Fire-fighting garment comprehensive performance evaluation prediction method and system based on multi-source data fusion

Abstract

The application discloses a fire-fighting clothes comprehensive performance evaluation prediction method and system based on multi-source data fusion, and the method comprises the following steps: constructing a scoring database; determining the weight of each body part and calculating the single comprehensive score under each test working condition; by introducing a heat exposure time weight function, a liquid threat weight function and a motion weight coefficient, a heat exposure ergonomics index, a heat coupling liquid resistance index and a heat coupling particle resistance index are respectively calculated; a fire-fighting clothes comprehensive performance prediction model is constructed and trained; the fire-fighting clothes comprehensive performance prediction model is used to output multi-dimensional performance indexes, and the system recommendation order of the fire-fighting clothes is obtained; the application has the advantages that the multi-dimensional performance of the fire-fighting clothes in a compound disaster environment is predicted, so as to evaluate whether the fire-fighting clothes is usable, and the safety of the firemen is ensured.

Description

Technical Field

[0001] This invention relates to the field of performance testing and intelligent evaluation technology for fire suits, specifically to a method and system for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion. Background Technology

[0002] Firefighting suits are core personal protective equipment that ensures the safety of firefighters in extreme fire environments. Modern fire scenes are complex, often accompanied by multiple combined threats such as high-temperature burning, chemical liquid splashes, and dense smoke and particulate matter. Therefore, the evaluation of firefighting suit performance must go beyond a single indicator and involve a comprehensive, systematic, and multi-dimensional assessment of disaster scenarios.

[0003] Currently, mainstream evaluation methods have significant limitations: existing standards involve independent testing of fire suits for flashover, liquid tightness, and particulate matter leakage, resulting in isolated evaluation dimensions. The evaluation results are fragmented and fail to reflect the overall protective performance of fire suits when subjected to consecutive high-temperature burning, liquid, and smoke particle attacks in actual fire scenes.

[0004] Furthermore, existing evaluation methods are insufficient for handling real-world, complex disaster scenarios. For example, thermal protection tests primarily assess the insulation value of clothing or fabrics. However, in actual combat, clothing hardens and shrinks after being burned, significantly hindering personnel movement. This inevitably impacts the clothing's resistance to liquids and particulate matter, a critical change for which objective evaluation methods are currently lacking. The specific ways in which the protective capability of clothing diminishes when facing threats of varying comprehensive disaster intensities remain a blind spot in safety assessments. Finally, existing methods rely on actual experiments, which are costly and time-consuming, making it impossible to quickly predict the performance of untested conditions or new fire suit designs. They cannot identify the specific weaknesses leading to performance deficiencies (such as which body part or under what severe conditions), hindering guidance for equipment improvement and selection. For instance, the fire suit environmental adaptability assessment method and system based on AHP-fuzzy comprehensive evaluation disclosed in Chinese Patent Publication No. CN114372682A cannot perform rapid performance prediction or identify the specific weaknesses leading to performance deficiencies.

[0005] Therefore, there is an urgent need for a method and system that can objectively, comprehensively, and intelligently evaluate and predict the multi-dimensional performance of firefighting suits in complex disaster environments. Summary of the Invention

[0006] The technical problem to be solved by this invention is how to predict the multi-dimensional performance of fire suits in complex disaster environments, so as to assess whether the fire suits are usable and ensure the safety of firefighters.

[0007] This invention solves the above-mentioned technical problems through the following technical means: This invention provides a method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion, comprising:

[0008] S1. The performance scores of fire suits under multiple body parts and multiple test conditions are obtained through the whole-body thermal exposure efficiency evaluation test, the whole-body thermal coupling anti-penetration characteristic evaluation test, and the whole-body thermal coupling anti-particle characteristic evaluation test, and are mapped to standard levels to build a scoring database. S2. Based on the scoring database, determine the weight of each body part and calculate the single comprehensive score under each test condition; based on the single comprehensive score, by introducing the heat exposure time weight function, the liquid threat weight function and the action weight coefficient, calculate the heat exposure efficiency index, the heat coupling liquid resistance index and the heat coupling particulate resistance index respectively. S3. Construct a comprehensive performance prediction model for fire suits using a multi-output regression multilayer perceptron model. Use several pre-selected features related to the comprehensive performance of fire suits as model inputs, and use the true values ​​of thermal exposure efficiency index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index as model output labels to train the model and obtain a trained comprehensive performance prediction model for fire suits. S4. Using the comprehensive performance prediction model of the fire suit, the predicted thermal exposure efficiency index, thermal coupling liquid resistance index, and thermal coupling particle resistance index are output. Different weight ratios are set according to different application scenario types to fuse the three performance indices of each fire suit to obtain a weighted comprehensive score. The fire suits are sorted in descending order according to the weighted comprehensive score to obtain the system recommended ranking.

[0009] Furthermore, the overall thermal exposure ergonomics evaluation test includes: Fire suits that have undergone flashover pretreatment for different durations were selected as test samples. Subjects of a predetermined body type wore the test samples and performed predetermined activities in a predetermined test environment. The overall physical restraint of key parts of the human body was scored.

[0010] Furthermore, the evaluation test of the overall thermal coupling anti-permeability characteristics includes: Firefighting suits pretreated with flashover were worn on mannequins, and liquid splash tests were conducted at different pressures and times. The staining rate of different parts of the colored underwear was analyzed and converted into standardized scores.

[0011] Furthermore, the evaluation test of the overall thermal coupling particulate protection characteristics includes: The fire suits pretreated with flashover were worn by the subjects, who performed prescribed actions of varying intensities inside the aerosol testing chamber. The leakage rate of each part of the suit was monitored in real time and compared with the benchmark data of the leakage rate of an untreated parallel sample. The comparison results were then converted into standardized scores.

[0012] Further, S2 includes: Construct the original rating matrix, which represents the behavioral test samples. The columns of the original rating matrix represent various body parts, and the elements of the original rating matrix correspond to the rating of each test sample at the corresponding body part. For each test condition, the arithmetic mean of the scores of all subjects or dummies at each body part under that test condition is calculated to form an average score vector. The average score vectors corresponding to all test conditions form an average score matrix. The CRITIC method was used to analyze the average score matrix and calculate the weight of each body part to obtain the body part weight vector. For each test condition, the average score vector and body part weight vector of that test condition are used to calculate a single comprehensive score representing the overall efficiency of that test condition through weighted fusion. A weighting function is introduced to calculate the original weights corresponding to each test condition, and then normalization is performed to obtain the normalized weights of each test condition. The final performance index is obtained by weighting and summing the single comprehensive scores of each test condition with their corresponding normalized weights. When the test condition is the flashover time and the weighting function is the heat exposure time weighting function, the calculated index is the heat exposure effectiveness index. When the test condition is a three-dimensional condition consisting of flashover time, liquid pressure, and injection time, and the weighting function is the liquid threat weighting function, the calculated index is the thermal coupling liquid resistance index. When the test condition is a three-dimensional condition consisting of flashover time, action condition, and environmental condition condition, and the weighting function is the product of the heat exposure time weighting function corresponding to the flashover time and the action weighting coefficient, the calculated index is the thermal coupling particulate resistance index.

