A multi-source fusion-based oil storage pit flame retardation performance testing system and method

Abstract

The application discloses a kind of based on multi-source fusion's oil storage pit flame retardant performance test system and method, the system includes control execution module, data module, decision module and evaluation module;Control execution module is used to simulate oil storage pit fire scene;Data module is used to real-time acquisition in simulated fire scene under the generation of multimodal physical parameter;Decision module is used to preprocess the multimodal physical parameter, carries out space-time alignment and feature fusion analysis to the multimodal physical parameter after preprocessing, obtains the unified feature representation after fusion, generates comprehensive combustion situation chart, according to the comprehensive combustion situation chart, generates adaptive control instruction, and sends to control execution module, adjusts oil storage pit fire scene;Evaluation module is used to receive historical data and the unified feature representation after fusion, by constructing flame retardant performance prediction model output to the quantitative evaluation and prediction report of oil storage pit flame retardant performance.The application realizes the intelligent test of the multi-source data of oil storage pit flame retardant performance.

Description

Technical Field

[0001] This invention relates to the field of oil storage pit testing technology, specifically to a multi-source fusion-based oil storage pit flame retardant performance testing system and method. Background Technology

[0002] As a critical safety barrier for oil storage and transportation facilities, the flame-retardant performance of oil storage pits directly affects the operational safety of oil and gas storage and transportation systems. Traditional flame-retardant performance testing of oil storage pits often relies on monitoring single physical parameters, such as surface temperature measurement or visual observation of flames. This approach has several technical shortcomings, including: static test scenarios that fail to simulate the dynamic coupling effects of changes in flame intensity, ambient wind field, and oil circuit conditions during a real fire; limited data acquisition dimensions that fail to comprehensively capture multimodal information such as flame spread dynamics, internal combustion characteristics of the medium, and concentration of gaseous products, leading to a one-sided assessment of the combustion situation; and a disconnect between data processing and decision-making, lacking in-depth fusion analysis of multi-source data, making it difficult to generate accurate quantitative results of flame-retardant performance and limiting testing efficiency and evaluation accuracy due to the inability to dynamically adjust the test scenario based on real-time conditions. With increasing industrial safety requirements, existing technologies can no longer meet the needs for comprehensive, dynamic, and accurate testing of the flame-retardant performance of oil storage pits. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a multi-source fusion-based oil storage pit flame retardant performance testing system and method to address the issues of disconnection from actual fire scenarios and information limitations of single sensors in the prior art. This system can accurately reproduce the fire state of oil storage pits under different combustion intensities and wind speed interference, and achieve a visual and accurate characterization of flame spatial distribution, temperature gradient, and spread trend.

[0004] To solve the above technical problems, the present invention adopts the following technical solution:

[0005] A multi-source fusion-based flame retardant performance testing system for oil storage pits includes:

[0006] The module consists of a control execution module, a data module, a decision-making module, and an evaluation module.

[0007] The control execution module is used to simulate an oil storage pit fire scenario.

[0008] The data module includes a heterogeneous sensor array for real-time acquisition of multimodal physical parameters generated by the control execution module under simulated fire scenarios.

[0009] The decision module comprises a data fusion unit, an edge computing unit and an instruction unit; the data fusion unit is configured to receive and pre-process the multi-modal physical parameters to obtain pre-processed multi-modal physical parameters; the edge computing unit is configured to perform spatio-temporal alignment and feature fusion analysis on the pre-processed multi-modal physical parameters to obtain a unified feature representation after fusion and generate a comprehensive combustion situation map; and the instruction unit is configured to generate adaptive control instructions according to the comprehensive combustion situation map and send the adaptive control instructions to the control execution module to adjust the oil pit fire scene.

[0010] The evaluation module is configured to receive historical data and the unified feature representation after fusion, and output a quantitative evaluation and prediction report on the fire-retardant performance of the oil pit by constructing a fire-retardant performance prediction model.

[0011] As a preferred scheme of the fire-retardant performance testing system for an oil pit based on multi-source fusion, in the scheme:

[0012] The heterogeneous sensor array comprises an optical acquisition unit, an acoustic acquisition unit, a temperature acquisition unit and a gas acquisition unit.

[0013] The optical acquisition unit is configured to acquire two-dimensional flame vision and infrared radiation information.

[0014] The acoustic acquisition unit is configured to acquire acoustic signals of combustion.

[0015] The temperature acquisition unit is configured to measure temperature distribution and obtain temperature data.

[0016] The gas acquisition unit is configured to acquire combustion product gas concentration and obtain gas data.

[0017] As a preferred scheme of the fire-retardant performance testing system for an oil pit based on multi-source fusion, in the scheme:

[0018] The edge computing unit performs spatio-temporal alignment on the pre-processed multi-modal physical parameters, and the spatio-temporal alignment comprises timestamp synchronization of the pre-processed multi-modal physical parameters based on a unified time reference and spatial position mapping of the pre-processed multi-modal physical parameters based on a unified spatial coordinate system, to obtain multi-modal physical parameters after spatio-temporal alignment.

[0019] The edge computing unit performs feature fusion analysis on the multi-modal physical parameters after spatio-temporal alignment, and the feature fusion analysis comprises extracting feature parameters related to combustion from the multi-modal physical parameters after spatio-temporal alignment and fusing the feature parameters into a unified feature representation based on a preset fusion algorithm.

[0020] Based on the fused unified feature representation, a comprehensive combustion situation map is generated, which visually represents the three-dimensional spatial distribution, temperature field gradient and spreading dynamic trend of the flame inside and on the surface of the gravel medium layer of the oil storage pit.

