A method and system for acquiring and analyzing data from a glow wire test

Abstract

The application discloses a glowing filament test data acquisition and analysis method and system, and relates to the technical field of data analysis, and comprises the following steps: actively verifying a glowing filament testing machine, collecting a plurality of rounds of target test data of each sensor to perform drift trend analysis, and evaluating the drift state of each sensor; correcting the temperature field distribution of the plurality of rounds of target test data according to the drift state of each sensor, extracting a target combustion characteristic parameter, training a combustion state classification model, and generating a target combustion timing grade evaluation; and judging whether the target combustion timing grade evaluation meets the application field direction, and outputting verification qualified if yes, or outputting verification unqualified if no. The application has the beneficial effect of improving the glowing filament test precision and combustion determination accuracy.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to a method and system for acquiring and analyzing data from a glow wire test. Background Technology

[0002] The glow wire test is a standard test method for evaluating the flame retardant properties of materials. It involves heating a nickel-chromium alloy wire of a specific specification to a preset temperature and then contacting it with a sample to observe whether the material ignites, the duration of combustion, and the behavior of dripping substances, thereby determining its fire resistance rating.

[0003] Existing glow wire testing machines suffer from signal attenuation or reference shift of temperature measuring elements such as thermocouples due to long-term high-temperature operation, resulting in distorted temperature field data. Secondly, traditional manual interpretation relies on operators' visual observation of flame height and duration, which is highly subjective and lacks consistency in judging transient combustion and dripping behavior, leading to poor accuracy in assessing the flame retardancy rating of the verification target. Summary of the Invention

[0004] To address the aforementioned technical problems, a method and system for acquiring and analyzing glow wire test data are provided. This technical solution solves the problems of existing glow wire testing machines, where temperature sensing elements such as thermocouples are prone to signal attenuation or reference shift due to long-term high-temperature operation, resulting in distorted temperature field data. Secondly, traditional manual interpretation relies on operators' visual observation of flame height and duration, which is subject to strong human subjectivity and lacks consistency in judging transient combustion and dripping behavior, leading to poor accuracy in assessing the flame retardancy rating of the verification target.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for acquiring and analyzing data from a glow wire test, comprising:

[0007] Actively verify the glow wire test machine, collect several rounds of target test data from each sensor to analyze drift trends, and evaluate the drift status of each sensor;

[0008] Based on the drift state of each sensor, the temperature field distribution of several rounds of target test data is corrected, target combustion characteristic parameters are extracted to train a combustion state classification model, and a target combustion time series rating is generated.

[0009] The target combustion timing level is assessed to determine whether it meets the requirements of the application area. If yes, the output verification is qualified; otherwise, the output verification is unqualified.

[0010] Furthermore, based on the preset temperature of the glow wire tester, the theoretical data acquisition range of the preset temperature of each sensor is determined, and a standardized reference data array of the preset temperature of each sensor is established.

[0011] Acquire target test data from several rounds of each sensor, divide the target test data from several rounds of each round ...

[0012] Using the timestamp of the data sampling rate as the observation window, the standardized control data array of the preset temperature of each sensor and the target test data of the preset temperature of each sensor are used as observation variables to extract the standardized control time series data of the preset temperature of each sensor and the target test time series data of the preset temperature of each sensor.

[0013] Using cosine similarity, the deviation angle between the standardized control time series data of the preset temperature of each sensor and the target test time series data of the preset temperature of each sensor is calculated, and the target test deviation time series data of the preset temperature of each sensor is selected.

[0014] Furthermore, a sensor status assessment model for the glow wire testing machine is established based on ARMA autoregressive moving average.

[0015] The target test deviation time series data of the preset temperature of each sensor is substituted into the sensor state evaluation model of the glow wire test machine. The target test deviation parameter value of the preset temperature of the sensor per unit time is used as the input, and the target test deviation drift value of the preset temperature of each sensor per unit time is used as the output.

[0016] Based on the entropy weight method, the ratio of the target test deviation drift value of the preset temperature of each sensor per unit time to the standardized control data array of the preset temperature of each sensor is calculated, and the target test deviation drift value of the preset temperature of each sensor per unit time is assigned a weight.

