Methods, systems and apparatus for vertical combustion testing of materials

Abstract

This invention proposes a method, system, and apparatus for vertical combustion testing of materials, relating to the technical fields of machine vision and flame retardant testing of materials. For the vertical combustion testing process, it integrates machine vision and utilizes a pre-trained flame region detection model to identify the flame region frame by frame in each combustion image. Based on the identified flame region, it obtains the core reference point and combustion endpoint of the flame region in the horizontal direction for each frame of the combustion image. It can extract the coordinate changes of the core reference point and combustion endpoint in each frame, and use the position data of the core reference point and combustion endpoint to control the movement of the flame application position, improving the dynamic response capability of the applied flame, ensuring that the flame always tracks the combustion endpoint, improving accuracy, avoiding full human intervention, increasing testing efficiency, and ensuring the safety of personnel. The apparatus proposed in this invention automatically controls the movement of the flame application position based on the position data of the core reference point and combustion endpoint, improving the accuracy of the test.

Description

Technical Field

[0001] This invention relates to the technical fields of machine vision and flame retardant testing of materials, and more specifically, to a method, system and apparatus for vertical burning testing of materials. Background Technology

[0002] Polymer materials are widely used in automobiles, home appliances, and many other fields. However, these materials are prone to causing various types of fires. Therefore, accurate and reliable flame retardant testing is required before large-scale production to evaluate their flame retardant performance. Currently, this evaluation is mainly achieved through horizontal and vertical burning tests. Taking the vertical burning test as an example, the standard requires that material samples be tested according to the specified flame application distance and time. Then, the flame retardant performance is evaluated based on the burning time of the material sample after flame application (usually using a Bunsen burner).

[0003] However, the current vertical burning test process has many limitations, such as: the test process requires full manual participation, including adjusting the position of the Bunsen lamp and recording the burning time, which is inefficient; the material sample is prone to complex changes such as shrinkage and bending during the burning process, and it is necessary to maintain the distance between the Bunsen lamp and the midpoint of the bottom surface of the material sample at 10±1mm in real time, which is difficult to control precisely by manual operation; the recording of the burning time of the material sample depends on subjective judgment and is related to the reaction speed of the tester, resulting in a large deviation between the test results and the actual results; the burning process of the material sample produces harmful fumes, which poses a health threat to the testers. Summary of the Invention

[0004] To address the insufficient effectiveness of current vertical combustion testing methods for materials, this invention proposes a method, system, and apparatus for vertical combustion testing of materials, which integrates machine vision to improve testing efficiency and safety, and enhance the reliability of test results.

[0005] To achieve the above-mentioned technical effects, the technical solution of the present invention is as follows: Firstly, this application proposes a method for testing the vertical burning of materials, comprising the following steps: Apply a flame to the material specimen to be tested; Real-time capture of each frame of combustion images of the material spline, and frame-by-frame identification of the flame region in each frame of combustion image using a pre-trained combustion flame region detection model; Based on the frame-by-frame identification of the flame region, the core reference point and the position data of the core reference point in the horizontal direction of the flame region in each frame of the combustion image are obtained. Using the core reference point of the flame region in the horizontal direction of each frame of the combustion image as a reference, a scanning reference line of the flame region in the vertical direction is constructed. Based on the scanning reference line, the combustion endpoint and position data of the flame region in each frame of the combustion image are obtained. The position data of the core reference point and the position data of the combustion endpoint, which are acquired frame by frame, are used as position control data. The movement trajectory of the flame application position is controlled by the position control data until the vertical combustion test ends, and the test results are obtained.

[0006] In this technical solution, machine vision is integrated for the vertical combustion test of materials. A pre-trained combustion flame area detection model is used to identify the flame area frame by frame in each combustion image. Based on the flame area identified frame by frame, the core reference point and combustion endpoint of the flame area in the horizontal direction of each combustion image are obtained. The coordinate changes of the core reference point and combustion endpoint in each frame can be extracted. The position data of the core reference point and combustion endpoint are used to control the movement of the flame application position, which improves the dynamic response capability of the applied flame, ensures that the flame always tracks the combustion endpoint, avoids full human intervention, improves test efficiency and accuracy, and also ensures the safety of the staff.

[0007] Preferably, the process of pre-training the combustion flame area detection model is as follows: Multiple sets of material samples were subjected to pre-vertical combustion tests, and the entire process of the pre-vertical combustion tests was recorded. Based on the results of the full process of the pre-vertical combustion tests, a dataset of material sample combustion images was obtained. The acquired material spline combustion image dataset is preprocessed; The pre-processed material spline combustion image dataset is used to train a pre-defined combustion flame region detection model to obtain a trained combustion flame region detection model.

[0008] Preferably, the full-process data recording is video data of the material sample combustion state recorded during the entire process of the pre-vertical combustion test, and the material sample combustion state includes: material sample combustion shape, material sample combustion flame outline, and material sample flame extinguishing; The video data is extracted frame by frame to obtain the material spline combustion state image corresponding to each video frame. The combustion areas of the spline in the material spline combustion state image are labeled. All labeled material spline combustion state images constitute a material spline combustion image dataset. The preprocessing of the material spline combustion image dataset includes: image data enhancement, initial random image rearrangement, and secondary random image rearrangement. Specifically, after image data enhancement, the images in the material spline combustion image dataset are subjected to initial random image rearrangement, and after the initial random image rearrangement, secondary random image rearrangement is performed.

[0009] By using the above-mentioned techniques, the generalization ability of the combustion flame region detection model can be improved by training the model with the preprocessed material spline combustion image dataset.