[0013] Furthermore, the heat exposure time weighting function is: ,in, Indicates the time of ignition The original weight of the corresponding heat exposure time, This is the preset critical time. This is the control factor.

[0014] Furthermore, the liquid threat weighting function is: ,in, Indicates test conditions The corresponding original weights of the liquid threat. Indicates pressure conditions and injection time The corresponding liquid splash weight and , , , To normalize the liquid pressure, and These are the preset minimum and maximum liquid pressures, respectively. To normalize the splash time, and These are the preset minimum and maximum splash times, respectively. This is a stress index that characterizes the degree to which stress contributes to a threat. This is a time index, representing the degree to which splash time contributes to the threat.

[0015] Further, S3 includes: Multiple features related to the comprehensive performance of fire suits were selected as inputs to a multi-output regression multilayer perceptron model. The thermal exposure ergonomics index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index were used as labels to construct the input feature set. The multi-output regression multilayer perceptron model consists of a sequentially connected input layer, three intermediate hidden layers, and an output layer with three output nodes. The hidden layers employ a topology design with progressively decreasing neuron counts, and each layer uses a modified linear unit as a non-linear activation function. The three neurons in the output layer use linear activation functions, corresponding to the predicted values ​​of the thermal exposure ergonomics index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index, respectively. The model was trained using the input feature set, with mean squared error used as the loss function during training. The model parameters were adjusted, and training stopped when a preset number of iterations or the minimum loss function was reached, resulting in a trained model. Periodic validation was implemented during the training phase; if the validation loss did not decrease within consecutive preset iterations, an early stopping mechanism was triggered.

[0016] Furthermore, S2 also includes a contribution tracing analysis based on the contribution of each test condition to the final performance, locating the failure conditions and weakest points of each protection dimension. The specific process is as follows: Under each test condition, the difference between the base score and the single comprehensive score under each flashover time is calculated. The product of the difference and the normalized time weight is used as the failure value. The maximum failure value under all flashover times is used as the failure condition. Under the test condition corresponding to the failure condition, the body part score is calculated by multiplying the arithmetic mean of the scores of each body part by the weight of each body part. The body part with the lowest score is used as the weakest body part. The base score is the single comprehensive score corresponding to the flashover time of 0.

[0017] This invention also provides a comprehensive performance evaluation and prediction system for fire suits based on multi-source data fusion, comprising the following steps: The scoring database construction module is used to obtain the performance scores of fire suits under multiple body parts and multiple test conditions through the whole-pack thermal exposure efficiency evaluation test, the whole-pack thermal coupling anti-permeability evaluation test, and the whole-pack thermal coupling anti-particle characteristic evaluation test, and map them to standard levels to build a scoring database. The multidimensional performance index calculation module is used to determine the weight of each body part based on the scoring database and calculate the single comprehensive score under each test condition; based on the single comprehensive score, by introducing the heat exposure time weight function, the liquid threat weight function and the action weight coefficient, the heat exposure efficiency index, the heat coupling liquid resistance index and the heat coupling particulate resistance index are calculated respectively. The performance prediction model building module is used to build a comprehensive performance prediction model for fire suits using a multi-output regression multilayer perceptron model. It takes several pre-selected features related to the comprehensive performance of fire suits as model inputs, and the real values ​​of thermal exposure efficiency index, thermal coupling liquid resistance index and thermal coupling particulate protection index as model output labels to train the model and obtain a trained comprehensive performance prediction model for fire suits. The intelligent prediction and diagnosis output module is used to output predicted thermal exposure efficiency index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index using the comprehensive performance prediction model of the fire suit. Different weight ratios are set according to different application scenario types to fuse the three performance indices of each fire suit to obtain a weighted comprehensive score. The fire suits are sorted in descending order according to the weighted comprehensive score to obtain the system recommended ranking.

[0018] The advantages of this invention are: (1) This invention obtains the performance scores of fire suits under multiple body parts and multiple test conditions through whole-assembly thermal exposure efficiency evaluation test, whole-assembly thermal coupling anti-permeability evaluation test and whole-assembly thermal coupling anti-particle characteristic evaluation test, and maps them to standard levels to construct a scoring database. Based on the scoring database, the thermal exposure efficiency index, thermal coupling liquid resistance index and thermal coupling anti-particle index are calculated respectively to predict the multi-dimensional performance of fire suits in complex disaster environments. According to different application scenario types, different weight ratios are set to fuse the three performance indices of each fire suit to obtain a weighted comprehensive score. Each fire suit is sorted in descending order according to the size of the weighted comprehensive score to obtain the system recommended ranking. In this way, it is possible to evaluate whether the fire suit is usable, ensure the safety of firefighters, and also locate the failure conditions and the weakest body parts of each protection dimension, which is convenient for rectification.

[0019] (2) The comprehensive performance evaluation method for fire suits provided by this invention no longer relies on single, isolated laboratory test indicators. Instead, it introduces dynamic, multi-dimensional performance degradation characterization under complex attack scenarios (such as time value weight and thermal-humid coupling amplification effect) as the core evaluation idea. This solves the problem that existing technologies cannot scientifically evaluate the residual comprehensive protective performance of fire suits in real complex fire scenes, and improves the objectivity, foresight, and accuracy of the evaluation. The integrated "experiment-evaluation-prediction-diagnosis" system constructed by this invention integrates objective experimental data, subjective perception scores, environmental coupling effects, and data-driven source analysis. It realizes data fusion and intelligent modeling of three dimensions: physical performance, chemical barrier performance, and ergonomic perception, which significantly improves the model's generalization ability, prediction accuracy, and decision support value.

[0020] (3) This invention utilizes a multi-output regression multilayer perceptron algorithm to process high-dimensional features and perform multi-task learning, which can effectively model the complex nonlinear mapping relationship between input parameters and three performance dimensions, and realize rapid and intelligent prediction and accurate diagnosis of shortcomings in the comprehensive performance of fire suits. When facing actual disaster relief scenarios, input feature parameters can be used to quickly respond and output a suitable fire suit selection scheme.