[0017] The drift state index of each sensor is calculated by normalizing the target test deviation drift value of the preset temperature of each sensor per unit time and weighting the target test deviation drift value of the preset temperature of each sensor per unit time.

[0018] Furthermore, based on the thermocouple target test data from several rounds of target test data from various sensors, the contact measurement local temperature of the thermocouple target test data is determined;

[0019] Based on the drift state indices of each sensor, the contact measurement local temperature of the thermocouple target test data is corrected to obtain the contact local corrected temperature per unit time of the thermocouple target test data.

[0020] Based on infrared thermal imaging target test data from several rounds of target test data from various sensors, the non-contact measurement two-dimensional temperature field per unit time of the target test data is determined.

[0021] The infrared spatial gradient of the target test data in a two-dimensional temperature field is extracted using the Laplace algorithm for non-contact measurement of the target test data.

[0022] By utilizing the infrared spatial gradient of the target test data, the contact local correction temperature of the thermocouple target test data per unit time is compensated, and the comprehensive compensated temperature field distribution of the target test data per unit time is obtained.

[0023] Furthermore, based on the combustion videos in several rounds of target test data from various sensors, the data is divided according to the video frame rate to obtain combustion frame image samples of the target test data;

[0024] A combustion flame segmentation model was established based on the U-Net convolutional neural network.

[0025] By substituting combustion frame image samples from the target experimental data into the combustion flame segmentation model, a combustion flame probability map per unit time is generated for the target experimental data. ;in, The probability of flame at the p-th pixel position in the combustion frame image at the t-th unit time of the target experimental data;

[0026] Based on the combustion flame probability diagram per unit time of the target test data, calculate the combustion area per unit time of the target test data;

[0027] Based on Savitzky-Golay filtering differentiation, the derivative of the combustion area per unit time of the target test data is calculated using the combustion area per unit time of the target test data;

[0028] Based on the combustion area per unit time and the derivative of the combustion area per unit time of the target test data, the combustion flame characteristic data of the target test data are constructed.

[0029] Furthermore, a combustion state classification model for the target test data is established based on the LSTM long short-term memory network;

[0030] Based on the combustion flame characteristic data of the target test data, establish the original feature vector of the combustion flame of the target test data;

[0031] The combustion state classification model based on target test data takes the original feature vector of the combustion flame of the target test data as input and the combustion level of the target test data per unit time as output.

[0032] Furthermore, based on the application area of ​​the verification target, the application requirements at different combustion temperatures in the application area are determined, and a combustion requirement vector per unit time is established for the application area.

[0033] The combustion rating per unit time of the target test data is standardized, and a vector transformation is performed using linear mapping to obtain the combustion rating capability vector per unit time of the target test data.

[0034] Calculate the spatial distance between the combustion demand vector per unit time in the application area and the combustion rating capability vector per unit time in the target test data;

[0035] Determine whether the spatial distance between the combustion demand vector per unit time in each application area and the combustion rating capability vector per unit time in the target test data meets the application area threshold. If yes, output "verification qualified"; otherwise, output "verification unqualified".

[0036] Furthermore, a glow wire test data acquisition and analysis system includes:

[0037] Drift assessment module, combustion determination module, and pass / fail judgment module;

[0038] The drift assessment module is used to actively verify the glow wire testing machine, collect several rounds of target test data from each sensor to analyze drift trends, and assess the drift status of each sensor.

[0039] The combustion adjudication module is electrically connected to the drift evaluation module. The combustion adjudication module is used to correct the temperature field distribution of several rounds of target test data according to the drift state of each sensor, extract target combustion feature parameters to train the combustion state classification model, and generate a target combustion time series rating.