[0021] As a preferred scheme of the oil storage pit flame-retardant performance test system based on multi-source fusion, wherein:

[0022] The spatio-temporal alignment comprises the following steps:

[0023] Step 1: Taking the geometric center of the oil storage pit as the origin, a three-dimensional spatial coordinate system containing the inside and surface of the gravel layer medium layer is established as a unified spatial reference datum of the preprocessed multi-modal physical parameters.

[0024] Step 2: A unified clock source is established to stamp a unified timestamp on the two-dimensional flame visual and infrared radiation information, the sound wave signal of combustion, the temperature data and the gas data.

[0025] Step 3: According to the installation position and viewing angle parameters of the optical acquisition unit, the two-dimensional flame visual and infrared radiation information are mapped into the three-dimensional spatial coordinate system; according to the embedding position of the acoustic acquisition unit inside the gravel medium layer, a corresponding relationship between the sound wave signal of combustion and the three-dimensional spatial coordinate system is established; the temperature data and the gas data are positioned to the acquisition position in the three-dimensional spatial coordinate system.

[0026] Step 4: The preprocessed multi-modal physical parameters corresponding to the same spatial region in the three-dimensional spatial coordinate system under the same timestamp are associated and integrated to generate the spatio-temporally aligned multi-modal physical parameters.

[0027] As a preferred scheme of the oil storage pit flame-retardant performance test system based on multi-source fusion, wherein:

[0028] The feature fusion analysis comprises the following steps:

[0029] Step 1: Extracting the characteristic parameters of combustion from the spatio-temporally aligned multi-modal physical parameters, the characteristic parameters including flame morphology and spreading dynamics characteristics, medium internal combustion characteristics, temperature field distribution characteristics and gas composition characteristics.

[0030] Step 2: Dimension reduction processing is performed on the characteristic parameters of combustion, and standardization processing is performed.

[0031] Step 3: A multi-layer perception model is constructed, and the standardized combustion characteristic parameters are input into the model for nonlinear transformation.

[0032] The multi-layer perception model sets multiple neurons and adopts a nonlinear activation function, extracts and combines nonlinear features in the normalized combustion feature parameters layer by layer, and generates a fused unified feature representation at the output layer, which maps multidimensional information of the oil storage pit combustion state.

[0033] Step 4, input the fused unified feature representation into the situation generation module to generate a comprehensive combustion situation map.

[0034] As a preferred scheme of the oil storage pit flame-retardant performance test system based on multi-source fusion of the present application, wherein:

[0035] The flame-retardant performance prediction model comprises an input layer, multiple hidden layers and an output layer; each hidden layer comprises multiple neurons.

[0036] The flame-retardant performance prediction model receives the fused unified feature representation through the input layer, performs linear combination operation on the unified feature representation through the hidden layer, and performs transformation through the nonlinear activation function; the number of neurons of the output layer is consistent with the number of flame-retardant performance indicators to be predicted, and the neurons of the output layer receive the feature representation output by the hidden layer and perform independent linear calculation to output the quantitative prediction value of the corresponding flame-retardant performance indicator.

[0037] As a preferred scheme of the oil storage pit flame-retardant performance test system based on multi-source fusion of the present application, wherein:

[0038] Based on the quantitative prediction value of the flame-retardant performance indicator, the flame-retardant performance indicator is classified according to the predetermined rating standard.

[0039] The flame-retardant performance indicator of the current test is compared and analyzed with the historical data in time sequence to generate a performance evolution trend prediction.

[0040] According to the performance evolution trend prediction, an optimization improvement suggestion is provided, which includes a cobblestone medium layer configuration scheme optimization suggestion and a structure design adjustment scheme; according to the optimization improvement suggestion, a quantitative evaluation and prediction report containing data visualization elements is output, the data visualization elements including performance indicator comparison charts, trend prediction curves and three-dimensional thermal distribution maps.

[0041] As a preferred scheme of the oil storage pit flame-retardant performance test system based on multi-source fusion of the present application, wherein:

[0042] The control execution module comprises a variable power flame injection unit, an oil line control unit and an environmental wind field simulation unit.

[0043] The variable power flame injection unit is used to inject multiple flames.

[0044] The oil line control unit is used to control the oil line of the oil storage pit.

[0045] The environmental wind field simulation unit is used for simulating an environmental wind field.

[0046] As a preferred scheme of the oil storage pit flame-retardant performance test system based on multi-source fusion, wherein:

[0047] The data fusion unit pre-processes the multi-modal physical parameters, and the pre-processing includes time synchronization, data format standardization and invalid data elimination.

[0048] Further, the application also proposes an oil storage pit flame-retardant performance test method based on multi-source fusion, comprising:

[0049] S1, a dynamic fire scene is constructed by a variable power flame injection unit, an oil line control unit and an environmental wind field simulation unit, and a fire burning state of the oil storage pit in the simulated fire scene is simulated.

[0050] S2, a heterogeneous sensor array composed of an optical acquisition unit, an acoustic acquisition unit, a temperature acquisition unit and a gas acquisition unit is used to acquire multi-modal physical parameters of the oil storage pit in the simulated fire scene in real time.

[0051] S3, the data fusion unit of the decision module receives the multi-modal physical parameters, pre-processes the multi-modal physical parameters by time synchronization, data format standardization and invalid data elimination, and outputs the pre-processed multi-modal physical parameters to the edge computing unit.

[0052] S4, the edge computing unit performs spatio-temporal alignment on the pre-processed multi-modal physical parameters, performs timestamp synchronization based on a unified time reference and space position mapping based on a unified space coordinate system, extracts combustion feature parameters from the spatio-temporal aligned multi-modal physical parameters through feature fusion analysis, and generates a comprehensive combustion situation map representing flame three-dimensional distribution, temperature gradient and spreading trend by fusing the multi-layer perception model into a unified feature representation.

