layout and bulkhead reflection characteristics). All data is interpolated and fitted before being stored in the database, allowing for real-time retrieval of a matching spray coverage model by inputting the current nozzle model, pressure, angle, and environmental parameters during runtime.

[0086] During coverage determination, the system first reads the 3D coordinates of the current fire source and the position and attitude information of each nozzle in the distance adjustment matrix. Then, it calls the nozzle coverage database, extracts the corresponding 3D spray cone model based on the real-time spray intensity, angle, and environmental conditions of each nozzle, and generates a spray coverage volume in a unified coordinate system. The fire source coordinates are mapped to this volume space, and it is determined whether the point is within the effective coverage volume of the spray cone. If the fire source is completely within the coverage volume of a nozzle, the nozzle is marked as "effective coverage"; if the fire source is located in an edge region or at the intersection of multiple nozzle coverage volumes, a secondary determination is made based on the droplet density threshold to confirm the sufficiency of the fire source's impact from the spray; if the fire source is not included in any nozzle coverage volume, it is marked as "insufficient coverage".

[0087] For example, in a specific embodiment, five nozzles, of types A and B, are arranged in a compartment. The spray pressure is set to 0.6 MPa, and the spray angles are 70 degrees and 90 degrees, respectively. The corresponding spray model in the database is called to generate the volumes of the five spray cones. Calculations show that the coordinates of the fire source located in the center of the compartment are only partially covered by nozzles 1 and 3, with a coverage density of less than 70%. Based on this, the coverage judgment result is "partial coverage," and nozzles 1 and 3 are marked as objects requiring optimization. The subsequent optimization module will adjust the nozzle angles and spray intensity based on this result to ensure that the fire source coordinates are completely covered. Through this process, the nozzle coverage database not only supports coverage range judgment but also provides accurate basic data support for subsequent dynamic optimization of spray parameters.

[0088] Based on the coverage assessment results, the spray intensity parameters of the sprinkler control command sequence are further dynamically adjusted to obtain the final sprinkler control command sequence. Specifically, when the coverage assessment results indicate that the fire source area is only partially covered by sprinklers or that there are edge blind spots, the intensity of sprinklers at the edge of the covered area is increased first, while the intensity of sprinklers in the overlapping covered areas is suppressed or time-staggered to avoid resource waste caused by redundant spraying. This adjustment process is completed jointly based on a preset spray energy efficiency model and an equipment load model. The spray energy efficiency model describes the fire extinguishing coverage efficiency per unit intensity, while the equipment load model reflects the changes in electrical and hydraulic loads of each sprinkler under different intensities.

[0089] In one specific embodiment, the "spray radius-intensity function" and "spray angle-height function" corresponding to each sprinkler head in the coverage database are first read, and the overlap index is calculated by combining the three-dimensional coordinates of the fire source and the spatial layout of the sprinklers. If the overlap index exceeds a preset overlap threshold (e.g., 0.35), the spray intensity of that sprinkler head is reduced by 10% to 30% to suppress the fire. Conversely, if the fire source is located at the edge of the coverage of two sprinklers and the overlap is less than 0.15, the spray intensity of the edge sprinkler head is increased to 80% to 90% of the rated upper limit to ensure the continuity of fire suppression coverage.

[0090] Furthermore, the current total load factor is also considered during the adjustment process. When the total load factor approaches a preset threshold (e.g., 0.9), a short-term time-sharing triggering mechanism is used for coordinated control. That is, some nozzles perform high-intensity spraying in the first cycle, and then enter a low-intensity maintenance mode, while other nozzles are activated again to reduce the instantaneous peak load. After this adaptive adjustment, a final nozzle control command sequence is formed to ensure complete coverage of the fire source area and optimal distribution of extinguishing intensity within the safe load range.

[0091] In summary, the present invention provides a high-pressure fine water mist fire extinguishing method and system for ship roll-on / roll-off spaces and special spaces, so as to realize real-time identification and location of abnormal heat sources and adaptive adjustment of spraying direction.

[0092] Reference Figure 2 The second embodiment of the present invention provides a high-pressure fine water mist fire extinguishing system for ship roll-on / roll-off spaces and special spaces, comprising: The data acquisition module is used to acquire temperature field data and ventilation airflow parameters, and integrate the temperature field data and ventilation airflow parameters into a heat source distribution map; The anomaly analysis module is used to extract and analyze heat sources in abnormal areas based on the heat source distribution mapping, and obtain multi-angle image data and heat source feature vectors. The fire source layout module is used to perform fire source location layout analysis based on the multi-angle image data and the heat source feature vector to obtain the nozzle layout and relative distance matrix. The path configuration module is used to optimize the nozzle direction and spray intensity parameters based on the nozzle layout and the relative distance matrix to obtain the fire extinguishing path configuration; The optimization instruction module is used to perform feedback loop iteration to obtain an optimized nozzle control instruction sequence when the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold. The final instruction module is used to transmit the nozzle control instruction sequence to the actuator network to activate the corresponding nozzle, perform spray range coverage judgment, and obtain the final nozzle control instruction sequence.

[0093] It should be noted that the high-pressure fine water mist fire extinguishing system for ship roll-on / roll-off spaces and special spaces provided in this embodiment of the invention is used to execute all the process steps of the high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0094] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an optimized instruction program. When the processor executes the computer program, it implements the steps in the embodiments of the high-pressure fine water mist fire extinguishing methods for various ship roll-on / roll-off spaces and special spaces described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments, such as the instruction optimization module.