[0010] Preferably, each frame of the burning image captured in real time is used as the input to the pre-trained burning flame region detection model, and the pre-trained burning flame region detection model is used to output the flame region recognition detection box of each frame of the burning image. The pre-trained combustion flame region detection model also identifies whether the combustion flame of the material spline has been extinguished. If, after q consecutive frames of combustion images are input into the combustion flame region detection model, the output of the combustion flame region detection model does not contain any flame region recognition detection boxes, then the combustion flame of the material spline has been extinguished. q is the set threshold for the flame extinguishing image frame.

[0011] Preferably, the process of obtaining the core reference point of the flame area in the horizontal direction and the coordinate data of the core reference point in each frame of the combustion image is as follows: Extract the boundary coordinates (x1, y1, x2, y2) of the flame region detection box, where (x1, y1) is the coordinate of the upper left corner of the flame region detection box, and (x2, y2) is the coordinate of the lower right corner of the flame region detection box. Take the average of the sum of the x-coordinates of the upper left and lower right corners, and determine the position point corresponding to this average as the core reference point of the flame region in the horizontal direction for each frame of the combustion image. The expression for calculating the position data center_x of the core reference point is as follows: center_x = (x1 + x2) / 2; Based on the boundary coordinates (x1, y1, x2, y2) of the flame region detection box, the flame region image ROI of the combustion image is extracted, and the flame region image ROI is converted into a grayscale image. Edge contour recognition is performed on the grayscale image to obtain a binarized edge map.

[0012] Preferably, when constructing the vertical scanning reference line of the flame region, on the binarized edge map, a perpendicular line is drawn from the core reference point to the lower border of the flame region recognition and detection box to obtain the vertical scanning reference line of the flame region. Starting from the bottom of the binarized edge map, pixel scanning is performed row by row along the scanning reference line. When an edge pixel is scanned, the edge pixel is taken as the combustion endpoint, and the ordinate boundary_y of the edge pixel in the current frame combustion image is recorded. The ordinate boundary_y is used as the position data of the combustion endpoint. Suppose that m consecutive burning images are captured in real time, the position data of the reference point acquired in the i-th frame is center_xi, and the position data of the burning endpoint is boundary_yi, i=1,2,...,m. The position data of the core reference point and the position data of the burning endpoint acquired frame by frame are combined and represented as position control data L. The expression of L is: L={(center_x1, boundary_y1), (center_x2, boundary_y2),...,(center_xi, boundary_yi),...,(center_xm, boundary_ym)}.

[0013] Preferably, the process of controlling the movement trajectory of the flame application position using position control data until the vertical combustion test ends and obtaining the test results is as follows: SA: Within the set flame application time, real-time position control data is used as the input to drive the flame application position change, moving the flame application position to apply flame to the test material sample for the first time; SB: After the flame application is complete, the flame is horizontally withdrawn, and the time t1 of the horizontal withdrawal of the flame is recorded; SC: Real-time detection of the combustion state of the material sample, which includes: the combustion shape of the material sample, the flame outline of the material sample, and the flame extinguishing of the material sample; if the flame of the material sample is extinguished, the flame extinguishing time t2 is recorded, and the duration of the first flame application ΔT1 is obtained by subtracting the flame extinguishing time t2 from the flame horizontal withdrawal time t1, and the flame is reset to the position before withdrawal; SD: Within the set flame application time, real-time position control data is used as the input to drive the flame application position change, moving the flame application position to apply flame to the test material sample a second time; SE: After flame application is complete, the flame is horizontally withdrawn, and the flame horizontal withdrawal time t3 is recorded; SF: Real-time detection of the combustion state of the material sample, which includes: the combustion shape of the material sample, the flame outline of the material sample, and the flame extinguishing of the material sample; if the flame of the material sample is extinguished, the flame extinguishing time t4 is recorded, and the duration of the second flame application ΔT2 is obtained by subtracting the flame extinguishing time t4 from the flame horizontal withdrawal time t3, and the flame is reset to the initial standby flame application position to wait for the next vertical combustion test of the material sample; SG: After the flame is extinguished, extract the flame extinguishing image and perform color conversion on the flame extinguishing image. Based on the color conversion result, determine the residual flame after the flame is extinguished. SH: Record the time t5 for the embers to extinguish after the flame is extinguished. Subtract the ember extinguishing time t5 from the flame extinguishing time t4 to obtain the ember duration ΔT3. SI: The flame retardancy rating of the material is evaluated based on the standards corresponding to the flame duration ΔT1 of the first flame application, the flame duration ΔT2 of the second flame application, and the afterglow duration ΔT3, and the test results are obtained.

[0014] Through the above-mentioned technical means, the real-time position control data is obtained by the combustion flame area detection model by identifying the flame area frame by frame in each captured combustion image. The real-time position control data can control the movement of the flame application position in real time, realizing real-time flame tracking.

[0015] Secondly, this application also proposes a material vertical combustion testing system, the system being used to implement the aforementioned material vertical combustion testing method, comprising: The flame application module is used to apply a flame to the sample of the material to be tested; The flame region recognition module is used to capture each frame of the combustion image of the material spline in real time, and to perform frame-by-frame recognition of the flame region in each frame of the combustion image using a pre-trained combustion flame region detection model. The horizontal reference point localization module, based on the flame region identified frame by frame, obtains the core reference point and the position data of the core reference point in the horizontal direction of the flame region in each frame of the combustion image. The combustion endpoint localization module uses the core reference point of the flame area in the horizontal direction of each frame of combustion image as a benchmark to construct a scanning reference line of the flame area in the vertical direction. Based on the scanning reference line, it obtains the combustion endpoint and the position data of the flame area of ​​each frame of combustion image. The control module uses the position data of the core reference point and the position data of the combustion endpoint acquired frame by frame as position control data. It uses the position control data to control the movement trajectory of the flame position until the vertical combustion test ends and the test results are obtained.