[0053] S5, the instruction unit of the decision module generates adaptive control instructions according to the comprehensive combustion situation map and sends them to the control execution module.

[0054] S6, the evaluation module receives historical data and fused unified feature representation, constructs a flame-retardant performance prediction model including an input layer, a hidden layer and an output layer, and outputs flame-retardant performance index quantitative prediction values.

[0055] Compared with the prior art, the application has the following technical effects:

[0056] The application can dynamically adjust fire scene parameters, accurately reproduce the oil pit fire state under different combustion intensity and wind speed interference, solve the problem of disconnection between traditional static test and actual fire scene, and improve the test authenticity by controlling the variable power flame injection, oil circuit control and environmental wind field simulation function of the execution module.

[0057] The application adopts an array of optical, acoustic, temperature, gas and other heterogeneous sensors to synchronously collect multi-dimensional parameters such as flame shape, medium internal combustion sound wave, temperature field distribution and gas composition, breaks through the information limitation of a single sensor, and provides complete data support for combustion situation analysis.

[0058] The application is provided with a decision module, a data fusion unit, a preprocessing unit, an edge computing unit, a spatio-temporal alignment unit, a feature fusion unit, a multi-modal physical parameter conversion unit, a three-dimensional combustion situation map generation unit, a flame space distribution visualization unit, a temperature gradient visualization unit, a spread trend visualization unit, a prediction module, a quantitative index output unit, a grade division unit, a trend prediction unit and a structure optimization suggestion unit. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 It is the overall structure diagram of the system of the application.

[0060] Figure 2 It is the flowchart of the multi-modal physical parameters after spatio-temporal alignment of the application.

[0061] Figure 3 It is the flowchart of generating a comprehensive combustion situation map of the application.

[0062] Figure 4 It is the overall implementation flowchart of the method of the application. DETAILED DESCRIPTION

[0063] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the specific embodiments of the application will be described in detail below with reference to the accompanying drawings.

[0064] In the following description, many specific details are set forth in order to provide a thorough understanding of the application, but the application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the scope of the application, therefore the application is not limited to the specific embodiments disclosed below.

[0065] As used herein, the term "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the implementation can be included in at least one implementation of the disclosure. The appearances of the phrase "in one embodiment" or "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.

[0066] To achieve the above object, the present application provides a multi-source fusion-based oil storage pit fire resistance performance test system, as shown in the figure, comprising a control execution module, a data module, a decision module and an evaluation module. Figure 1 The control execution module is used to simulate the fire scene of the oil storage pit; specifically:

[0067] The control execution module comprises a variable power flame injection unit, an oil line control unit and an environmental wind field simulation unit.

[0068] The variable power flame injection unit is used to inject multiple flames.

[0069] The oil line control unit is used to control the oil line of the oil storage pit, and switches the inflow and outflow oil lines of the oil storage pit by receiving the adaptive control instructions of the instruction unit in the decision module, for changing the flame simulation scene.

[0070] The environmental wind field simulation unit is used to simulate the environmental wind field.

[0071] The data module comprises a heterogeneous sensor array, which is used to collect multiple modal physical parameters generated by the control execution module in the simulated fire scene in real time; specifically:

[0072] The heterogeneous sensor array comprises an optical acquisition unit, an acoustic acquisition unit, a temperature acquisition unit and a gas acquisition unit.

[0073] The optical acquisition unit is arranged above and around the scene simulation area, and is used to collect two-dimensional flame vision and infrared radiation information; the acoustic acquisition unit is arranged in a three-dimensional spatial distribution manner inside the cobblestone medium layer of the oil storage pit, and is used to collect sound wave signals of combustion; the temperature acquisition unit is arranged in a distributed configuration on the surface and different depths inside the cobblestone medium layer, and is used to measure temperature distribution and obtain temperature data; the gas acquisition unit is arranged above the scene simulation area, and is used to collect combustion product gas concentration and obtain gas data.

[0074]

[0075] ​The decision module comprises a data fusion unit, an edge computing unit and an instruction unit; the data fusion unit is configured to receive and pre-process the multi-modal physical parameters to obtain pre-processed multi-modal physical parameters; the edge computing unit is configured to perform spatio-temporal alignment and feature fusion analysis on the pre-processed multi-modal physical parameters to obtain a unified feature representation after fusion, generate a comprehensive combustion situation map; the instruction unit is configured to generate adaptive control instructions according to the comprehensive combustion situation map and send the adaptive control instructions to the control execution module to adjust the oil pit fire scene; specifically:

[0076] The data fusion unit pre-processes the multi-modal physical parameters to realize optimization of multi-source heterogeneous raw data in three dimensions of time alignment, format unification and quality reliability, and the pre-processing includes time synchronization, data format standardization and invalid data elimination; the specific content includes:

[0077] A unified time reference is established through a high-precision clock synchronization mechanism to ensure consistency of data collected by different collection units in the time dimension;

[0078] A time synchronization component is configured for the heterogeneous sensor array to keep the local clock of all devices in high-precision synchronization with the system master clock; a unified format time mark is added to the two-dimensional flame visual and infrared radiation information, the sound signal of combustion, the temperature data and the gas data to clearly indicate the accurate time of data collection; for the difference in sampling frequency of different collection units, data interpolation or down-sampling processing is performed to keep the interval of various data on the time axis consistent;

[0079] The two-dimensional flame visual and infrared radiation information is subjected to format conversion and key feature extraction; the information is image data, which is converted into structured flame shape parameters and associated with corresponding time marks; the sound signal of combustion is subjected to segmentation processing and feature extraction to convert the continuous sound wave stream into structured data containing time domain and frequency domain features; the temperature data is converted into intuitive physical quantity values, and a three-dimensional distribution matrix is constructed in combination with the spatial position of the temperature collection unit; the gas data is subjected to normalization processing to unify the dimension representation of different gas components and calculate the concentration trend parameter;