[0021] (4) In existing research methods, it is generally believed that the weights for flashover experiments at different times are consistent, without actively assigning dynamic weights. This overlooks a key issue: when the heat exposure time exceeds the critical value t0 of the clothing, the overall state of the fire suit will show obvious hardening and shrinkage, and its protective performance will change drastically, with a cliff-like decline in performance. This invention uses a heat exposure time weighting function to quantify the weights of these conditions. When t < t0, the weight increases slowly; when t > t0, the weight increases rapidly, amplifying the weight of extreme conditions exceeding the critical time, making this part of the performance have a greater impact on the final index. This is designed specifically for the thermal aging characteristics of fire suits. A parameter control coefficient k2 is designed in this function. The larger the value of k2, the steeper the function curve is near t0, which means that the burn area of ​​the fire suit increases sharply near the critical time point, and the thermal protection performance declines more rapidly. This is applicable to materials with high heat sensitivity. Once the critical temperature is exceeded, the material will deform and shrink rapidly, leading to a cliff-like decline in performance. A smaller k2 value results in a flatter function curve near t0, meaning that the burn area of ​​the fire suit increases more slowly near the critical time point, and the performance degradation is gradual. This applies to materials with good thermal stability, where the performance decline is relatively gradual.

[0022] (5) The liquid threat weighting function of this invention considers the weights of three-dimensional working conditions (heat exposure, liquid pressure, and liquid splashing time) rather than considering a single factor. This helps to realize comprehensive weight calculation in heat-liquid coupling scenarios, enabling the model to accurately assess the residual liquid resistance of fire suits under the real fire sequence of "heat exposure first, then liquid splashing". The weighting function is constructed using an exponential function, which can effectively depict the product coupling of two strongly correlated factors in the fluid penetration process of fire suit fabrics. At the same time, when clothing undergoes heat exposure, the fabric properties will inevitably change, such as increased porosity leading to decreased impermeability, and heat exposure causing loosening or cracking at the seams. These changes significantly increase the liquid penetration sensitivity of the clothing. From a mathematical point of view, compared with intact clothing, this creates an amplification effect: when pressure or time increases, the weight value increases exponentially, which can accurately reflect the increased sensitivity of clothing to liquid threats after heat damage.

[0023] (6) This invention constructs an original scoring matrix. The matrix form facilitates subsequent multi-source data fusion analysis. Each row corresponds to the actual perception of a single subject, avoiding the loss of individual variation information due to direct averaging. This is beneficial for capturing the fluctuations of each part among different individuals during subsequent objective weighting. The average score of each part is calculated at each flashover time. Through averaging, the data reflects the overall trend of clothing performance at that time point, improving the stability and representativeness of the evaluation and eliminating interference from individual differences. The objective weighting of body parts based on the CRITIC method can automatically reduce the weight of redundant indicators that are highly correlated with other indicators, avoiding double scoring and making the final weight allocation more accurately reflect the independent contribution of each body part. The weights are automatically calculated based on measured data, eliminating interference from human factors and ensuring the objectivity and reproducibility of the evaluation results. By combining the CRITIC weights with the scoring status of each part, the weakest body part can be accurately located, providing a direct reference for equipment improvement.

[0024] (7) This invention intuitively reflects the degree of performance degradation after the critical time by calculating each efficiency index, which can be directly used for horizontal comparison between fire suits, making it convenient to qualitatively determine the performance of fire suits. By quantifying the contribution of each working condition to the final efficiency index in the form of a product, the identification of the failure of clothing performance in different parts has a clear quantitative basis, clearly indicating "under what conditions the performance is worst" and "which part fails first", providing direct guidance for equipment improvement and targeted protection, and has application value. Attached Figure Description

[0025] Figure 1 This is a flowchart of the comprehensive performance evaluation and prediction method for fire suits based on multi-source data fusion disclosed in an embodiment of the present invention; Figure 2This is a model architecture diagram of the comprehensive performance evaluation and prediction method for fire suits based on multi-source data fusion disclosed in the embodiments of the present invention; Figure 3 This is a flowchart illustrating a specific application of the fire suit comprehensive performance evaluation and prediction method based on multi-source data fusion disclosed in this invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, 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.

[0027] Example 1 like Figure 1 As shown, Embodiment 1 of the present invention provides a method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion, comprising the following steps: S1: Through the whole-body thermal exposure ergonomics evaluation test, the whole-body thermal coupling liquid resistance evaluation test, and the whole-body thermal coupling particulate protection evaluation test, the original performance data of the fire suit under multiple body parts and multiple test conditions are obtained. After standardization processing, a unified multi-dimensional scoring database is constructed.

[0028] S2: Based on the scoring database, the CRITIC objective weighting method is used to determine the weight of each body part in each protection dimension, and the fuzzy comprehensive evaluation method is used to calculate the single comprehensive score under each single test condition; by introducing the heat exposure time weight function, liquid threat weight function and action weight coefficient, the heat exposure efficiency index (HTEI), heat coupling liquid resistance index (HCRI) and heat coupling particulate resistance index (HPRI) are calculated respectively to form a multi-dimensional performance output set after heat exposure; at the same time, based on contribution analysis, the most severe failure conditions and the weakest body parts in each dimension are located, and a performance evaluation rule base is constructed.

[0029] S3: Using the test conditions in the scoring database as features and the calculated three-dimensional performance indices (HTEI, HCRI, HPRI) as labels, construct and train a comprehensive performance prediction model for fire suits using the multi-output regression multilayer perceptron (MLP) algorithm.

[0030] S4: Using the trained prediction model, output the predicted thermal exposure efficiency index, thermal coupling liquid resistance index, and thermal coupling particulate protection index. According to different application scenario types, set different weight ratios to fuse the three performance indices of each fire suit to obtain a weighted comprehensive score. Sort each fire suit in descending order according to the weighted comprehensive score to obtain the system's recommended ranking. Perform rapid performance prediction and intelligent diagnosis for new fire suit designs or unknown working conditions and mission scenarios, and output a comprehensive report including quantitative performance index, comprehensive protection level, and root cause analysis.

[0031] The following details the calculation of the Heat Exposure Efficiency Index (HTEI), the Heat Coupled Liquid Resistance Index (HCRI), and the Heat Coupled Particulate Resistance Index (HPRI), as well as the corresponding traceability process.

[0032] 1. To evaluate the retention of functional activity of firefighting suits after heat exposure, a whole-suit heat exposure efficiency evaluation test was designed and implemented. The calculation and source tracing (contribution analysis) process of the Heat Exposure Efficiency Index (HTEI) is as follows: 1) Construction of the original rating matrix Sample pretreatment: Samples of fire suits from the same batch were selected and subjected to heat exposure pretreatment for different durations using a standard flashover testing apparatus (heat flux density 84 kW / m²). The flashover time set was set as: T = {0, 1, 2, 3, 4, 5, 6, 7, 8} seconds, with a total of 9 levels. Fifteen independent repeated flashover tests were conducted for each flashover time. The 0-second sample served as an untreated control baseline.