[0016] Thirdly, this application proposes a material vertical combustion testing device, the device comprising: the material vertical combustion testing system, and further comprising: a device body, a material strip clamp for holding material strips and a moving mechanism, wherein the flame area identification module and the material strip clamp are both disposed on the device body, and the moving mechanism drives the flame application module to move closer to or away from the material strip clamp.

[0017] Preferably, the moving mechanism includes a first driver, a driving slider, a driving slider guide rail, a second driver, a supporting slider, and a supporting slider guide rail; The flame application module is mounted on the support slider, the support slider guide rail is arranged vertically, the support slider is slidably connected to the support slider guide rail, and the first driver drives the support slider to slide. The support slider guide rail is connected to the drive slider. The drive slider guide rail is arranged in a horizontal direction. The drive slider is slidably connected to the drive slider guide rail from left to right. The second driver drives the drive slider to slide.

[0018] Compared with the prior art, the beneficial effects of the technical solution of the present invention are: This invention proposes a method, system, and apparatus for vertical combustion testing of materials. For the vertical combustion testing process, it integrates machine vision and utilizes a pre-trained flame region detection model to identify the flame region frame by frame in each combustion image. Based on the identified flame region, it obtains the core reference point and combustion endpoint of the flame region in the horizontal direction for each frame. The coordinate changes of the core reference point and combustion endpoint in each frame can be extracted. The position data of the core reference point and combustion endpoint are used to control the movement of the flame application position, improving the dynamic response capability of the applied flame, ensuring that the flame always tracks the combustion endpoint, improving accuracy, avoiding full human intervention, increasing testing efficiency, and ensuring the safety of personnel. The apparatus proposed in this invention automatically controls the movement of the flame application position based on the position data of the core reference point and combustion endpoint, improving the accuracy of the test. Attached Figure Description

[0019] Figure 1 A schematic flowchart illustrating the vertical combustion test method for materials proposed in this embodiment of the invention; Figure 2 This image shows the flame region recognition result output by the pre-trained combustion flame region detection model proposed in this embodiment of the invention. Figure 3 This is a schematic diagram illustrating the conversion of a flame region image into a grayscale image, as proposed in an embodiment of the present invention. Figure 4 A schematic diagram showing the edge contour of the flame region extracted in an embodiment of the present invention; Figure 5 A schematic diagram showing the vertical scanning reference line of the flame region proposed in the embodiments of the present invention; Figure 6 This is a schematic diagram showing the change of the position data of the core reference point of the m-frame continuous combustion image proposed in this embodiment of the invention with the frame order. Figure 7 A schematic diagram of curve B showing the change of the position data of the combustion endpoints of m consecutive combustion images in the embodiments of the present invention with the frame order; Figure 8 A schematic diagram illustrating the flame afterburner after color conversion as proposed in an embodiment of the present invention; Figure 9 This diagram illustrates the structure of the vertical combustion testing system for materials proposed in this embodiment of the invention. Figure 10 This diagram illustrates the structure of the vertical combustion testing device for materials proposed in an embodiment of the present invention.

[0020] Among them, 1-material spline holder; 2-Bunsen lamp; 3-high-definition industrial camera; 4-second driver; 5-first driver; 6-drive slider; 7-support slider; 8-drive slider guide rail; 9-support slider guide rail; 10-device body. Detailed Implementation

[0021] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. To better illustrate this embodiment, some parts of the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions; It is understandable to those skilled in the art that some well-known details may be omitted from the accompanying drawings.

[0022] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] The positional relationships depicted in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent. Example 1 This embodiment proposes a method for vertical burning testing of materials. All embodiments in this application use a 50W vertical burning test as an example. The flowchart of this method can be found here. Figure 1 This includes the following steps: S1: Apply flame to the material sample to be tested.

[0024] S2: Real-time capture of each frame of combustion images of the material spline, and frame-by-frame identification of the flame region in each frame of combustion image using a pre-trained combustion flame region detection model.

[0025] S3: Based on the flame region identified frame by frame, obtain the core reference point and the position data of the core reference point in the horizontal direction of the flame region in each frame of the combustion image.

[0026] S4: Using the core reference point of the flame area in the horizontal direction of each frame of the combustion image as a reference, construct the scanning reference line of the flame area in the vertical direction. Based on the scanning reference line, obtain the combustion endpoint and position data of the flame area in each frame of the combustion image.

[0027] S5: Using the position data of the core reference point and the position data of the combustion endpoint acquired frame by frame as position control data, the movement trajectory of the flame application position is controlled by the position control data until the vertical combustion test ends, and the test results are obtained.

[0028] The vertical combustion test method for materials proposed in this embodiment is generally divided into three parts. The first part involves real-time identification of the combustion state of the material sample and extraction of the core reference point and combustion endpoint for each frame, recording the position data. The second part involves controlling the movement of the flame application position based on the position data (in vertical combustion tests, a Bunsen burner is generally used to apply the flame, which is equivalent to controlling the movement trajectory of the Bunsen burner). The third part involves obtaining the test results based on the entire vertical combustion test process. Currently, the vertical combustion test standard stipulates that during the test, a strip-shaped material sample with a length of 125±5mm is used, and the material sample is vertically clamped and fixed, with the central axis of the Bunsen burner remaining vertical. The material sample is then flamed twice with a flame at a fixed height. During the flame application process, the Bunsen burner nozzle must be 10±1mm away from the midpoint of the bottom surface of the material sample for 10±0.5s, and the Bunsen burner is moved to track changes in the length and position of the material sample.