[0040] The qualification judgment module is electrically connected to the combustion determination module. The qualification judgment module is used to determine whether the target combustion time sequence level meets the application field direction. If yes, it outputs verification qualified; otherwise, it outputs verification unqualified.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0042] This invention proposes a data acquisition and analysis scheme for glow wire tests. By constructing an intelligent glow wire test evaluation system, it employs multi-source data fusion and deep learning technology to accurately determine the fire resistance performance of materials. The system first uses an ARMA model and entropy weighting method to dynamically monitor and compensate for sensor drift, then combines infrared gradient correction and U-Net flame segmentation technology to reconstruct a high-precision temperature field. Next, it utilizes an LSTM network to analyze combustion time-series characteristics and establish a combustion level classification model. Finally, it achieves quantitative comparison of requirements and capabilities through vector space matching. This scheme significantly improves the testing accuracy of glow wire tests and the accuracy of combustion determination. Attached Figure Description

[0043] Figure 1This is a flowchart of a method for acquiring and analyzing data from a glowing wire test.

[0044] Figure 2 This is a framework diagram of a hot wire test data acquisition and analysis system; Detailed Implementation

[0045] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0046] Reference Figure 1 As shown, a method for acquiring and analyzing data from a glow wire test includes:

[0047] Step 1: Actively verify the glow wire test machine, collect several rounds of target test data from each sensor, perform drift trend analysis, and evaluate the drift status of each sensor;

[0048] Step one includes the following:

[0049] Step 101: Based on the preset temperature of the glow wire tester, determine the theoretical data acquisition range of the preset temperature of each sensor, and establish a standardized reference data array of the preset temperature of each sensor.

[0050] Acquire target test data from several rounds of each sensor, divide the target test data from several rounds of each round ...

[0051] Using the timestamp of the data sampling rate as the observation window, the standardized control data array of the preset temperature of each sensor and the target test data of the preset temperature of each sensor are used as observation variables to extract the standardized control time series data of the preset temperature of each sensor and the target test time series data of the preset temperature of each sensor.

[0052] Using cosine similarity, the deviation angle between the standardized control time-series data of the preset temperature of each sensor and the target experimental time-series data of the preset temperature of each sensor is calculated. The target experimental deviation time-series data of the preset temperature of each sensor are then selected, as follows: ,in, The time series data of the target test deviation of the k-th preset temperature of the i-th sensor at the t-th unit time. This is the standardized comparison time-series data for the k-th preset temperature of the i-th sensor at the t-th unit time. For the target test time series data of the k-th preset temperature of the i-th sensor at the t-th unit time, This represents the total number of preset temperatures;

[0053] Step 102: Based on ARMA autoregressive moving average, establish a sensor status evaluation model for the glow wire testing machine;

[0054] The target test deviation time series data of the preset temperature of each sensor is substituted into the sensor state evaluation model of the glow wire testing machine. The target test deviation parameter value of the preset temperature of the sensor per unit time is used as input, and the target test deviation drift value of the preset temperature of each sensor per unit time is used as output, as follows: ,

[0055] in, Let $t$ be the target test deviation drift value of the $k$-th preset temperature of the $i$-th sensor at the $t$-th unit time. For constant terms, Let j be the autoregressive coefficient. Let be the target test deviation drift value of the i-th sensor at the k-th preset temperature in the ti-th unit time. Let j be the coefficient of the moving average order. Let be the predicted residual for the ti-th unit time. Let be the random error term at time t. The total number of autoregressive orders. This represents the total number of moving average orders.

[0056] Based on the entropy weight method, the ratio of the target test deviation drift value of the preset temperature of each sensor per unit time to the standardized control data array of the preset temperature of each sensor is calculated, and the target test deviation drift value of the preset temperature of each sensor per unit time is assigned a weight, as follows: ,

[0057] in, The weight of the target test deviation drift value of the i-th sensor at the k-th preset temperature in the t-th unit time is given. Let be the entropy of the target test deviation drift value of the i-th sensor at the k-th preset temperature in the t-th unit time. The ratio of the target test deviation drift value of the i-th sensor at the k-th preset temperature in the t-th unit time to the standardized control data array of the preset temperatures of each sensor;

[0058] The drift state index of each sensor is calculated by normalizing the target test deviation drift value of the preset temperature of each sensor per unit time and weighting the target test deviation drift value of the preset temperature of each sensor per unit time, as follows: ,

[0059] in, Let be the drift state index of the i-th sensor at the t-th unit time. Let $t$ be the absolute value of the target test deviation drift at the $k$-th preset temperature of the $i$-th sensor in the $t$-th unit time. This represents the total number of sensors inside the glow wire testing machine;