[0080] A targeted filtering algorithm is used to remove noise from the multi-modal physical parameters, smooth the high-frequency fluctuations caused by environmental interference, and retain the real physical characteristic changes. Based on statistical rules and cross-frame checking methods, abnormal data points are identified, including sudden changes in values beyond the normal range and logically inconsistent feature information. The identified abnormal data is repaired using interpolation methods. For time-continuous data, the trend is completed by adjacent valid data. For spatially distributed data, spatial interpolation is performed using the information of the surrounding acquisition units. Through the logical correlation of multi-modal physical parameters, consistency checking is performed to eliminate abnormal values that are inconsistent between different parameters. If a collection unit continuously outputs abnormal data, a fault alarm is triggered and the data channel is switched to the backup channel.

[0081] The pre-processed multi-modal physical parameters are time and space aligned by the edge computing unit. The time and space alignment includes establishing a unified time reference through a high-precision clock synchronization mechanism, synchronizing the timestamps of the pre-processed multi-modal physical parameters based on the unified time reference, and mapping the pre-processed multi-modal physical parameters based on a unified spatial coordinate system to obtain time and space aligned multi-modal physical parameters. As shown in Figure 2 The specific content includes:

[0082] Step 1: Establish a three-dimensional spatial coordinate system including the interior and surface of the cobblestone medium layer with the geometric center of the oil storage pit as the origin, as the unified spatial reference benchmark for the pre-processed multi-modal physical parameters.

[0083] Specifically, the length direction of the oil storage pit, i.e., from the oil inlet to the oil outlet, is defined as the x-axis, the width direction, i.e., the horizontal direction perpendicular to the length direction, is defined as the y-axis, and the vertical direction perpendicular to the bottom surface of the oil storage pit is defined as the z-axis, forming a right-handed screw coordinate system.

[0084] Three or more calibration marker points are set on the inner wall of the oil storage pit. The optical positioning module in the heterogeneous sensor array identifies the marker points in real time, dynamically calibrates the coordinate system, and ensures the deviation of the coordinate system from the physical structure of the oil storage pit.

[0085] The precise coordinates of the lens center of the optical acquisition unit, the array center point of the acoustic acquisition unit, the probe of the temperature acquisition unit, and the probe of the gas acquisition unit in the three-dimensional coordinate system are measured by the total station, and stored in the sensor parameter database of the data fusion unit as reference data for spatial mapping.

[0086] Step 2: Establish a unified clock source to apply a unified timestamp to the two-dimensional flame vision and infrared radiation information, the sound wave signal of combustion, temperature data, and gas data.

[0087] Specifically, a GPS synchronous clock module can be used as a master clock source to ensure the stability of the clock reference; meanwhile, a constant temperature crystal oscillator is configured as a backup clock source, which is automatically switched to other available devices when the GPS signal is lost.

[0088] A clock synchronization network is constructed, a master clock node is deployed in a core controller of the control execution module, and each acquisition unit is configured with a synchronization system slave clock module to realize real-time synchronization with the master clock.

[0089] When each acquisition unit collects data, the local synchronization system slave clock module generates an accurate time stamp in the format of xx year xx month xx day xx hour xx minute xx second, which is embedded in the original data through the data frame header field to ensure the forced association of time information and measurement data.

[0090] Step 3. According to the installation position and viewing angle parameters of the optical acquisition unit, map the two-dimensional flame vision and infrared radiation information to the three-dimensional space coordinate system; according to the embedding position of the acoustic acquisition unit inside the cobblestone medium layer, establish the correspondence between the sound signal of the combustion and the three-dimensional space coordinate system; and position the temperature data and gas data to the specific collection position in the three-dimensional space coordinate system.

[0091] The three-dimensional space mapping method comprises:

[0092] The optical acquisition unit is parameterized by Zhang's calibration method to obtain an intrinsic matrix, including focal length, principal point coordinates, distortion coefficient, extrinsic matrix rotation matrix and translation vector; the extrinsic matrix is determined by measuring the distance from the camera to the origin and the angles between the camera and the x-axis, y-axis and z-axis;

[0093] Based on the extrinsic matrix, the two-dimensional image pixel coordinates are converted into three-dimensional space coordinates, the image is corrected for distortion to eliminate the influence of lens distortion; the normalized image coordinates are calculated according to the row and column coordinates of the pixel points in the image in combination with the intrinsic matrix; the normalized image coordinates are mapped to the three-dimensional space coordinate system through inverse transformation of the extrinsic matrix to obtain three-dimensional point cloud data of the flame region; if multiple optical acquisition units are deployed, the iterative closest point algorithm is used to register and fuse the three-dimensional point cloud data of different viewing angles to eliminate the viewing angle blind area and improve the three-dimensional reconstruction accuracy;

[0094] The acoustic acquisition unit adopts an n×n three-dimensional array layout, the sound source position is calculated by measuring the time difference of the arrival of the sound wave signal at each array element, specifically, the Chan algorithm can be used to solve the nonlinear positioning equation set to realize the three-dimensional space positioning of the combustion sound source signal.

[0095] According to the sound wave propagation characteristics of the cobblestone medium layer, the sound wave propagation velocity correction curve at different depths is measured through pre-experiment, a velocity-depth mapping model is established, a depth compensation coefficient is introduced in the positioning calculation to correct the propagation time difference error.

[0096] The three-dimensional coordinates of the located combustion sound source signal are compared with the data of the temperature collection unit in the same area. When the temperature value corresponding to the sound source coordinates exceeds the ignition threshold, the sound wave signal is confirmed as an effective combustion signal, otherwise it is marked as an environmental interference signal and eliminated.