[0029] In this embodiment, a machine vision algorithm is used to identify the flame region of each frame of the combustion image frame by frame using a pre-trained combustion flame region detection model. For the vertical combustion test of materials, machine vision is integrated, and the pre-trained combustion flame region detection model is used to identify the flame region of each frame of the combustion image frame by frame. Based on the flame region identified frame by frame, the core reference point and combustion endpoint of the flame region in the horizontal direction of each frame of the combustion image are obtained. The coordinate changes of the core reference point and combustion endpoint in each frame can be extracted. The position data of the core reference point and combustion endpoint are used to control the movement of the flame application position, which improves the dynamic response capability of the applied flame, ensures that the flame always tracks the combustion endpoint, avoids full human intervention, improves test efficiency and accuracy, and also ensures the safety of the staff.

[0030] Example 2 In this embodiment, the process of pre-training the combustion flame region detection model is as follows: S21: Perform pre-vertical combustion tests on multiple sets of material samples and record the entire process of the pre-vertical combustion tests. Based on the results of the full process data recording of the pre-vertical combustion tests, obtain a dataset of material sample combustion images. S22: Preprocess the acquired material spline combustion image dataset; S23: Use the preprocessed material spline combustion image dataset to train the preset combustion flame region detection model to obtain the trained combustion flame region detection model.

[0031] In this embodiment, the full-process data recording is video data of the material sample combustion state recorded throughout the entire process of the pre-vertical combustion test. The material sample combustion state includes: material sample combustion shape, material sample combustion flame outline, and material sample flame extinguishing. The video data is extracted frame by frame to obtain the material spline combustion state image corresponding to each video frame. The combustion areas of the spline in the material spline combustion state image are labeled. All labeled material spline combustion state images constitute a material spline combustion image dataset. The preprocessing of the material spline combustion image dataset includes: image data enhancement, initial random image rearrangement, and secondary random image rearrangement. Specifically, after image data enhancement, the images in the material spline combustion image dataset are subjected to initial random image rearrangement, and after the initial random image rearrangement, secondary random image rearrangement is performed.

[0032] In practice, multiple sets of splines underwent pre-vertical combustion tests. High-definition industrial cameras were deployed to record the entire pre-vertical combustion test process, ensuring the acquisition of clear video data of the spline combustion morphology, flame outline, and flame extinguishing. Then, a Python script (using an OpenCV module) was used to extract images of the flame combustion and extinguishing states for each frame, forming an image set. Using the Labelimg tool, the combustion areas of the images in the image set were labeled using bounding boxes and semantic tags, automatically generating a Labels folder containing the labeled information, thus constructing a material spline combustion image dataset.

[0033] In this embodiment, the labeled images are enhanced in color space, geometric transformation, and image quality using a Python script (Albumentations library), thereby expanding the size of the image dataset. This allows subsequent model predictions to use richer image data and improves the accuracy of the prediction results.

[0034] After image data augmentation of the material spline combustion image dataset, an initial random rearrangement of the images is performed. A Python script (using the `Train_test_split` function from the Scikit-learn library) divides the positive and negative samples (flame extinguished) of the combustion image dataset into a 6:2:2 ratio. During this division, an initial random rearrangement is performed, and then the positive and negative samples are merged to form complete training, validation, and test sets. The training set is used for model training, the validation set is used to adjust model performance, and the test set is used for the final evaluation of the combustion flame region detection model. Following the initial random rearrangement, a second random rearrangement is performed. A Python script (using the `Random.shuffle` function from the NumPy library) globally shuffles the merged training, validation, and test sets a second time to enhance data randomness and avoid potential impacts of sample order on model training. Then, the preprocessed material spline combustion image dataset is used to train the combustion flame region detection model, which improves the model's generalization ability.

[0035] In this embodiment, the selected preset combustion flame region detection model is the YOLOv8n lightweight object detection model. When training the preset combustion flame region detection model using a preprocessed material spline combustion image dataset, the training epochs are set to 200, the initial learning rate to 0.001, and the batch size to 16. The optimizer used is Adam adaptive moment estimation, and L2 weight decay regularization is employed to suppress overfitting. The training set obtained from the signing process is input into the YOLOv8n lightweight object detection model for training. During training, the performance of the YOLOv8n lightweight object detection model is evaluated in real time using a validation set. Performance monitoring metrics include loss function, accuracy, recall, and mean precision. Based on the feedback from the validation set, the model's hyperparameters are dynamically adjusted, and training is iteratively continued until the model converges (the loss function tends to stabilize, and the evaluation metrics show no significant improvement). The optimal model weight file (Best.pt) is saved for subsequent recognition and detection, thus obtaining the pre-trained combustion flame region detection model.

[0036] When using the pre-trained combustion flame region detection model for flame region recognition, the saved optimal model weights are loaded, and each frame of the real-time captured combustion image is used as the input to the pre-trained combustion flame region detection model. The pre-trained combustion flame region detection model outputs a flame region recognition detection box for each frame of the combustion image. For vertical combustion test scenarios, the trained combustion flame region detection model proposed in this embodiment captures the test process in real time, performs flame region recognition on each frame of the image, and outputs a detection box with confidence and a label. The flame region recognition result image can be found in [reference needed]. Figure 2 ,exist Figure 2 In the test box, the burn point is indicated, along with a confidence level of 0.89.

[0037] In this embodiment, the pre-trained flame region detection model also simultaneously identifies whether the flame of the material spline has been extinguished. If, after inputting q consecutive frames of combustion images into the flame region detection model, the model outputs no flame region detection boxes, then the flame of the material spline is extinguished. Here, q is the set threshold for the number of extinguished image frames. In this embodiment, negative samples are not labeled, i.e., they have no detection boxes. Python evaluation determines that the flame is extinguished when there are no flames in 3 consecutive frames.

