Strip steel shearing fault detection method and device, electronic equipment and storage medium

Abstract

The invention discloses a strip steel shearing fault detection method and device, electronic equipment and a storage medium, and relates to the technical field of strip steel shearing fault detection. According to the technical scheme, an image sample set of the welding machine shear is obtained; training a preset convolutional neural network model according to the image sample set until the convolutional neural network model accurately recognizes image samples in a cut fault state in the image sample set; as the convolutional neural network model can accurately identify the shearing fault, when the welding machine shear executes the shearing task, the current image of the shearing discharge port is input into the trained convolutional neural network model; and determining whether the welding machine shear is in a normal working state or a shear fault state according to an output result of the trained convolutional neural network model. According to the technical scheme, the detection efficiency is high, meanwhile, error detection caused by manual naked eye observation in the prior art is reduced, and then the efficiency and accuracy of shearing fault detection are both considered.

Description

Technical Field

[0001] The present invention relates to the technical field of strip steel shearing fault detection, and in particular to a strip steel shearing fault detection method, device, electronic equipment and storage medium. Background Art

[0002] In order to achieve continuous production of strip steel, it is necessary to weld the head and tail of the strip steel in adjacent production plans. Taking the cold-rolled 2# galvanized unit on a certain production line as an example, a narrow lap welder is required to weld the head and tail of the strip steel. Therefore, the narrow lap welder is one of the important single equipment and a key equipment to ensure the continuous and stable operation of the unit. The welding quality of the welding machine greatly affects the production efficiency and production cost of the continuous annealing unit. Due to the magnetism of the shear blades used for shearing the head and tail of the strip steel, the resulting sheared strips are prone to abnormal discharge, which in turn affects the welding quality and causes frequent production line shutdowns. In response to the above technical problems, there is currently a lack of effective technical means to deal with them. Summary of the invention

[0003] The present invention provides a strip shearing fault detection method, device, electronic equipment and storage medium, which can accurately detect whether a shearing fault occurs in a welding machine, and takes into account both the efficiency and accuracy of shearing fault detection.

[0004] The embodiment of the present invention provides the following solution:

[0005] In a first aspect, an embodiment of the present invention provides a method for detecting a strip shearing fault, the method comprising:

[0006] Acquire an image sample set of the welding shear, wherein the image sample set is a collection of images of the shearing outlet of the welding shear when the welding shear is in a normal working state and a shearing fault state;

[0007] Training a preset convolutional neural network model according to the image sample set until the convolutional neural network model accurately identifies image samples in the image sample set that have a shearing fault state;

[0008] When the welding shear performs a shearing task, the current image of the shear outlet is input into the trained convolutional neural network model, wherein the shearing task is to perform the head and tail shearing of the strip to be welded;

[0009] According to the output results of the trained convolutional neural network model, it is determined whether the welding shear is in a normal working state or a shear fault state.

[0010] In an optional embodiment, when the welding shear performs a shearing task, the current image of the shearing outlet is input into a trained convolutional neural network model, including:

[0011] Determine whether the welding shear performs the shearing task according to the current image of the shearing outlet;

[0012] If so, all current images in the preset time period are input into the trained convolutional neural network model.

[0013] In an optional embodiment, judging whether the welding shear performs the shearing task according to the current image of the shearing outlet includes:

[0014] Extract the target area in the current image of the shearing outlet according to the preset pixel frame;

[0015] The target area is processed in grayscale and binarized in turn to obtain the number of white pixels in the target area;

[0016] When the number of white pixel points is greater than a preset number threshold, it is determined that the welding shear performs the shearing task.

[0017] In an optional embodiment, determining whether the welding shear is in a normal working state or a shearing fault state according to the output result of the trained convolutional neural network model includes:

[0018] The judgment output result indicates whether the amount of waste discharged from the shear outlet in a preset period of time is less than 2;

[0019] If not, make sure the welding shear is in normal working condition;

[0020] If so, it is determined that the welding shear is in a shear fault state.

[0021] In an optional embodiment, after determining that the welding shear is in a shear failure state, the method further includes:

[0022] A pop-up window is displayed according to the waste discharge quantity represented by the output result, and the display time of the pop-up window is kept not less than the preset time threshold.