[0097] The probe position coordinates of each temperature collection unit are determined by the total station during installation and stored in the sensor parameter database. When the measurement data is output, the preset coordinate information is automatically associated to form a data structure of [time stamp, probe position coordinates, temperature value];

[0098] The gas collection unit uses distributed sampling probes. Each probe is associated with a three-dimensional coordinate system through a pipe number. The mapping relationship between the pipe number and the coordinates is stored in the sensor parameter database. When the gas data is output, the corresponding spatial coordinates are automatically matched through the pipe number to ensure the accurate correspondence between the concentration data and the collection position.

[0099] The spatial positioning accuracy of the temperature collection unit and the gas collection unit is calibrated periodically by standard heat sources and standard gas sources. If the deviation exceeds the rated value of the gas, the coordinates are re-measured and the sensor parameter database is updated.

[0100] Step 4: Associate and integrate the preprocessed multi-modal physical parameters corresponding to the same spatial region in the three-dimensional coordinate system under the same time stamp to generate multi-modal physical parameters after time and space alignment.

[0101] The three-dimensional space of the oil storage pit is divided into several spatial units according to the grid size of Mcm x Mcm x Mcm. Each unit is assigned a unique spatial index code in the format of x interval_y interval_z interval, which serves as the spatial reference for data association.

[0102] In addition, a time and space matching rule is set, which includes time matching and space matching, and the specific content includes:

[0103] Time matching is used to filter preprocessed multi-modal physical parameters with an actual time stamp difference not greater than the rated time stamp difference, which is considered as measurement data of the same time segment.

[0104] Space matching associates data through spatial index coding to map preprocessed multi-modal physical parameters to corresponding spatial units. If the spatial coordinates of a certain data fall within the boundary range of a unit, it is considered as a valid value.

[0105] According to the precision characteristic setting data of the acquisition unit, data weights are set, the weights include a weight of temperature data, a weight of two-dimensional flame vision and infrared radiation information, a weight of gas data, and a weight of a sound wave signal of combustion, a comprehensive characteristic value of a space unit is generated through a weighted fusion algorithm, and the weights can be dynamically adjusted according to a combustion stage;

[0106] When logical contradictions occur in the acquisition parameters of the same space unit, a secondary verification mechanism is started, the source of abnormal data is judged through query of a historical data correlation model, and suspicious data is marked for manual review.

[0107] The edge computing unit performs feature fusion analysis on the spatio-temporally aligned multi-modal physical parameters, including extracting feature parameters related to combustion from the spatio-temporally aligned multi-modal physical parameters, and fusing the feature parameters into a unified feature representation based on a preset fusion algorithm. Figure 3 As shown in the figure, specifically:

[0108] Step 1, extract the feature parameters of combustion from the spatio-temporally aligned multi-modal physical parameters, including flame morphology and spreading dynamics characteristics, medium internal combustion characteristics, temperature field distribution characteristics, and gas composition characteristics.

[0109] Among them, the flame morphology and spreading dynamics characteristics include area, perimeter, circularity, kinematics characteristics (including spreading direction and speed calculated based on optical flow method of continuous frame images), texture characteristics, and surface temperature distribution statistical characteristics (including maximum temperature, average temperature, and high temperature area) provided by infrared thermal images; the medium internal combustion characteristics include energy, amplitude, frequency domain characteristics (including frequency spectrum obtained by Fourier transform, main frequency), time-frequency domain characteristics (including frequency components at a specific time obtained by wavelet transform); the temperature field distribution characteristics include temperature values at different depths, temperature gradient, temperature rise rate, and heat flux density values; the gas composition characteristics include the concentrations of various gases and their rates of change over time.

[0110] Step 2, perform dimensionality reduction processing on the feature parameters of combustion to eliminate redundant information between features, and standardize the feature parameters of different dimensions; feature dimensionality reduction can select the most effective and representative features, reduce data dimension and computational complexity, and improve the efficiency and accuracy of the fusion model;

[0111] Principal component analysis, linear discriminant analysis, etc. can be used for dimensionality reduction, and the physical features most related to the flame retardant performance can be directly selected according to domain knowledge;

[0112] Step 3, construct a multi-layer perception model, and input the standardized combustion feature parameters into the model for nonlinear transformation.

[0113] The multi-layer perception model sets multiple neurons and adopts a nonlinear activation function, extracts and combines high-order nonlinear features in the standardized combustion feature parameters layer by layer, and generates a fused unified feature representation at an output layer, which maps multidimensional information of the combustion state of the oil storage pit.

[0114] In step 4, the fused unified feature representation is input into a situation generation module to generate a comprehensive combustion situation map, which visually represents the three-dimensional spatial distribution, temperature field gradient, and dynamic trend of the flame inside and on the cobblestone medium layer of the oil storage pit. Different colors represent different temperature regions, arrow vector diagrams represent the direction and speed of flame spread, and the depth or brightness of the color represents the severity of the combustion.

[0115] Based on the comprehensive combustion situation map, the instruction unit first determines the current fire scene risk level by extracting key indicators such as flame spread speed, high-temperature area, and gas concentration exceeding standard value; then, according to the risk level, it matches the control target, calls the preset control strategy library to generate quantitative adjustment instructions, and clearly defines the power increase / decrease amplitude of the variable power flame jet unit, the flow regulation parameters of the oil way control unit, and the wind speed and direction adjustment values of the environmental wind field simulation unit; After the instructions are sent to the control execution module in a standardized format, the temperature change and flame shape adjustment effect are checked in real time through the sensor feedback, and if the expected result is not achieved, a secondary instruction optimization is triggered to achieve precise dynamic adjustment of the oil storage pit fire scene.