[0060] When using it, refer to steps 101 to 102:

[0061] As a further step, by establishing a standardized control data array for the sensor and collecting time-series data from the target test, and using cosine similarity to filter out deviation time-series data, and combining the ARMA model and entropy weight method to calculate drift state indicators, accurate drift trend analysis of the glow wire testing machine sensor is achieved. The beneficial effects are: 1. Improved drift detection sensitivity based on the standardized control array and dynamic time-series analysis; 2. Objective allocation of weights using the entropy weight method, avoiding existing subjective judgment biases; 3. Quantitative assessment of the degree of sensor performance degradation, effectively ensuring the stable operation of the glow wire testing machine.

[0062] Step 2: Correct the temperature field distribution of several rounds of target test data based on the drift state of each sensor, extract target combustion characteristic parameters to train the combustion state classification model, and generate target combustion time series level assessment.

[0063] Step two includes the following:

[0064] Step 201: Based on the thermocouple target test data from several rounds of target test data from each sensor, determine the contact measurement local temperature of the thermocouple target test data;

[0065] Based on the drift state indices of each sensor, the contact measurement local temperature of the thermocouple target test data is corrected to obtain the contact local corrected temperature per unit time of the thermocouple target test data.

[0066] Based on infrared thermal imaging target test data from several rounds of target test data from various sensors, the non-contact measurement two-dimensional temperature field per unit time of the target test data is determined.

[0067] The infrared spatial gradient of the target test data in a two-dimensional temperature field is extracted using the Laplace algorithm for non-contact measurement of the target test data.

[0068] The infrared spatial gradient of the target test data is used to compensate for the local contact correction temperature per unit time of the thermocouple target test data, resulting in the comprehensive compensated temperature field distribution per unit time of the target test data, as follows: ,

[0069] in, The comprehensive compensated temperature field distribution of the target experimental data at the t-th unit time. For the first Thermocouple sensor weights, For the thermocouple target test data, the data at the t-th unit time is... A thermocouple sensor contacts a localized temperature correction point. For the infrared thermal imaging sensor, the reliable adaptive coefficient is... The infrared spatial gradient of the target test data;

[0070] Step 202: Based on the combustion videos in several rounds of target test data from each sensor, divide the data according to the video frame rate to obtain combustion frame image samples of the target test data;

[0071] A combustion flame segmentation model was established based on the U-Net convolutional neural network.

[0072] By substituting combustion frame image samples from the target experimental data into the combustion flame segmentation model, a combustion flame probability map per unit time is generated for the target experimental data. ;in, The probability of flame at the p-th pixel position in the combustion frame image at the t-th unit time of the target experimental data;

[0073] Based on the combustion flame probability diagram per unit time of the target test data, calculate the combustion area per unit time of the target test data;

[0074] Based on Savitzky-Golay filtering differentiation, the derivative of the combustion area per unit time of the target test data is calculated using the combustion area per unit time of the target test data, as follows: ,

[0075] in, The burning area at the t-th unit time in the target test data. The actual area of ​​the burning frame image. The derivative of the instantaneous change in combustion area with respect to unit time in the target test data is given. These are the differential weighting coefficients. For symmetrical time windows, Index for summation of integrals within a symmetric time window The time interval between adjacent data points within a symmetrical time window;

[0076] Combustion flame characteristic data are constructed based on the combustion area per unit time and the derivative of the combustion area per unit time of the target test data.

[0077] Step 203: Based on the LSTM long short-term memory network, establish a combustion state classification model for the target test data;

[0078] Based on the combustion flame characteristic data of the target test data, establish the original feature vector of the combustion flame of the target test data;

[0079] The combustion state classification model based on target test data uses the original feature vector of the combustion flame from the target test data as input and the combustion level per unit time from the target test data as output, as follows: ,

[0080] in, The combustion rating is the value of the target test data at the t-th unit time. The interface for the combustion state classification model of the target test data contains functions. The combustion rating is the value of the target test data at the (t-1)th unit time. Let be the original feature vector of the combustion flame at the t-th unit time of the target experimental data. The target experimental data represents the maximum combustion flame area per unit time. The integral of the combustion flame area. This is a temporary time variable within the integral;