[0038] In this embodiment, the process of obtaining the core reference point and coordinate data of the flame area in the horizontal direction of each frame of the combustion image is as follows: For the flame region detection box of each frame of the burning image, based on the Box.xyxy attribute output by the model and combined with the Astype method of the NumPy library, the boundary coordinates (x1, y1, x2, y2) of the flame region detection box are extracted, where (x1, y1) is the coordinate of the upper left corner of the flame region detection box and (x2, y2) is the coordinate of the lower right corner of the flame region detection box. The average of the sum of the x-coordinates of the upper left and lower right corners is taken, and the position point corresponding to the average value is determined as the core reference point of the flame region in the horizontal direction of each frame of the burning image. The expression for calculating the position data center_x of the core reference point is as follows: center_x = (x1 + x2) / 2; This value represents the core reference point of the flame area in the horizontal direction. Subsequently, by tracking the change of center_x with each frame until the flame is extinguished, the horizontal displacement trend of the combustion endpoint during a single flame application can be quantified. At the same time, it can also be used as a horizontal reference position to find the vertical displacement of the combustion endpoint.

[0039] Based on the boundary coordinates (x1, y1, x2, y2) of the flame region detection box, the flame region image ROI of the combustion image is extracted using the NumPy array slicing operation (img[y1:y2, x1:x2]). In specific implementation, the OpenCV image storage format ((height, width, channels), corresponding to (y-axis range, x-axis range, color channels)) is followed. The slicing format of img[y1:y2, x1:x2] is used to directly crop the region within the detection box from the original image, ensuring that the cropped ROI completely contains the flame shape, providing an analysis object for subsequent analysis of the vertical displacement of the combustion endpoint. The flame region image ROI is then converted into a grayscale image, as shown below. Figure 3 As shown, this eliminates color interference, providing a high-contrast, low-noise input basis for subsequent edge detection. Then, based on the Canny edge detection algorithm, edge contour recognition is performed on the grayscale image to obtain a binarized edge map, as shown below. Figure 4 As shown, it can effectively extract continuous, single-pixel-width edge contours of the flame region, providing a feature basis for the localization of the combustion endpoint.

[0040] When constructing the vertical scanning reference line of the flame region, on the binarized edge map, a perpendicular line is drawn from the core reference point to the lower border of the flame region recognition and detection box, resulting in the vertical scanning reference line of the flame region as shown below. Figure 5 As shown, starting from the bottom of the binarized edge map, pixel scanning is performed row by row upwards along the scanning reference line. When an edge pixel is scanned, it is taken as the burning endpoint. The edge pixel value is 255, which is greater than 0, while the non-edge pixel value is 0. Figure 3The image shows a white edge outline and a black non-edge area. The ordinate (boundary_y) of the edge pixels in the current frame's burning image is recorded, and this ordinate (boundary_y) is used as the position data of the burning endpoint.

[0041] Subsequently, the vertical displacement trend of the combustion endpoint during a single flame application can be quantified by tracking the changes in boundary_y with each frame until the flame is extinguished.

[0042] Suppose that m consecutive burning images are captured in real time, the position data of the reference point acquired in the i-th frame is center_xi, and the position data of the burning endpoint is boundary_yi, i=1,2,...,m. The position data of the core reference point and the position data of the burning endpoint acquired frame by frame are combined and represented as position control data L. The expression of L is: L={(center_x1, boundary_y1), (center_x2, boundary_y2),...,(center_xi, boundary_yi),...,(center_xm, boundary_ym)}.

[0043] Using the frame order of position control data L as the x-axis and the position data of the core reference point corresponding to each frame order in position control data L as the y-axis, a curve A is plotted to show the change of the position data of the core reference point with the frame order in m consecutive combustion images. A schematic diagram of the curve can be found in [reference needed]. Figure 6 Using the frame sequence of position control data L as the x-axis and the position data of the combustion endpoint corresponding to each frame sequence in position control data L as the y-axis, plot curve B of the change in the position data of the combustion endpoint as a function of the frame sequence in m consecutive combustion images. A schematic diagram of the curve can be found in [reference needed]. Figure 7 Then, the movement trajectory of the flame application position is controlled using position control data until the vertical combustion test ends, and the process of obtaining the test results is as follows: SA: Within the set flame application time, real-time position control data is used as the input to drive the flame application position change, moving the flame application position to apply flame to the test material sample for the first time; SB: After the flame application is complete, the flame is horizontally withdrawn, and the time t1 of the horizontal withdrawal of the flame is recorded; SC: Real-time detection of the combustion state of the material sample, which includes: the combustion shape of the material sample, the flame outline of the material sample, and the flame extinguishing of the material sample; if the flame of the material sample is extinguished, the flame extinguishing time t2 is recorded, and the duration of the first flame application ΔT1 is obtained by subtracting the flame extinguishing time t2 from the flame horizontal withdrawal time t1, and the flame is reset to the position before withdrawal; SD: Within the set flame application time, real-time position control data is used as the input to drive the flame application position change, moving the flame application position to apply flame to the test material sample a second time; SE: After flame application is complete, the flame is horizontally withdrawn, and the flame horizontal withdrawal time t3 is recorded; SF: Real-time detection of the combustion state of the material sample, which includes: the combustion shape of the material sample, the flame outline of the material sample, and the flame extinguishing of the material sample; if the flame of the material sample is extinguished, the flame extinguishing time t4 is recorded, and the duration of the second flame application ΔT2 is obtained by subtracting the flame extinguishing time t4 from the flame horizontal withdrawal time t3, and the flame is reset to the initial standby flame application position to wait for the next vertical combustion test of the material sample; SG: After the flame is extinguished, extract the flame extinguishing image and perform color conversion on the flame extinguishing image. Based on the color conversion result, determine the residual flame after the flame is extinguished. In this embodiment, the process of extracting the flame extinguishing image is similar to acquiring the flame combustion image. Image frames are extracted from the video recorded during the vertical combustion test. The process of determining the afterglow of the flame after extinguishing the flame involves color conversion of the flame extinguishing image, such as... Figure 8 As shown, the colors of the flame extinguishing image are converted from the RGB color space to the HSV color space. A threshold is set for the red / orange flame, and the color range is defined as [0,100,100] to [15,255,255]. Colors that fit into this range are converted to white, and the remaining areas are black, thus determining the residual burn.