[0116] The evaluation module is used to receive historical data and the fused unified feature representation, and outputs a quantitative evaluation and prediction report of the fire retardant performance of the oil storage pit by constructing a fire retardant performance prediction model; specifically:

[0117] The fire retardant performance prediction model includes an input layer, multiple hidden layers, and an output layer; the hidden layer is set to 3 layers, and the number of neurons is 256, 128, and 64 in sequence, following the progressive logic of high-dimensional feature compression, high-order feature extraction, and key feature focusing, and adapting to the nonlinear mapping requirements from 32-dimensional input features to output indicators;

[0118] A three-dimensional spatial grid McmxMcmxMcis used as the basic unit, and each grid contains combustion feature parameters; the fused unified feature representation is transmitted to the input layer of the fire retardant performance prediction model in the form of a matrix of [timestamp, spatial grid index, multi-modal feature vector], and the spatial grid index corresponds to the encoding of the x, y, and z intervals, ensuring that the input data strictly matches the spatiotemporal alignment benchmark;

[0119] The first hidden layer corresponds to the global correlation analysis of the three-dimensional space grid features, the second hidden layer focuses on the extraction of the combustion dynamic trend features, and the third hidden layer realizes the feature analysis of multi-index prediction; a LeakyReLU activation function is adopted to adapt to the nonlinear mutation characteristics of temperature, concentration and other features in the combustion process, and is used to process the nonlinear problems of acoustic signal distortion correction and temperature gradient calculation;

[0120] The training data set is constructed based on the historical test data of the sensor parameter database, each group of data is spatio-temporally aligned and feature-fused to generate labeled samples, and the sample label is the measured value of the artificial labeled flame retardant performance index; a mean square error loss function is used, and dynamic weights are set for different flame retardant performance indexes; the Adam optimizer is used as the optimization algorithm, the initial learning rate is set, the convergence accuracy is improved through the learning rate decay strategy, and an abnormal data checking mechanism is introduced in the training process to eliminate invalid samples;

[0121] The number of output layer neurons is consistent with the number of flame retardant performance indexes to be predicted, the neurons of the output layer receive the feature representation of the hidden layer output, and perform independent linear calculation to output the quantitative prediction value of the corresponding flame retardant performance index; wherein the flame retardant performance indexes to be predicted include quantitative parameters directly related to the safety of the oil storage pit: the ratio of the actual flame spread distance per unit time to the theoretical spread distance without flame retardant measures, the duration of maintaining the temperature of the key area of the oil storage pit (such as the surface and internal of the cobblestone layer) below the ignition point, the proportion of time for the concentration of flammable gas in the combustion product to drop below the safety threshold, the proportion of area of the cobblestone layer structure that does not occur thermal failure based on the temperature field distribution calculation, and the output format (for example, the high temperature tolerance time output is 50±10 seconds);

[0122] When the spatio-temporal alignment result of the new test data deviates from the model training data by more than a preset threshold, incremental training is started; real-time access to the characteristic parameters of combustion, and dynamically update the training data set through a sliding window.

[0123] Based on the quantitative prediction value of the flame retardant performance index, the flame retardant performance index is graded according to the predetermined rating standard;

[0124] Based on the core index, a rating system is constructed, and the weight distribution is flame spread inhibition rate, high temperature tolerance time, gas product control rate, and flame retardant structure integrity coefficient, which corresponds to the flame retardant performance index, and the grade threshold is set as excellent, good, qualified and unqualified, and the comprehensive score is obtained by weighted summation;

[0125] The deviation rate of the current index from the historical same scene data is calculated, the deviation rate = (current index measured value - historical same scene data mean) ÷ historical same scene data mean × 100%, and the deviation distribution is visualized through a heat map, and the areas with deviation anomalies are highlighted;

[0126] The performance evolution trend prediction is generated by comparing the current tested flame-retardant performance index with historical data in time sequence; the trend of continuous test data can be extracted by time series decomposition, and the performance evolution trend prediction is generated by combining the predicted value of the future index output by the flame-retardant performance prediction model;

[0127] Optimization improvement suggestions are provided according to the performance evolution trend prediction, the suggestions including cobblestone medium layer configuration scheme optimization suggestions and structure design adjustment schemes; according to the optimization improvement suggestions, a quantitative evaluation and prediction report containing data visualization elements is output, the data visualization elements including performance index comparison charts, trend prediction curves and three-dimensional thermal distribution maps;

[0128] The cobblestone medium layer configuration scheme optimization suggestions are: based on the high-temperature accumulation area in the three-dimensional thermal distribution map, it is suggested to increase the proportion of high-heat-conductivity cobblestones at the corresponding position to reduce local heat accumulation; according to the visualization result of the flame spread path, the main flame spread direction is determined by the flame movement vector, the thickness of the cobblestone layer is increased to improve the longitudinal flame-retardant ability; combined with the gas product control rate data, if the gas concentration in a certain area decreases slowly, it is suggested to optimize the cobblestone particle size distribution in this area to enhance the pore air permeability;

[0129] The performance index comparison chart can compare the four core indexes of the current test and the historical optimal data by using a radar chart, and the radar axis scale is set based on the rating standard threshold; the comprehensive rating results under different test scenarios are displayed by a column chart, the horizontal axis is marked with scene parameters, the vertical axis corresponds to the rating level, the column color coding is matched with the level, for example, green for excellent and yellow for good; the trend prediction curve has a time axis and a performance index value axis, the curve uses a dashed line to represent the predicted section and a solid line to represent the measured section, and the four indexes are distinguished by different colors; the three-dimensional thermal distribution map visualizes the internal temperature distribution of the cobblestone layer based on the temperature data after space-time alignment by using volume rendering technology, the temperature color scale gradually changes from blue to red, and the temperature profile at different depths can be viewed by interactive operation; the flame spread trajectory is extracted based on the time sequence change of the flame shape in the thermal map, the thickness of the trajectory line corresponds to the spread speed, and the weak flame-retardant area is intuitively displayed.