[0033] Subjects and Evaluation Sites: Fifteen healthy male volunteers meeting firefighter anthropometric standards were recruited as subjects. A set of 10 key body parts to be evaluated was defined: P = {head and neck, chest, back, shoulders, elbows, waist, hips, knees, thighs, and calves}.

[0034] Experimental Procedure and Scoring: Under standard environmental conditions (20°C, 65% RH), each subject was randomly assigned to wear clothing samples pretreated for different flashover times and complete a prescribed sequence of standard activity postures (including standing, walking, bending over, squatting, running, and jumping). Subjects provided real-time subjective scores based on the physical restriction felt in each area, using a scale of 1-10 (1 point: completely unable to move; 10 points: freely mobile).

[0035] Construction of the original scoring matrix: All experimental data were organized into an original scoring matrix X. The rows of this matrix correspond to each experimental sample (9 flashover times × 15 subjects = 135 rows), and the columns correspond to 10 body parts. Matrix elements x ij Record the score of the i-th sample at the j-th location. As shown in Table 1, the complete flashover time is presented. Examples of raw scores from the subjects and other raw scores from the blast time.

[0036] Table 1. Examples of some original scores from subjects at different flashover times.

[0037] The original rating matrix mentioned above is part of the rating database. The experimental acquisition process of the original rating matrix is ​​not an improvement of this invention, but rather the basic data of this invention. In practical applications, the original rating matrix can be given directly.

[0038] 2) Calculate the average score for each part at each flashover time. For each flashover time Calculate the arithmetic mean of the scores for each body part of all subjects at that time point to form an average score vector. .in, Indicates the time of ignition The average score for the Nth body part. All The corresponding average rating vectors form the average rating matrix S (9 rows × 10 columns). The average rating matrix is ​​shown in Table 2: Table 2 Example of Average Rating Matrix

[0039] 3) Analyze the average rating matrix S using the CRITIC method to calculate the weight vector for each body part, including the following sub-steps: The average rating matrix S is standardized to obtain the standardized matrix Z. Calculate the standard deviation of each column (i.e., each body part) of the standardized matrix. This serves as the contrast intensity for that area; Calculate the correlation coefficient matrix between each column, and then calculate the conflict index for each part: ,in, Indicates that the j-th body part is related to the j-th body part. The Pearson correlation coefficient between body parts measures the degree of linear correlation between scores of two body parts. Its value ranges from [ ]. [1,1], the closer the absolute value is to 1, the stronger the correlation between the two parts; the closer it is to 0, the weaker the correlation. By analyzing all... Summing the parts, we get The larger the value, the lower the overall correlation between this part and other parts, the more independent information it carries, and the higher the weight it should be given in objective weighting.

[0040] Calculate the amount of information in each part ; The information content is normalized to obtain the weights of each part. ,satisfy

[0041] The weighting results are shown in Table 3. The body part weight vector is obtained as follows: w = [0.0792, 0.0555, 0.0555, 0.0772, 0.1184, 0.0724, 0.0313, 0.3978, 0.0324, 0.0803].

[0042] Table 3: Examples of Objective Weighting Calculation Results Using the CRITIC Method

[0043] 4) Calculate the single comprehensive score for each flashover time. For each ignition time Using the average rating vector at that time point The weighted vector w of body parts is used to calculate a single comprehensive score representing the overall work efficiency at that point in time, through weighted fusion. As shown in Table 4, Indicates the time of ignition Next Average score for each body part.

[0044] Table 4. Examples of Single-Instance Comprehensive Scores for Each Flashover Time

[0045] 5) Define the heat exposure time weighting function and calculate the normalized weights. A sigmoid function is introduced as a weighting function for thermal exposure time to characterize the importance of the nonlinear growth of longer flashover tolerance time:

[0046] in, Indicates the time of ignition The original weight of the corresponding heat exposure time, The critical time is preset based on the inflection point of the thermal protection performance degradation of clothing materials. These are control coefficients. According to Table 4... The decay trend was determined to have a significant inflection point around t=4.0 seconds, therefore, we take... Calculate the original weights of the heat exposure time for each flashover time. Then, normalization is performed to obtain the normalized weights for heat exposure time: In this embodiment, there are a total of 9 flashover time points (0s, 1s, 2s, 3s, 4s, 5s, 6s, 7s, 8s), so k=1,2,…,9, corresponding to the above 9 time points respectively. The normalized weights of the heat exposure time are shown in Table 5.

[0047] Table 5 Example of normalized heat exposure time weight calculation

[0048] 6) Calculate the Heat Exposure Efficiency Index (HTEI) The overall score for each explosion duration Its corresponding normalized weight for heat exposure time By performing a weighted summation, the final heat exposure ergonomics index is obtained: .

[0049] The calculation process and results are shown in Table 6. The calculation result is: the heat exposure efficiency index (HTEI) of the fire suit is 4.08 (out of 10).

[0050] Table 6: Example of calculation process

[0051] 7) Source analysis of contribution According to the contribution tracing method proposed in this invention, an in-depth analysis of HTEI is conducted to locate the most severe failure condition and the weakest body part. Under each test condition, the difference between the baseline score and the single comprehensive score at each flashover time is calculated. The product of this difference and the normalized weight is taken as the failure value. The maximum failure value across all flashover times is taken as the failure condition. Under the test condition corresponding to the failure condition, the body part score is calculated by multiplying the arithmetic mean of the scores for each body part by the weight of each body part. The body part with the lowest score is taken as the weakest body part. The baseline score is the single comprehensive score corresponding to a flashover time of 0. Specifically, in this embodiment, the most severe failure condition under the heat exposure ergonomics test condition is located: Analysis revealed that, The failure value was 1.434 seconds (as shown in Table 7), indicating that the negative effect on the overall HTEI was most significant when the flashover time reached 8 seconds. This represents the basic score for the thermal exposure ergonomics test conditions.

[0052] Table 7 Examples of the most severe failure conditions

[0053] Identifying the weakest body parts: during the critical flashover time Instantly, combined with the scores of each part. (Table 2) and weights (Table 3) Calculate the weighted score for each part. .

[0054] Table 8: Contribution analysis of each part per second

[0055] As shown in Table 8, the elbow had the lowest weighted score (0.118). Therefore, the elbow was identified as the weakest part of the body.