[0095] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0096] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0097] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0098] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0099] If the modules / units integrated into the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or system capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0100] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0101] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces, characterized in that, include: Acquire temperature field data and ventilation airflow parameters, and integrate the temperature field data and ventilation airflow parameters into a heat source distribution map; Based on the heat source distribution mapping, heat sources in abnormal areas are extracted and analyzed to obtain multi-angle image data and heat source feature vectors; Based on the multi-angle image data and the heat source feature vector, fire source location layout analysis is performed to obtain the nozzle layout and relative distance matrix; Based on the nozzle layout and the relative distance matrix, the nozzle direction and spray intensity parameters are optimized to obtain the fire extinguishing path configuration. When the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold, a feedback loop iteration is executed to obtain an optimized nozzle control command sequence. The nozzle control command sequence is transmitted to the actuator network to activate the corresponding nozzle, perform a spray range coverage judgment, and obtain the final nozzle control command sequence.

2. The high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces according to claim 1, characterized in that, The process of integrating the temperature field data and the ventilation airflow parameters into a heat source distribution map includes: The temperature field data and the ventilation airflow parameters are timestamped to obtain the original dataset with time series. The original dataset is subjected to data denoising processing to obtain a smoothed dataset; Based on the smoothed dataset, the heat source intensity and location are calculated to obtain the heat source distribution dataset; A gridded interpolation analysis is performed on the heat source distribution dataset to obtain a dynamically updated heat source distribution map.

3. The high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces according to claim 1, characterized in that, The step of extracting and analyzing heat sources in abnormal areas based on the heat source distribution mapping to obtain multi-angle image data and heat source feature vectors includes: Based on the temperature field data and ventilation airflow parameters in the heat source distribution mapping, the correlation between temperature fluctuations and ventilation airflow is calculated to obtain abnormal temperature fluctuations; When the abnormal temperature fluctuation exceeds a preset fluctuation threshold, the abnormal temperature fluctuation is divided into network interpolation regions to obtain the abnormal heat source region. Extract multi-angle image data corresponding to the abnormal heat source region; The multi-angle image data are subjected to data denoising and enhancement processing respectively to obtain preprocessed image data; Texture and brightness features are extracted from the preprocessed image data, and a heat source feature vector is generated by combining the support vector machine algorithm.

4. The high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces according to claim 1, characterized in that, The step of performing fire source location layout analysis based on the multi-angle image data and the heat source feature vector to obtain the nozzle layout and relative distance matrix includes: Feature point matching and alignment are performed on the multi-angle image data to obtain registered image data; Based on the registered image data, multi-view image data integration is performed to obtain unified image data; Based on the unified image data and the heat source feature vector, spatial geometric calculations are performed to obtain the three-dimensional coordinates of the fire source. Based on the three-dimensional coordinates of the fire source, and combined with a preset three-dimensional model of the ship's internal structure, the nozzle layout is extracted. Perform Euclidean distance calculations on the three-dimensional coordinates of the fire source and the nozzle layout to obtain a relative distance matrix.

5. The high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces according to claim 4, characterized in that, The step of optimizing the nozzle direction and spray intensity parameters based on the nozzle layout and the relative distance matrix to obtain the fire extinguishing path configuration includes: When the relative distance in the relative distance matrix is ​​greater than a preset distance threshold, coordinate transformation is performed to adjust the three-dimensional coordinates of the fire source and the nozzle layout to obtain an adjustment distance matrix. The nozzle direction and spray intensity parameters are optimized on the adjusted distance matrix to obtain the fire extinguishing path configuration.

6. The high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces according to claim 1, characterized in that, When the spray intensity parameter in the fire extinguishing path configuration is greater than a preset equipment load threshold, a feedback loop iteration is executed to obtain an optimized sprinkler head control command sequence, including: When the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold, the number of iterations is adjusted based on the preset historical iteration data to obtain the updated number of iterations; Based on the number of update iterations, the spray intensity parameters are optimized to obtain an optimized nozzle control command sequence.

7. The high-pressure fine water mist fire extinguishing method for ship roll-on / roll-off spaces and special spaces according to claim 4, characterized in that, The step of transmitting the nozzle control command sequence to the actuator network to activate the corresponding nozzle, performing a spray range coverage determination, and obtaining the final nozzle control command sequence includes: Based on a preset nozzle coverage database, determine whether the spray coverage area includes the three-dimensional coordinates of the fire source, and obtain the coverage determination result; Based on the coverage determination result, the spray intensity parameter of the nozzle control command sequence is adjusted to obtain the final nozzle control command sequence.

8. A high-pressure fine water mist fire extinguishing system for ship roll-on / roll-off spaces and special spaces, characterized in that, include: The data acquisition module is used to acquire temperature field data and ventilation airflow parameters, and integrate the temperature field data and ventilation airflow parameters into a heat source distribution map; The anomaly analysis module is used to extract and analyze heat sources in abnormal areas based on the heat source distribution mapping, and obtain multi-angle image data and heat source feature vectors. The fire source layout module is used to perform fire source location layout analysis based on the multi-angle image data and the heat source feature vector to obtain the nozzle layout and relative distance matrix. The path configuration module is used to optimize the nozzle direction and spray intensity parameters based on the nozzle layout and the relative distance matrix to obtain the fire extinguishing path configuration; The optimization instruction module is used to perform feedback loop iteration to obtain an optimized nozzle control instruction sequence when the spray intensity parameter in the fire extinguishing path configuration is greater than the preset equipment load threshold. The final instruction module is used to transmit the nozzle control instruction sequence to the actuator network to activate the corresponding nozzle, perform spray range coverage judgment, and obtain the final nozzle control instruction sequence.

Images