[0130] The application also provides a storage pit flame-retardant performance test method based on multi-source fusion, as shown in Figure 4 , comprising

[0131] S1, a dynamic fire scene is constructed by a variable power flame injection unit, an oil path control unit and an environmental wind field simulation unit to simulate the fire burning state of the storage pit;

[0132] S2, a heterogeneous sensor array composed of an optical acquisition unit, an acoustic acquisition unit, a temperature acquisition unit and a gas acquisition unit is used to acquire the multi-modal physical parameters of the storage pit in the simulated fire scene in real time;

[0133] S3, the data fusion unit of the decision module receives the multi-modal physical parameters, pre-processes, including time synchronization, data format standardization and invalid data elimination, and outputs the pre-processed multi-modal physical parameters to the edge computing unit;

[0134] S4, the edge computing unit performs spatio-temporal alignment on the pre-processed multi-modal physical parameters, performs timestamp synchronization based on a unified time reference and space position mapping based on a unified spatial coordinate system, a feature fusion analysis extracts combustion feature parameters from the spatio-temporal aligned multi-modal physical parameters, and fuses them into a unified feature representation through a multi-layer perception model to generate a comprehensive combustion situation map representing the three-dimensional distribution of flame, temperature gradient and spreading trend;

[0135] S5, the instruction unit of the decision module generates adaptive control instructions according to the comprehensive combustion situation map and sends them to the control execution module;

[0136] S6, the evaluation module receives historical data and fused unified feature representation, constructs a flame retardant performance prediction model including an input layer, a hidden layer and an output layer, and outputs a quantitative prediction value of the flame retardant performance index.

[0137] Importantly, it should be noted that the constructions and arrangements of the present application shown in the various different exemplary embodiments are illustrative only. Although only a few embodiments have been described in detail in this disclosure, many modifications are possible (e.g., variations in sizes, dimensions, structures, shapes and proportions of the various elements, values of parameters, mounting arrangements, use of materials, colors, orientations, etc.) without materially departing from the novel teachings and advantages of the subject matter described in this disclosure. For example, elements shown as integrally formed can be constructed of multiple parts or elements, the position of elements can be reversed or otherwise varied, and the nature or number of elements or positions can be altered or varied. Accordingly, all such modifications are intended to be included within the scope of the present application. The order or sequence of any process or method steps can be changed, or reordered, according to alternative embodiments. Any "means plus function" clauses are intended to cover the structures described herein as performing claimed functions and not only structural equivalents, but also equivalent structures. Other substitutions, modifications, changes, and omissions can be made in the design, operating conditions, and arrangement of the exemplary embodiments without departing from the scope of the present application. Accordingly, the present application is not limited to the particular embodiments described in this disclosure, but extends to any embodiments that would still fall within the scope of the appended claims.

[0138] Furthermore, in the interest of providing a concise description of exemplary embodiments, not all features of an actual implementation can be described, that is, those related to the

[0139] It will be appreciated that in the development of any actual implementation, as in any engineering or design project, numerous implementation-specific decisions can be made. Such development efforts might be complex and time-consuming, but would nevertheless be a routine undertaking of design, fabrication, and manufacture for those of ordinary skill having the benefit of this disclosure.

Claims

1. A multi-source fusion-based oil storage pit flame retardation performance test system, characterized in that, The application relates to a fire extinguishing performance evaluation system for oil storage pits. The system comprises a control execution module, a data module, a decision module and an evaluation module. The control execution module is used for simulating a fire scene in an oil storage pit. The data module comprises a heterogeneous sensor array which is used for collecting multi-modal physical parameters generated by the control execution module in the simulated fire scene in real time. The heterogeneous sensor array comprises an optical acquisition unit, an acoustic acquisition unit, a temperature acquisition unit and a gas acquisition unit. The optical acquisition unit is used for collecting two-dimensional flame vision and infrared radiation information. The acoustic acquisition unit is used for collecting sound wave signals of combustion. The temperature acquisition unit is used for measuring temperature distribution and obtaining temperature data. The gas acquisition unit is used for collecting combustion product gas concentration and obtaining gas data. The decision module comprises a data fusion unit, an edge computing unit and an instruction unit. The data fusion unit is used for receiving and preprocessing the multi-modal physical parameters to obtain preprocessed multi-modal physical parameters. The edge computing unit is used for performing time-space alignment and feature fusion analysis on the preprocessed multi-modal physical parameters to obtain fused unified feature representation and generate a comprehensive combustion situation map. The instruction unit is used for generating adaptive control instructions according to the comprehensive combustion situation map and sending the adaptive control instructions to the control execution module to adjust the fire scene in the oil storage pit. Specifically, the edge computing unit performs time-space alignment on the preprocessed multi-modal physical parameters. The time-space alignment comprises timestamp synchronization of the preprocessed multi-modal physical parameters based on a unified time reference and space position mapping of the preprocessed multi-modal physical parameters based on a unified space coordinate system to obtain time-space aligned multi-modal physical parameters. The edge computing unit performs feature fusion analysis on the time-space aligned multi-modal physical parameters. The feature fusion analysis comprises extracting combustion-related feature parameters from the time-space aligned multi-modal physical parameters and fusing the feature parameters into unified feature representation based on a preset fusion algorithm. The combustion-related feature parameters include flame shape and spreading dynamics characteristics, medium internal combustion characteristics, temperature field distribution characteristics and gas composition characteristics. Based on the fused unified feature representation, a comprehensive combustion situation map is generated. The comprehensive combustion situation map visually represents the three-dimensional space distribution, temperature field gradient and spreading dynamic trend of the flame in the internal and surface of the cobblestone medium layer of the oil storage pit. The control execution module comprises a variable power flame injection unit, an oil line control unit and an environmental wind field simulation unit. The variable power flame injection unit is used for injecting multiple flames. The oil line control unit is used for controlling the oil line of the oil storage pit. The environmental wind field simulation unit is used for simulating an environmental wind field. The evaluation module is used for receiving historical data and the fused unified feature representation, constructing a fire extinguishing performance prediction model and outputting a quantitative evaluation and prediction report of the fire extinguishing performance of the oil storage pit. The fire extinguishing performance prediction model comprises an input layer, multiple hidden layers and an output layer. Each hidden layer comprises multiple neurons. The flame-retardant performance prediction model receives the fused uniform feature representation through the input layer, performs linear combination operation on the uniform feature representation through the hidden layer, and transforms through the nonlinear activation function; the number of output layer neurons is consistent with the number of flame-retardant performance indicators to be predicted, the neurons of the output layer receive the feature representation output by the hidden layer, and perform independent linear calculation to output the quantitative prediction value of the corresponding flame-retardant performance indicator.