[0056] 2. A certain model of fire-fighting suit was used for thermal coupling liquid resistance evaluation. The calculation and tracing process of the thermal coupling liquid resistance index (HCRI) is as follows: 1) Construction of the original rating matrix The experiment was designed as a three-dimensional coupled test: nine key flashover pretreatment times (0s, 1s, 2s, 3s, 4s, 5s, 6s, 7s, 8s), four liquid pressure levels (50 kPa, 100 kPa, 150 kPa, 200 kPa), and six injection times (10s, 20s, 30s, 40s, 50s, 60s) were selected, resulting in 216 test conditions. Each condition was tested three times. Ten key body parts were defined for evaluation: P = {head and neck, chest, back, shoulder, elbow, waist, hip, knee, thigh, calf}. After the experiment, the permeability of each body part was converted into a ten-point score.

[0057] Construction of the original scoring matrix: All experimental data were organized into an original scoring matrix X. The rows of this matrix correspond to each experimental sample (9 flashover times × 4 pressure conditions × 6 injection times × 3 replicates = 648 rows), and the columns correspond to 10 body parts. Matrix elements x ij Record number i The sample at the th j The scores for each part are shown in Table 9, which presents the original scores of partial staining rates under different working conditions in this embodiment.

[0058] Table 9. Examples of raw scores for dummy staining rate

[0059] The original rating matrix mentioned above is part of the rating database. The experimental acquisition process of the original rating matrix is ​​not an improvement of this invention, but rather the basic data of this invention. In practical applications, the original rating matrix can be given directly.

[0060] 2) Calculate the average score of each part under each three-dimensional working condition. For each three-dimensional working condition , Corresponding to the flashover time, pressure condition, and injection time, the arithmetic mean of the scores for each body part of all mannequins under this condition is calculated to form an average score vector. , N This indicates the total number of body parts evaluated. Indicates the first N Each body part under test conditions The average score is calculated based on the following. For example, in the overall thermal exposure ergonomics evaluation test, if the total number of body parts is 10 (including head and neck, chest, back, shoulders, elbows, waist, hips, knees, thighs, and calves), then... This indicates that the lower leg area is under working conditions. c l The average score is calculated based on the average score vectors of all corresponding body parts. This matrix is ​​composed of the average score vectors for each corresponding body part. S (216 rows × 10 columns).

[0061] Table 10. Examples of average scores for each part under various three-dimensional working conditions.

[0062] 3) Objective weighting of body parts based on the CRITIC method For the average rating matrix S The CRITIC method was used for analysis to calculate the objective weights of each body part, resulting in the body part weight vector: w = [0.0656, 0.0522, 0.1098, 0.2693, 0.0558, 0.1163, 0.0213, 0.1248, 0.0889, 0.0961], as shown in Table 11.

[0063] Table 11: Example of objective weighting calculation results

[0064] 4) Calculate the single comprehensive score under each three-dimensional working condition. For each three-dimensional working condition The average score vector of this three-dimensional working condition is used. The weighted vector w of body parts is used to calculate a single comprehensive score representing the overall work efficiency at that point in time, through weighted fusion. Calculation method and The results are the same, only the data used in the calculation are different. The calculation results are shown in Table 12.

[0065] Table 12 Examples of Single-Instance Comprehensive Scores for Each 3D Working Condition

[0066] 5) Define the three-dimensional load case weight function and calculate the normalized weights. The liquid threat weighting function is ,in, Indicates test conditions The corresponding original weights for liquid threats; Indicates pressure conditions and injection time The corresponding liquid splash weight and , , , To normalize the liquid pressure, and These are the preset minimum and maximum liquid pressures, respectively. To normalize the splash time, and These are the preset minimum and maximum splash times, respectively. This is a stress index that characterizes the degree to which stress contributes to a threat. The time index characterizes the contribution of splash time to the threat level; parameters... and Determined by regression analysis of experimental data , ,Right now Examples of the calculation results are shown in Table 13.

[0067] Table 13 Examples of Original Weights for Each 3D Working Condition

[0068] 6) Calculate the thermal coupling resistance index (HCRI). The original weights of the liquid threat are normalized to obtain the normalized weights of the liquid threat. The normalization results are shown in Table 14. The calculation method is the same as that used in the calculation of the normalized weight of heat exposure time in the Heat Exposure Efficiency Index (HTEI), and will not be elaborated here.

[0069] Table 14 Examples of Normalized Weights

[0070] The single comprehensive score of each three-dimensional working condition Its corresponding normalized weights for liquid threats The weighted summation is shown in Table 15, yielding a final HCRI value of 5.74 (out of 10). Table 15: Example of HCRI calculation process

[0071] 7) Source analysis of contribution According to the contribution tracing method proposed in this invention, HCRI is analyzed in depth to locate the most severe failure condition and the weakest body part. Under each test condition, the difference between the base score and the single comprehensive score at each flashover time is calculated. The product of this difference and the normalized weight is taken as the failure value. The maximum failure value across all flashover times is taken as the failure condition. Under the test condition corresponding to the failure condition, the body part score is calculated by multiplying the arithmetic mean of the scores for each body part by the weight of each body part. The body part with the lowest score is taken as the weakest body part. The base score is the single comprehensive score corresponding to a flashover time of 0. Specifically, in this embodiment, the most severe failure condition under the thermal coupling liquid resistance test condition is located: ,in, This is the basic score corresponding to the thermal coupling liquid resistance test condition. Analysis revealed that during the flashover time... Liquid pressure Liquid injection time Under certain operating conditions, the negative effect on overall HCRI is most significant.

[0072] Identifying the weakest body parts: during the flashover time Liquid pressure Liquid injection time Under the working conditions, the scores of each part are combined. With weight Calculate the weighted score for each part. The calculation results are shown in Table 16.

[0073] Table 16 Flashover Time Liquid pressure Liquid injection time Analysis of the contribution of each part under working conditions

[0074] As shown in Table 16, the weighted score for the buttocks was the lowest (0.092). Therefore, the buttocks were identified as the weakest body part.

[0075] 3. A certain model of fire-fighting suit was used for thermal coupling particulate repellency performance evaluation. The calculation and traceability process of the Thermal Coupling Particulate Repellency Index (HPRI) is as follows: 1) Construction of the original rating matrix The experiment was designed as a three-dimensional coupled test: eight key flashover pretreatment times (1s, 2s, 3s, 4s, 5s, 6s, 7s, 8s), three action conditions (standing still for 10 minutes, walking for 10 minutes, running for 10 minutes), and four environmental conditions (25°C, 55%RH, 30°C, 60%RH, 35°C, 65%RH, 40°C, 70%RH) were selected, forming a total of 96 test conditions. Each condition was tested three times.

[0076] Define the set of 4 key body parts to be evaluated: P = {chest, waist, thigh, calf}.

[0077] After the experiment, the baseline data of the inward leakage rate of each body part were compared and converted into a ten-point score.