[0081] When using this method, refer to steps 101 to 103:

[0082] As a further development, by combining drift compensation and temperature field gradient correction of thermocouples and infrared thermal imagers, a comprehensive compensated temperature field distribution is generated. U-Net is used to segment the combustion flame and extract the combustion area and its derivative. A dynamic flame feature vector is constructed and substituted into an LSTM model to fuse time-series temperature and flame morphology features, outputting a flammability rating conforming to the UL94 standard. The beneficial effects include: 1. Improved temperature field measurement accuracy; 2. Increased flame segmentation accuracy; 3. Reduced flammability rating classification error.

[0083] Step 3: Determine whether the target combustion timing level meets the requirements of the application area. If yes, output verification qualified; otherwise, output verification unqualified.

[0084] Step 301: Based on the application area direction of the verification target, determine the application requirements of the application area direction at different combustion temperatures, and establish the combustion requirement vector of the application area direction per unit time.

[0085] The combustion rating per unit time of the target test data is standardized, and a vector transformation is performed using linear mapping to obtain the combustion rating capability vector per unit time of the target test data.

[0086] The spatial distance between the combustion demand vector per unit time in the application area and the combustion rating capability vector per unit time in the target test data is calculated as follows: ,

[0087] in, The combustion demand vector per unit time for the application area. Combustion rating capability vector per unit time of target test data Spatial distance between them;

[0088] Determine whether the spatial distance between the combustion demand vector per unit time in each application area and the combustion rating capability vector per unit time in the target test data meets the application area threshold. If yes, output "verification qualified"; otherwise, output "verification unqualified".

[0089] Reference Figure 2 As shown, a glow wire test data acquisition and analysis system includes:

[0090] Drift assessment module, combustion determination module, and pass / fail judgment module;

[0091] The drift assessment module is used to actively verify the glow wire testing machine, collect several rounds of target test data from each sensor to analyze drift trends, and assess the drift status of each sensor.

[0092] The combustion adjudication module is electrically connected to the drift evaluation module. The combustion adjudication module is used to correct the temperature field distribution of several rounds of target test data according to the drift state of each sensor, extract target combustion feature parameters to train the combustion state classification model, and generate a target combustion time series rating.

[0093] The qualification judgment module is electrically connected to the combustion determination module. The qualification judgment module is used to determine whether the target combustion time sequence level meets the application field direction. If yes, it outputs verification qualified; otherwise, it outputs verification unqualified.

[0094] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A glowing filament test data acquisition analysis method, characterized by, The method comprises the following steps: According to the preset temperature of the glow wire tester, the theoretical data acquisition range of the preset temperature of each sensor is determined, and a standardized control data array of the preset temperature of each sensor is established; According to the preset temperature, the target test data of several rounds of each sensor is divided to obtain the preset temperature target test data of each sensor; According to the time stamp of the data sampling rate as the observation window, the standardized control time series data of the preset temperature of each sensor and the target test time series data of the preset temperature of each sensor are extracted as observation variables; Using cosine similarity, the deviation angle between the standardized control time series data of the preset temperature of each sensor and the target test time series data of the preset temperature of each sensor is calculated, and the target test deviation time series data of the preset temperature of each sensor is selected; Based on ARMA autoregressive moving average, a sensor state evaluation model of the glow wire tester is established; According to the target test deviation time series data of the preset temperature of each sensor, the sensor state evaluation model of the glow wire tester is substituted, and the target test deviation parameter value of the preset temperature of the sensor per unit time is taken as the input, and the target test deviation drift value of the preset temperature of each sensor per unit time is taken as the output; Based on the entropy weight method, the ratio of the target test deviation drift value of the preset temperature of each sensor per unit time to the standardized control data array of the preset temperature of each sensor is calculated, and the weight of the target test deviation drift value of the preset temperature of each sensor per unit time is given; The target test deviation drift value of the preset temperature of each sensor per unit time is normalized, and the weight of the target test deviation drift value of the preset temperature of each sensor per unit time is calculated to obtain the drift state index of each sensor; According to the drift state of each sensor, the temperature field distribution of the target test data of several rounds is corrected, the target combustion characteristic parameter is extracted, the combustion state classification model is trained, and the target combustion time sequence level evaluation is generated; According to the target combustion time sequence level evaluation, it is judged whether it meets the application field direction, if yes, the verification is qualified, and if not, the verification is unqualified.