[0044] SH: Record the time t5 for the embers to extinguish after the flame is extinguished. Subtract the ember extinguishing time t5 from the flame extinguishing time t4 to obtain the ember duration ΔT3. SI: The flame retardancy rating of the material is evaluated based on the standards corresponding to the flame duration ΔT1 of the first flame application, the flame duration ΔT2 of the second flame application, and the afterglow duration ΔT3, and the test results are obtained.

[0045] In this embodiment, a set of tests requires testing 5 material samples. The two flame tests mentioned above are for testing a single sample. The tests need to be repeated 5 times to evaluate the flame retardancy rating. The flame retardancy rating includes: V-0, V-1, and V-2.

[0046] Among them, the flame duration (ΔT1, ΔT2) of a single material sample is less than or equal to 10s, and the total afterflame time t of a group of 5 samples under any state adjustment is less than or equal to 50s, and ΔT2+ΔT3 is less than or equal to 30s, and the afterflame does not burn to the clamping fixture, and if there are drips, the drips do not ignite the cotton pad, then it is judged as V-0. The criteria for determining the flame retardancy rating can be found in Table 1.

[0047] Table 1

[0048] Through the above-mentioned technical means, the real-time position control data is obtained by the combustion flame area detection model by identifying the flame area frame by frame in each captured combustion image. The real-time position control data can control the movement of the flame application position in real time, realizing real-time flame tracking.

[0049] Example 3 This embodiment also proposes a material vertical combustion testing system, such as... Figure 8 As shown, the system is used to implement the material vertical combustion test method described in the foregoing embodiments, including: The flame application module is used to apply a flame to the sample of the material to be tested; The flame region recognition module is used to capture each frame of the combustion image of the material spline in real time, and to perform frame-by-frame recognition of the flame region in each frame of the combustion image using a pre-trained combustion flame region detection model. The horizontal reference point localization module, based on the flame region identified frame by frame, obtains the core reference point and the position data of the core reference point in the horizontal direction of the flame region in each frame of the combustion image. The combustion endpoint localization module uses the core reference point of the flame area in the horizontal direction of each frame of combustion image as a benchmark to construct a scanning reference line of the flame area in the vertical direction. Based on the scanning reference line, it obtains the combustion endpoint and the position data of the flame area of ​​each frame of combustion image. The control module uses the position data of the core reference point and the position data of the combustion endpoint acquired frame by frame as position control data. It uses the position control data to control the movement trajectory of the flame position until the vertical combustion test ends and the test results are obtained.

[0050] Example 4 This embodiment proposes a material vertical combustion testing device, see [link / reference] Figure 9 The device includes: the material vertical combustion test system described in the foregoing embodiment, and further includes: device body 10, material strip clamp 1 for holding material strips and a moving mechanism. The flame area identification module and the material strip clamp 1 are both disposed on the device body 10, and the moving mechanism drives the flame application module to move closer to or away from the material strip clamp 1.

[0051] In this embodiment, the flame area recognition module includes: a high-definition industrial camera 3 and a recognition module encapsulating the combustion flame area detection model described in embodiments 1-2. See also... Figure 9The moving mechanism includes a first driver 5, a driving slider 6, a driving slider guide rail 8, a second driver 4, a supporting slider 7, and a supporting slider guide rail 9. In this embodiment, both the first driver 5 and the second driver 4 are motors. The flame-emitting module in this embodiment is a Bunsen lamp 2, which is mounted on the supporting slider 7. The supporting slider guide rail 9 is arranged vertically, and the supporting slider 7 is slidably connected to the supporting slider guide rail 9. The first driver 5 drives the supporting slider 7 to slide. The supporting slider guide rail 9 is connected to the driving slider 6, which is arranged horizontally. The driving slider 6 is slidably connected to the driving slider guide rail 8. The second driver 4 drives the driving slider 6 to slide.

[0052] In practice, the material sample to be tested is vertically fixed in the material sample holder 1 to ensure that the material sample hangs vertically. The Bunsen lamp 2 is pre-positioned in the initial standby position (at a horizontal distance of ≥150mm from the sample), and the flame parameters (height and flame color) are adjusted according to the standard requirements.

[0053] In this embodiment, the second driver 4 is reversed by the Siemens TIA Portal control program, which in turn drives the slider 6 to move the Bunsen lamp 2 to the center point of the bottom surface of the material sample (the initial flame position). Once in position, the PLC timing command (Ton module) is triggered to prepare for flame application.

[0054] The initial flame tracking control of Bunsen burner 2 is performed. After Bunsen burner 2 reaches the initial flame position, the PLC starts the flame application after a 10-second timer. During the flame application, the high-definition industrial camera 3 captures the entire test process. Based on the combustion endpoint identified by the vertical combustion test method proposed in Examples 1-2, real-time position data (center_xi, boundary_yi) is extracted. Figure 6 The trend of center_x during the test process is shown as follows: Figure 7 The trend of boundary_y during the test is shown. A Socket connection is established with the Siemens PLC by calling the Snap7 library through a Python script. The extracted position control data center_xi and boundary_yi are written into the PLC's DB data block in real time and converted into displacement data according to the ratio. The Siemens control program reads the displacement data in the DB block and drives the second driver 4 and the first driver 5 to work together. This drives the slider 6 and the support slider 7 to adjust the spatial position of the Bunsen lamp 2, ensuring that the flame always tracks the combustion endpoint.