[0078] Construction of the original scoring matrix: All experimental data were organized into an original scoring matrix X. The rows of this matrix correspond to each experimental sample (a total of 8 flashover times × 3 action conditions × 4 environmental condition conditions × 3 replicates = 288 rows), and the columns correspond to 4 body parts. Matrix elements x ij Record number i The sample at the th j The scores for each part are shown in Table 17, which is an example of comparing the baseline data of the leakage rate from the fire suit worn by the volunteer with the original scores.

[0079] Table 17 Example of Inward Leakage Rate Benchmark Data Comparison with Original Scores

[0080] The original rating matrix mentioned above is part of the rating database. The experimental acquisition process of the original rating matrix is ​​not an improvement of this invention, but rather the basic data of this invention. In practical applications, the original rating matrix can be given directly.

[0081] 2) Calculate the average score of each part under each three-dimensional working condition. For each three-dimensional working condition , Let these represent the flashover time, action conditions, temperature, and humidity, respectively. Calculate the arithmetic mean of the scores for each body part of all volunteers under these three-dimensional conditions to form an average score vector. .all The average score matrix S (96 rows × 4 columns) is composed of the average score vectors of each body part. As shown in Table 18.

[0082] Table 18 Examples of average scores for each part under various three-dimensional working conditions

[0083] 3) Based on the objective weighting of body parts using the CRITIC method, the average score matrix S is analyzed using the CRITIC method to calculate the objective weight of each body part, resulting in the body part weight vector: w = [0.3932, 0.1608, 0.2805, 0.1655], as shown in Table 19.

[0084] Table 19: Objective weighted calculation results

[0085] 4) Calculate the single comprehensive score under each three-dimensional working condition. For each three-dimensional working condition The average score vector of this three-dimensional working condition is used. The weighted vector w of the work location is used to calculate a single comprehensive score representing the overall work efficiency at that point in time through weighted fusion. As shown in Table 20.

[0086] Table 20 Examples of Single-Instance Comprehensive Scores for Each 3D Working Condition

[0087] 5) Define the three-dimensional working condition weight function and calculate the normalized weights. Specifically, determine the motion weight coefficients through the range of motion of real people wearing the suits in experiments. The weighting coefficient for standing still for 10 minutes is 1.0, the weighting coefficient for walking for 10 minutes is 1.2, and the weighting coefficient for running for 10 minutes is 1.7.

[0088] Calculate test conditions Corresponding original weights ,right Normalization is performed to obtain normalized weights. As shown in Table 21.

[0089] Table 21 Examples of Original Weights for Each 3D Working Condition

[0090] 6) Calculate the thermal coupling particulate repellency index (HPRI) In one possible embodiment, for each three-dimensional working condition Original weights Normalization is performed to obtain normalized weights. Each three-dimensional working condition Single comprehensive score Its corresponding normalized weights The weighted summation yields the final HPRI value, as shown in Table 22, which is 6.2713 (out of 10 points).

[0091] Table 22: Example of HPRI Calculation Process

[0092] 7) Source analysis of contribution According to the contribution tracing method proposed in this invention, HPRI is analyzed in depth to locate the most severe failure condition and the weakest body part. Under each test condition, the difference between the base score and the single comprehensive score at each flashover time is calculated. The product of this difference and the normalized weight is taken as the failure value. The maximum failure value across all flashover times is taken as the failure condition. Under the test condition corresponding to the failure condition, the body part score is calculated by multiplying the arithmetic mean of the scores for each body part by the weight of each body part. The body part with the lowest score is taken as the weakest body part. The base score is the single comprehensive score corresponding to a flashover time of 0. Specifically, in this embodiment, the most severe failure condition under the thermal coupling particulate repellency test condition is located: Analysis revealed that the detonation time The negative effects on overall HPRI are most significant when the conditions are equal to 8 seconds, 10 minutes of running, 40°C, and 70%RH. This represents the basic score for the thermal coupling particulate matter protection performance test conditions.

[0093] Identifying the weakest body parts: during the flashover time Running for 10 minutes in an environment with 40°C and 70% RH, combined with scores for each body part. With weight Calculate the weighted score for each part. As shown in Table 23.

[0094] Table 23 Flashover Time Contribution analysis of each part under the following conditions: running for 10 minutes, environmental conditions of 40°C and 70%RH.

[0095] As shown in Table 23, the lower leg had the lowest weighted score (0.662). Therefore, the lower leg was identified as the weakest part of the body.

[0096] To provide a clearer understanding of the specific construction process of the comprehensive performance prediction model of this invention, the following detailed explanation is provided with specific examples: 1. Dataset Construction The measured data obtained from the three experiments mentioned above (HTEI, HCRI, HPRI) are integrated. It should be noted that the data obtained is not limited to the data of different working conditions in the above embodiments, and a dataset for model training and validation is constructed.

[0097] A total of 550 valid samples were collected. Each sample corresponds to a defined combination of "fire suit - pretreatment conditions - test conditions" and includes the actual values ​​of HTEI, HCRI, and HPRI calculated by the method described above.

[0098] 2. Input Feature Set and Output Labels Eighteen features closely related to the overall performance of firefighting suits were selected as model inputs, as detailed in Table 24 below: Table 24 Input Feature Information

[0099] Output labels (3D): The true values ​​of HTEI, HCRI, and HPRI for each sample (out of 10).

[0100] 3. Model Architecture and Training Based on the scoring database and the calculated multidimensional performance output set (HTEI, HCRI, HPRI), a comprehensive performance prediction model for fire suits is constructed and trained, specifically including the following steps: The input feature set (18-dimensional) and output label set (3-dimensional) are preprocessed and encoded. After standardizing the feature variables, they are randomly divided into training set (385 samples), validation set (83 samples) and test set (83 samples) in a ratio of 70%:15%:15%.

[0101] A multilayer perceptron architecture incorporating a deep neural network is defined. This architecture consists of an input layer (18 neurons, consistent with the input feature dimension), three hidden layers, and an output layer with three output nodes. The hidden layers employ a progressively decreasing topology (32, 16, and 8 neurons respectively), with each layer using a modified linear unit as the non-linear activation function. The three neurons in the output layer correspond to the predicted values ​​of three performance metrics: thermal exposure ergonomics (HTEI), thermal coupling liquid resistance (HCRI), and thermal coupling particulate resistance (HPRI), respectively, and employ linear activation functions. To improve the model's generalization ability, a dropout layer (dropout rate set to 0.1) is introduced into the network, combined with L2 weight decay (decay coefficient λ = 0.001). Mean squared error is used as the loss function during model training. The adaptive moment estimation algorithm is selected for optimization, with an initial learning rate set to 0.001, and parameters are updated through mini-batch gradient descent. Periodic validation is implemented during the training phase. If the validation loss does not decrease within 20 consecutive iterations, an early stopping mechanism is triggered to prevent overfitting. The maximum number of training iterations is set to 300. The model converges after approximately 120 training iterations.