2. The multi-source fusion-based oil storage pit fire resistance performance test system according to claim 1, characterized in that, The spatio-temporal alignment includes the following steps: Step 1: Taking the geometric center of the oil storage pit as the origin, a three-dimensional space coordinate system containing the inside and surface of the cobblestone medium layer is established as a unified spatial reference datum of the preprocessed multi-modal physical parameters; Step 2: Establish a unified clock source to stamp a unified timestamp on the two-dimensional flame vision and infrared radiation information, the sound wave signal of combustion, temperature data and gas data; Step 3: According to the installation position and viewing angle parameters of the optical acquisition unit, the two-dimensional flame vision and infrared radiation information are mapped into the three-dimensional space coordinate system; according to the embedding position of the acoustic acquisition unit inside the cobblestone medium layer, the corresponding relationship between the sound wave signal of combustion and the three-dimensional space coordinate system is established; the temperature data and gas data are positioned to the collection position in the three-dimensional space coordinate system; Step 4: The preprocessed multi-modal physical parameters corresponding to the same spatial region in the three-dimensional space coordinate system at the same timestamp are associated and integrated to generate the multi-modal physical parameters after spatio-temporal alignment.

3. The multi-source fusion-based oil storage pit fire resistance performance test system according to claim 1, characterized in that, The feature fusion analysis includes the following steps: Step 1: Extracting the combustion feature parameters from the multi-modal physical parameters after spatio-temporal alignment; Step 2: Dimensionality reduction processing and standardization processing are performed on the combustion feature parameters; Step 3: Constructing a multi-layer perception model, inputting the standardized combustion feature parameters into the model, and performing nonlinear transformation; The multi-layer perception model sets multiple neurons and uses a nonlinear activation function to extract and combine nonlinear features in the standardized combustion feature parameters layer by layer, and generates a fused uniform feature representation in the output layer, which maps multi-dimensional information of the combustion state of the oil storage pit; Step 4: Input the fused uniform feature representation into the situation generation module to generate a comprehensive combustion situation map.

4. The multi-source fusion-based oil storage pit fire resistance performance test system according to claim 1, characterized in that, Based on the quantitative prediction value of the flame-retardant performance indicators, the flame-retardant performance indicators are graded according to predetermined rating standards; Perform time series comparison analysis on the current tested flame-retardant performance indicators and historical data to generate performance evolution trend prediction; According to the performance evolution trend prediction, optimization improvement suggestions are provided, which include cobblestone medium layer configuration scheme optimization suggestions and structure design adjustment schemes; According to the optimization improvement suggestions, output a quantitative evaluation and prediction report containing data visualization elements, including performance indicator comparison charts, trend prediction curves and three-dimensional thermal distribution maps.

5. The multi-source fusion-based oil storage pit fire resistance performance test system according to claim 1, characterized in that, The data fusion unit pre-processes the multi-modal physical parameters, and the pre-processing includes time synchronization, data format standardization and invalid data elimination.

6. The method applied to the multi-source fusion-based oil storage pit flame retardation performance test system according to any one of claims 1-5, characterized in that, It includes: S1, a dynamic fire scene is constructed by a variable power flame injection unit, an oil circuit control unit and an environmental wind field simulation unit to simulate the fire burning state of the oil storage pit; S2, a heterogeneous sensor array composed of an optical acquisition unit, an acoustic acquisition unit, a temperature acquisition unit and a gas acquisition unit is used to acquire multi-modal physical parameters of the oil storage pit in the simulated fire scene in real time; S3, the data fusion unit of the decision module receives the multi-modal physical parameters, performs preprocessing such as time synchronization, data format standardization and invalid data elimination, and outputs the preprocessed multi-modal physical parameters to the edge computing unit; S4, the edge computing unit performs spatio-temporal alignment on the preprocessed multi-modal physical parameters, synchronizes the timestamps based on a unified time reference and maps the spatial positions based on a unified spatial coordinate system, extracts the combustion feature parameters from the spatio-temporal aligned multi-modal physical parameters through feature fusion analysis, fuses them into a unified feature representation through a multi-layer perception model, and generates a comprehensive combustion situation map representing the three-dimensional distribution of the flame, the temperature gradient and the spreading trend; S5, the instruction unit of the decision module generates adaptive control instructions according to the comprehensive combustion situation map and sends them to the control execution module; S6, the evaluation module receives historical data and fused unified feature representation, constructs a flame retardant performance prediction model including an input layer, a hidden layer and an output layer, and outputs a quantitative prediction value of the flame retardant performance index.

Images