[0055] The Bunsen burner is horizontally withdrawn, and its flame status is monitored. After the Bunsen burner completes 10 seconds of flame application, the second driver 4 is driven to rotate forward using the Siemens control program, which in turn drives the slider 6 to move along the drive slider guide rail 8, causing the Bunsen burner 2 to horizontally withdraw 150mm away from the sample. In specific implementation, if the combustion endpoint is still changing during the horizontal withdrawal of the Bunsen burner 2, the flame displacement data can be fed back to control the adjustment of the Bunsen burner's position, ensuring that the initial position of the combustion endpoint of the sample meets the standard requirement distance during the secondary flame application. Of course, the displacement change of the combustion endpoint when it is still changing is generally within the allowable displacement change error range. Based on the method proposed in Examples 1 and 2, the flame status is detected in real time. When no effective flame characteristics are detected for 3 consecutive frames or more, the flame is determined to be extinguished. If the flame is detected to be extinguished, the Bunsen burner must be immediately reset and a secondary flame application performed, even if it has not withdrawn more than 150mm away from the sample.

[0056] The second flame application control of the Bunsen burner is performed. Once the flame is determined to be extinguished, the following process is initiated: Based on the Bunsen burner's position before the initial removal, the Siemens control program drives the second actuator 4 to operate, which in turn drives the slider 6 to move along the slider guide rail 8, resetting the Bunsen burner 2 to its initial position. (Specifically, if the combustion endpoint is still changing after the Bunsen burner 2 is horizontally removed to a distance of 150mm from the sample, before resetting the Bunsen burner 2 to its initial position, the flame displacement data detected during the initial removal process can be used to reset the Bunsen burner 2 to the specific secondary flame application position). After the Bunsen burner is reset, the aforementioned process is repeated, with the Bunsen burner 2 applying a secondary flame to the material sample at a 10-second flame application cycle, real-time identification of the combustion endpoint position data, and dynamic tracking of the Bunsen burner 2.

[0057] After the 10-second flame application period, the Bunsen burner 2 is horizontally withdrawn to a distance of 150mm from the sample. If no valid flame characteristics are detected for three consecutive frames, the flame is determined to be extinguished after the second flame application. When the flame is finally determined to be extinguished, the second flame application cycle ends, and the PLC control program drives the Bunsen burner 2 to return to the initial standby flame application position, waiting for the next combustion test of the sample to be tested. The initial standby flame application position is the flame application position that has been determined before the flame application according to the specified material vertical combustion test standard.

[0058] The embodiments described are merely examples to clearly illustrate the present invention and are not intended to limit the implementation of the invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively describe all possible implementations. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for testing the vertical burning of materials, characterized in that, Includes the following steps: Apply a flame to the material specimen to be tested; Real-time capture of each frame of combustion images of the material spline, and frame-by-frame identification of the flame region in each frame of combustion image using a pre-trained combustion flame region detection model; Based on the flame region identified frame by frame, the core reference point and the position data of the core reference point in the horizontal direction of the flame region in each frame of the combustion image are obtained. Using the core reference point of the flame region in the horizontal direction of each frame of the burning image as a benchmark, a scanning reference line for the flame region in the vertical direction is constructed. Based on the scanning reference line, the burning endpoint and position data of the flame region in each frame of the burning image are obtained. Pixels are scanned upwards line by line along the scanning reference line. When an edge pixel is scanned, the edge pixel is taken as the burning endpoint. The position data of the core reference point and the position data of the combustion endpoint, which are acquired frame by frame, are used as position control data. The movement trajectory of the flame application position is controlled by the position control data until the vertical combustion test ends, and the test results are obtained.

2. The method for vertical burning testing of materials according to claim 1, characterized in that, The process of pre-training the combustion flame region detection model is as follows: Multiple sets of material samples were subjected to pre-vertical combustion tests, and the entire process of the pre-vertical combustion tests was recorded. Based on the results of the full process of the pre-vertical combustion tests, a dataset of material sample combustion images was obtained. The acquired material spline combustion image dataset is preprocessed; The pre-processed material spline combustion image dataset is used to train a pre-defined combustion flame region detection model to obtain a trained combustion flame region detection model.

3. The method for vertical burning testing of materials according to claim 2, characterized in that, The full-process data record consists of video data of the material spline combustion state recorded during the entire process of the pre-vertical combustion test. The material spline combustion state includes: material spline combustion shape, material spline combustion flame outline, and material spline flame extinguishing. The video data is extracted frame by frame to obtain the material spline combustion state image corresponding to each video frame. The spline combustion area in the material spline combustion state image is labeled. All labeled material spline combustion state images constitute a material spline combustion image dataset. The preprocessing of the material spline combustion image dataset includes: image data enhancement, initial random image rearrangement, and secondary random image rearrangement. Specifically, after image data enhancement, the images in the material spline combustion image dataset are subjected to initial random image rearrangement, and after the initial random image rearrangement, secondary random image rearrangement is performed.

4. The method for vertical burning testing of materials according to claim 1, characterized in that, Each frame of the burning image captured in real time is used as the input to the pre-trained burning flame region detection model, and the pre-trained burning flame region detection model is used to output the flame region recognition detection box of each frame of the burning image. The pre-trained combustion flame region detection model also identifies whether the combustion flame of the material spline has been extinguished. If, after q consecutive frames of combustion images are input into the combustion flame region detection model, the output of the combustion flame region detection model does not contain any flame region recognition detection boxes, then the combustion flame of the material spline has been extinguished. q is the set threshold for the flame extinguishing image frame.