[0102] Finally, the trained model was evaluated on the test set, and its overall performance is shown in Table 25 below: Table 25 Performance of the trained model

[0103] The constructed comprehensive performance prediction model, using 18 key features as input, can accurately predict the multi-dimensional performance indices (HTEI, HCRI, HPRI) of fire suits under complex coupled conditions. The model has a simplified structure and high prediction accuracy (average R² = 0.90), enabling rapid and reliable assessment of the comprehensive performance of fire suits under untested conditions. The assessment architecture is as follows: Figure 2 As shown.

[0104] Example 2 Embodiment 2 of this invention mainly provides a specific application of the method proposed in Embodiment 1 in the selection and diagnosis of fire-fighting equipment. The specific process is as follows: Figure 3 As shown, a fire brigade in a certain city received an alarm about a leak and fire in a chemical plant's storage tank area. The command center needs to select the best protective equipment for the task force that is about to enter the core area to perform the valve shut-off operation.

[0105] Step 1: Task Parameter Input. The commander inputs the task type "Chemical Plant Tank Area Leakage Fire", the estimated exposure time of 30 minutes, environmental parameters (temperature 70°C, humidity 80%), threat assessment (liquid threat level: high; particulate matter threat level: medium-high) and personnel body size information.

[0106] The system automatically matches preset scenario weighting rules based on task type. For the "hazardous chemical spill response" scenario, HCRI is the core indicator to ensure firefighters are protected from chemical penetration injury, therefore it is given the highest weight; HPRI is second; HTEI has the lowest weight under the premise of ensuring safety. The specific weight allocation is: HCRI 50%, HPRI 30%, HTEI 20%. See Table 26 for details.

[0107] Table 26 Weighting of Core Metrics in Different Task Scenarios

[0108] Step 2: Automatic system matching and status mapping. The system initially screens three candidate garments (Type A, Type B, and Type C) from the existing active-duty firefighting suits. Based on the fire temperature, exposure time, and garment material parameters, the system estimates the comprehensive equivalent flashover time for each garment (e.g., Type A: 5.8 seconds, Type B: 6.1 seconds, Type C: 4.5 seconds).

[0109] Step 3: Feature Vector Construction and Model Prediction. For each garment, an 18-dimensional feature vector as proposed in Example 1 is constructed and input into the pre-trained model in Example 1. The predicted values ​​of HTEI, HCRI, and HPRI are instantly obtained. The predicted values ​​of HTEI, HCRI, and HPRI for type A garment are 7.6, 8.2, and 6.8, respectively; for type B garment, they are 7.1, 8.8, and 7.2, respectively; and for type C garment, they are 8.0, 5.5, and 7.4, respectively.

[0110] Step 4: Weighted overall score and recommendation ranking.

[0111] Based on the scene weights set in the first step, calculate the weighted overall score for each candidate outfit: Type A: 7.6 × 0.2 + 8.2 × 0.5 + 6.8 × 0.3 = 1.52 + 4.10 + 2.04 = 7.66 Type B: 7.1 × 0.2 + 8.8 × 0.5 + 7.2 × 0.3 = 1.42 + 4.40 + 2.16 = 7.98 Type C: 8.0 × 0.2 + 5.5 × 0.5 + 7.4 × 0.3 = 1.60 + 2.75 + 2.22 = 6.57 Based on the weighted score, type B scored the highest (7.98), followed by type A (7.66), and type C scored the lowest (6.57). The system recommends the following ranking: type B first, type A second, and type C third.

[0112] Step 5: Rule Base Diagnosis and Report Generation. The system combines predicted values ​​and performance evaluation rule bases to generate a selection report that includes quantitative indices and qualitative diagnoses, as shown in Table 27 below.

[0113] Table 27 Selection Report

[0114] The method in Example 1 is not limited to the application in the scenario of Example 2, but can also be extended to scenarios such as equipment procurement demonstration, in-service equipment health management, training and simulation.

[0115] Example 3 Embodiment 3 of the present invention also provides a comprehensive performance evaluation and prediction system for fire suits based on multi-source data fusion, used to execute the method described in Embodiment 1, including: The scoring database construction module is used to obtain the performance scores of fire suits under multiple body parts and multiple test conditions through the whole-pack thermal exposure efficiency evaluation test, the whole-pack thermal coupling anti-permeability evaluation test, and the whole-pack thermal coupling anti-particle characteristic evaluation test, and map them to standard levels to build a scoring database. The multidimensional performance index calculation module is used to determine the weight of each body part based on the scoring database and calculate the single comprehensive score under each test condition; based on the single comprehensive score, by introducing the heat exposure time weight function, the liquid threat weight function and the action weight coefficient, the heat exposure efficiency index, the heat coupling liquid resistance index and the heat coupling particulate resistance index are calculated respectively. The performance prediction model building module is used to build a comprehensive performance prediction model for fire suits using a multi-output regression multilayer perceptron model. It takes several pre-selected features related to the comprehensive performance of fire suits as model inputs, and the real values ​​of thermal exposure efficiency index, thermal coupling liquid resistance index and thermal coupling particulate protection index as model output labels to train the model and obtain a trained comprehensive performance prediction model for fire suits. The intelligent prediction and diagnosis output module is used to output predicted thermal exposure efficiency index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index using the comprehensive performance prediction model of the fire suit. Different weight ratios are set according to different application scenario types to fuse the three performance indices of each fire suit to obtain a weighted comprehensive score. The fire suits are sorted in descending order according to the weighted comprehensive score to obtain the system recommended ranking.

[0116] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion, characterized in that, include: S1. The performance scores of fire suits under multiple body parts and multiple test conditions are obtained through the whole-body thermal exposure efficiency evaluation test, the whole-body thermal coupling anti-penetration characteristic evaluation test, and the whole-body thermal coupling anti-particle characteristic evaluation test, and are mapped to standard levels to build a scoring database. S2. Based on the scoring database, determine the weight of each body part and calculate the single comprehensive score under each test condition; based on the single comprehensive score, by introducing the heat exposure time weight function, the liquid threat weight function and the action weight coefficient, calculate the heat exposure efficiency index, the heat coupling liquid resistance index and the heat coupling particulate resistance index respectively. S3. Construct a comprehensive performance prediction model for fire suits using a multi-output regression multilayer perceptron model. Use several pre-selected features related to the comprehensive performance of fire suits as model inputs, and use the true values ​​of thermal exposure efficiency index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index as model output labels to train the model and obtain a trained comprehensive performance prediction model for fire suits. S4. Using the comprehensive performance prediction model of the fire suit, the predicted thermal exposure efficiency index, thermal coupling liquid resistance index, and thermal coupling particle resistance index are output. Different weight ratios are set according to different application scenario types to fuse the three performance indices of each fire suit to obtain a weighted comprehensive score. The fire suits are sorted in descending order according to the weighted comprehensive score to obtain the system recommended ranking.