2. The glow wire test data acquisition and analysis method according to claim 1, wherein: According to the thermocouple target test data of each sensor in the target test data of several rounds, the contact measurement local temperature of the thermocouple target test data is determined; According to the drift state index of each sensor, the contact local correction temperature of the thermocouple target test data per unit time is obtained by correcting the contact measurement local temperature of the thermocouple target test data; According to the infrared thermal image target test data of the target test data of several rounds of each sensor, the non-contact measurement two-dimensional temperature field of the target test data per unit time is determined; ​ The infrared data space gradient of the target test data is extracted by using the Laplace algorithm to non-contact measure the two-dimensional temperature field of the target test data; The contact local correction temperature of the target test data per unit time is compensated by using the infrared data space gradient of the target test data, and a comprehensive compensation temperature field distribution of the target test data per unit time is obtained.

3. The hot wire test data acquisition and analysis method according to claim 2, characterized in that: According to the combustion video in the target test data of several rounds of each sensor, the combustion frame image samples of the target test data are obtained by dividing according to the video frame rate; Based on the U-Net convolutional neural network, a combustion flame segmentation model is established; According to the combustion frame image sample in the target test data, the combustion flame segmentation model is substituted to generate the combustion flame probability graph of the target test data in unit time ; wherein is the flame probability of the pth pixel position of the combustion frame image in the tth unit time of the target test data Based on the combustion flame probability map of the target test data per unit time, the combustion area of the target test data per unit time is calculated; Based on the Savitzky-Golay filter differential, the combustion area derivative of the target test data per unit time is calculated by using the combustion area of the target test data per unit time; According to the combustion area of the target test data per unit time and the combustion area derivative of the target test data per unit time, the combustion flame feature data of the target test data is established.

4. The hot wire test data acquisition and analysis method according to claim 3, characterized in that: Based on the LSTM long short-term memory network, a combustion state classification model of the target test data is established; According to the combustion flame feature data of the target test data, a combustion flame original feature vector of the target test data is established; Based on the combustion state classification model of the target test data, the combustion flame original feature vector of the target test data is taken as input, and the combustion grade of the target test data per unit time is taken as output.

5. The hot wire test data acquisition and analysis method according to claim 4, characterized in that: Based on the to-be-applied field direction of the verification target, the application demand of different combustion temperatures in the to-be-applied field direction is determined, and a combustion demand vector per unit time in the to-be-applied field direction is established; The combustion grade per unit time of the target test data is standardized and vector converted by using linear mapping to obtain a combustion grade capability vector per unit time of the target test data; The spatial distance between the combustion demand vector per unit time in the to-be-applied field direction and the combustion grade capability vector per unit time of the target test data is calculated; It is judged whether the spatial distance between the combustion demand vector per unit time in the to-be-applied field direction and the combustion grade capability vector per unit time of the target test data meets the application field threshold value, if yes, output verification qualified, if not, output verification unqualified.

6. A hot wire test data acquisition and analysis system characterized by, For realizing the hot wire test data acquisition and analysis method according to any one of claims 1-5, comprising: Drift evaluation module, combustion arbitration module, qualified judgment module; The drift evaluation module is used to actively verify the hot wire tester, collect several rounds of target test data of each sensor for drift trend analysis, and evaluate the drift state of each sensor; The combustion adjudication module is electrically connected with the drift evaluation module, and is configured to correct a temperature field distribution of target test data of several rounds according to the drift state of each sensor, extract a target combustion characteristic parameter, train a combustion state classification model, and generate a target combustion timing grade evaluation; The qualified judgment module is electrically connected with the combustion adjudication module, and is configured to judge whether the target combustion timing grade evaluation meets the application field direction, and if yes, output verification qualified, and if not, output verification unqualified.

Images