5. The method for vertical burning testing of materials according to claim 4, characterized in that, The process of acquiring the core reference point and its position data in the horizontal direction of the flame area in each frame of the combustion image is as follows: Extract the boundary coordinates (x1, y1, x2, y2) of the flame region recognition detection box, where (x1, y1) is the coordinate of the upper left corner of the flame region recognition detection box, and (x2, y2) is the coordinate of the lower right corner of the flame region recognition detection box. Take the average of the sum of the x-coordinates of the upper left corner coordinate and the x-coordinates of the lower right corner coordinate, and determine the position point corresponding to the average value as the core reference point of the flame region in the horizontal direction of each frame of the combustion image. The calculation expression of the position data center_x of the core reference point is: center_x=(x1+x2) / 2; Based on the boundary coordinates (x1, y1, x2, y2) of the flame region detection box, the flame region image ROI of the combustion image is extracted, and the flame region image ROI is converted into a grayscale image. Edge contour recognition is performed on the grayscale image to obtain a binarized edge map.

6. The method for vertical burning testing of materials according to claim 5, characterized in that, When constructing the vertical scanning reference line of the flame region, on the binarized edge map, a perpendicular line is drawn from the core reference point to the lower border of the flame region recognition and detection box to obtain the vertical scanning reference line of the flame region. Starting from the bottom of the binarized edge map, pixel scanning is performed row by row along the scanning reference line. When an edge pixel is scanned, the edge pixel is taken as the combustion endpoint, and the ordinate boundary_y of the edge pixel in the current frame combustion image is recorded. The ordinate boundary_y is used as the position data of the combustion endpoint. Suppose that m consecutive burning images are captured in real time, the position data of the core reference point acquired in the i-th frame is center_xi, and the position data of the burning endpoint is boundary_yi, i=1,2,...,m. The position data of the core reference point and the position data of the burning endpoint acquired frame by frame are combined into position control data L, and the expression of L is: L={(center_x1, boundary_y1), (center_x2, boundary_y2),...,(center_xi, boundary_yi),...,(center_xm, boundary_ym)}.

7. The method for vertical burning testing of materials according to claim 6, characterized in that, The process of controlling the movement trajectory of the flame application position using position control data until the vertical combustion test ends and obtaining the test results is as follows: SA: Within the set flame application time, move the flame application position to apply flame to the test material sample for the first time; SB: After the flame application is complete, the flame is horizontally withdrawn, and the time t1 of the horizontal withdrawal of the flame is recorded; SC: Real-time detection of the combustion state of the material sample, which includes: the combustion shape of the material sample, the flame outline of the material sample, and the flame extinguishing of the material sample; if the flame of the material sample is extinguished, the flame extinguishing time t2 is recorded, and the duration of the first flame application ΔT1 is obtained by subtracting the flame extinguishing time t2 from the flame horizontal withdrawal time t1, and the flame is reset to the position before withdrawal; SD: Within the set flame application time, real-time position control data is used as the input to drive the flame application position change, moving the flame application position to apply flame to the test material sample a second time; SE: Flame application complete, remove the flame horizontally, and record the flame horizontal removal time t3; SF: Real-time detection of the combustion state of the material sample, which includes: the combustion shape of the material sample, the flame outline of the material sample, and the flame extinguishing of the material sample; if the flame of the material sample is extinguished, the flame extinguishing time t4 is recorded, and the duration of the second flame application ΔT2 is obtained by subtracting the flame extinguishing time t4 from the flame horizontal withdrawal time t3, and the flame is reset to the initial standby flame application position to wait for the next vertical combustion test of the material sample; SG: After the flame is extinguished, extract the flame extinguishing image and perform color conversion on the flame extinguishing image. Based on the color conversion result, determine the residual flame after the flame is extinguished. SH: Record the time t5 for the embers to extinguish after the flame is extinguished. Subtract the ember extinguishing time t5 from the flame extinguishing time t4 to obtain the ember duration ΔT3. SI: The flame retardancy rating of the material is evaluated based on the standards corresponding to the flame duration ΔT1 of the first flame application, the flame duration ΔT2 of the second flame application, and the afterglow duration ΔT3, and the test results are obtained.

8. A vertical combustion testing system for materials, characterized in that, The system is used to implement the material vertical combustion test method according to any one of claims 1 to 7, including: The flame application module is used to apply a flame to the sample of the material to be tested; The flame region recognition module is used to capture each frame of the combustion image of the material spline in real time, and to perform frame-by-frame recognition of the flame region in each frame of the combustion image using a pre-trained combustion flame region detection model. The horizontal reference point localization module, based on the flame region identified frame by frame, obtains the core reference point and the position data of the core reference point in the horizontal direction of the flame region in each frame of the combustion image. The combustion endpoint localization module uses the core reference point of the flame area in the horizontal direction of each frame of combustion image as a benchmark to construct a scanning reference line of the flame area in the vertical direction. Based on the scanning reference line, it obtains the combustion endpoint and the position data of the flame area of ​​each frame of combustion image. The control module uses the position data of the core reference point and the position data of the combustion endpoint acquired frame by frame as position control data. It uses the position control data to control the movement trajectory of the flame position until the vertical combustion test ends and the test results are obtained.

9. A material vertical combustion testing device, characterized in that, The device includes: the material vertical combustion test system of claim 8, and further includes: a device body, a material strip clamp for holding material strips and a moving mechanism, wherein the flame area identification module and the material strip clamp are both disposed on the device body, and the moving mechanism drives the flame application module to move closer to or away from the material strip clamp.

10. The material vertical combustion testing device according to claim 9, characterized in that, The moving mechanism includes a first driver, a driving slider, a driving slider guide rail, a second driver, a supporting slider, and a supporting slider guide rail; The flame application module is mounted on the support slider, the support slider guide rail is arranged vertically, the support slider is slidably connected to the support slider guide rail, and the first driver drives the support slider to slide. The support slider guide rail is connected to the drive slider. The drive slider guide rail is arranged in a horizontal direction. The drive slider is slidably connected to the drive slider guide rail from left to right. The second driver drives the drive slider to slide.

Images