2. The method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion according to claim 1, characterized in that, The overall thermal exposure efficiency evaluation test includes: Fire suits that have undergone flashover pretreatment for different durations were selected as test samples. Subjects of a predetermined body type wore the test samples and performed predetermined activities in a predetermined test environment. The overall physical restraint of key parts of the human body was scored.

3. The method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion according to claim 1, characterized in that, The evaluation test of the overall thermal coupling anti-permeability characteristics includes: Firefighting suits pretreated with flashover were worn on mannequins, and liquid splash tests were conducted at different pressures and times. The staining rate of different parts of the colored underwear was analyzed and converted into standardized scores.

4. The method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion according to claim 1, characterized in that, The evaluation test of the overall thermal coupling anti-particle characteristics includes: The fire suits pretreated with flashover were worn by the subjects, who performed prescribed actions of varying intensities inside the aerosol testing chamber. The leakage rate of each part of the suit was monitored in real time and compared with the benchmark data of the leakage rate of an untreated parallel sample. The comparison results were then converted into standardized scores.

5. The method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion according to claim 1, characterized in that, S2 includes: Construct the original rating matrix, which represents the behavioral test samples. The columns of the original rating matrix represent various body parts, and the elements of the original rating matrix correspond to the rating of each test sample at the corresponding body part. For each test condition, the arithmetic mean of the scores of all subjects or dummies at each body part under that test condition is calculated to form an average score vector. The average score vectors corresponding to all test conditions form an average score matrix. The CRITIC method was used to analyze the average score matrix and calculate the weight of each body part to obtain the body part weight vector. For each test condition, the average score vector and body part weight vector of that test condition are used to calculate a single comprehensive score representing the overall efficiency of that test condition through weighted fusion. A weighting function is introduced to calculate the original weights corresponding to each test condition, and then normalization is performed to obtain the normalized weights of each test condition. The final performance index is obtained by weighting and summing the single comprehensive scores of each test condition with their corresponding normalized weights. When the test condition is the flashover time and the weighting function is the heat exposure time weighting function, the calculated index is the heat exposure effectiveness index. When the test condition is a three-dimensional condition consisting of flashover time, liquid pressure condition, and injection time, and the weighting function is the liquid threat weighting function, the calculated index is the thermal coupling liquid resistance index. When the test condition is a three-dimensional condition consisting of flashover time, action condition, and environmental condition condition, and the weighting function is the product of the heat exposure time weighting function and the action weighting coefficient, the calculated index is the thermal coupling particulate resistance index.

6. The method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion according to claim 5, characterized in that, The heat exposure time weighting function is: ,in, Indicates the time of ignition The original weight of the corresponding heat exposure time, This is the preset critical time. This is the control factor.

7. The method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion according to claim 6, characterized in that, The liquid threat weighting function is: ,in, Indicates test conditions The corresponding original weights of the liquid threat. Indicates pressure condition and injection time The corresponding liquid splash weight and , , , To normalize the liquid pressure, and These are the preset minimum and maximum liquid pressures, respectively. To normalize the splash time, and These are the preset minimum and maximum splash times, respectively. This is a stress index that characterizes the degree to which stress contributes to a threat. This is a time index, representing the degree to which splash time contributes to the threat.

8. The method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion according to claim 5, characterized in that, S3 includes: Multiple features related to the comprehensive performance of fire suits were selected as inputs to a multi-output regression multilayer perceptron model. The thermal exposure ergonomics index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index were used as labels to construct the input feature set. The multi-output regression multilayer perceptron model consists of a sequentially connected input layer, three intermediate hidden layers, and an output layer with three output nodes. The hidden layers employ a topology design with progressively decreasing neuron counts, and each layer uses a modified linear unit as a non-linear activation function. The three neurons in the output layer use linear activation functions, corresponding to the predicted values ​​of the thermal exposure ergonomics index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index, respectively. The model was trained using the input feature set, with mean squared error used as the loss function during training. The model parameters were adjusted, and training stopped when a preset number of iterations or the minimum loss function was reached, resulting in a trained model. Periodic validation was implemented during the training phase; if the validation loss did not decrease within consecutive preset iterations, an early stopping mechanism was triggered.

9. The method for comprehensive performance evaluation and prediction of fire suits based on multi-source data fusion according to claim 5, characterized in that, The S2 also includes a contribution tracing analysis based on the contribution of each test condition to the final performance, locating the failure conditions and weakest points of each protection dimension. The specific process is as follows: Under each test condition, the difference between the base score and the single comprehensive score under each flashover time is calculated. The product of the difference and the normalized weight is taken as the failure value. The maximum failure value under all flashover times is taken as the failure condition. Under the test condition corresponding to the failure condition, the body part score is calculated by multiplying the arithmetic mean of the scores of each body part by the weight of each body part. The body part with the lowest score is taken as the weakest body part. The base score is the single comprehensive score corresponding to the flashover time of 0.

10. A comprehensive performance evaluation and prediction system for fire suits based on multi-source data fusion, comprising the method described in any one of claims 1-9, characterized in that, include: The scoring database construction module is used to obtain the performance scores of fire suits under multiple body parts and multiple test conditions through the whole-pack thermal exposure efficiency evaluation test, the whole-pack thermal coupling anti-permeability evaluation test, and the whole-pack thermal coupling anti-particle characteristic evaluation test, and map them to standard levels to build a scoring database. The multidimensional performance index calculation module is used to determine the weight of each body part based on the scoring database and calculate the single comprehensive score under each test condition; based on the single comprehensive score, by introducing the heat exposure time weight function, the liquid threat weight function and the action weight coefficient, the heat exposure efficiency index, the heat coupling liquid resistance index and the heat coupling particulate resistance index are calculated respectively. The performance prediction model building module is used to build a comprehensive performance prediction model for fire suits using a multi-output regression multilayer perceptron model. It takes several pre-selected features related to the comprehensive performance of fire suits as model inputs, and the real values ​​of thermal exposure efficiency index, thermal coupling liquid resistance index and thermal coupling particulate protection index as model output labels to train the model and obtain a trained comprehensive performance prediction model for fire suits. The intelligent prediction and diagnosis output module is used to output predicted thermal exposure efficiency index, thermal coupling liquid resistance index, and thermal coupling particulate resistance index using the comprehensive performance prediction model of the fire suit. Different weight ratios are set according to different application scenario types to fuse the three performance indices of each fire suit to obtain a weighted comprehensive score. The fire suits are sorted in descending order according to the weighted comprehensive score to obtain the system recommended ranking.

Images