[0023] In an optional embodiment, obtaining an image sample set of a welding shear includes:

[0024] Collecting a plurality of first images of different sizes when the welding machine shear is in a normal working state, and a plurality of second images of different sizes when the shear is in a shear failure state;

[0025] Performing random preprocessing on the plurality of first images and the plurality of second images respectively to increase the number of the first images and the second images, wherein the preprocessing includes rotation, cropping, scaling and flipping;

[0026] An image sample set is obtained according to a set consisting of a plurality of first images and a plurality of second images after preprocessing.

[0027] In an optional embodiment, a preset convolutional neural network model is trained according to the image sample set until the convolutional neural network model accurately identifies image samples in the image sample set that have a shearing fault state, including:

[0028] Divide the image sample set into a training set and a test set according to a preset ratio, and configure initial parameters for the convolutional neural network model;

[0029] The convolutional neural network model with initial parameter configuration is trained according to the training set, and the training ratio of misclassified image samples is adjusted according to the number of training times;

[0030] The trained convolutional neural network model is tested according to the test set, and when the test results meet the preset goals, it is determined that the convolutional neural network model has completed training.

[0031] In a second aspect, an embodiment of the present invention further provides a device for detecting a strip shearing failure, the device comprising:

[0032] An acquisition module is used to acquire an image sample set of the welding shear, wherein the image sample set is an image set of the shear outlet of the welding shear when the welding shear is in a normal working state and a shearing fault state;

[0033] A training module, used for training a preset convolutional neural network model according to the image sample set, until the convolutional neural network model accurately identifies image samples in the image sample set that are in a shearing fault state;

[0034] An input module, used for inputting the current image of the shear outlet into the trained convolutional neural network model when the welding shear performs a shearing task, wherein the shearing task is to perform a head and tail shearing task of the strip to be welded;

[0035] The determination module is used to determine whether the welding machine shear is in a normal working state or a shearing fault state according to the output result of the trained convolutional neural network model.

[0036] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory is coupled to the processor and stores instructions, and when the instructions are executed by the processor, the electronic device executes the steps of any one of the methods in the first aspect.

[0037] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the methods in the first aspect.

[0038] Compared with the prior art, the method, device, electronic device and storage medium for detecting strip shearing failure of the present invention have the following advantages:

[0039] The technical solution of the present invention is to obtain an image sample set of the welding shear; train a preset convolutional neural network model according to the image sample set until the convolutional neural network model accurately identifies the image samples in the image sample set that are in a shearing fault state; because the convolutional neural network model can accurately identify shearing faults, when the welding shear performs a shearing task, the current image of the shearing outlet is input into the trained convolutional neural network model; according to the output result of the trained convolutional neural network model, it is determined whether the welding shear is in a normal working state or a shearing fault state. The technical solution implements intelligent detection of shearing faults through artificial intelligence, has high detection efficiency, and at the same time reduces the error detection of manual naked eye observation in the prior art, thereby taking into account the efficiency and accuracy of shearing fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flow chart of a method for detecting a strip shearing failure provided by an embodiment of the present invention;

[0042] Figure 2 An image sample of a welding shear in a shearing fault state provided by an embodiment of the present invention;

[0043] Figure 3 An image sample of a welding machine shear in a normal state provided by an embodiment of the present invention;

[0044] Figure 4 A schematic diagram of a pop-up window display under an abnormal working state provided by an embodiment of the present invention;

[0045] Figure 5 A logical schematic diagram of a method for detecting a strip shearing failure provided by an embodiment of the present invention;

[0046] Figure 6 A schematic structural diagram of a device for detecting strip shearing failure provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention belong to the scope of protection of the embodiments of the present invention.

[0048] See also Figure 1 , Figure 1 The present invention provides a flow chart of a method for detecting a shearing fault of a strip steel, which can be applied to a control terminal of a welding shear, or can be used in an independently set fault detection terminal to implement shearing fault detection of the welding shear. It is sufficient to be able to run the detection method, and no specific limitation is made here. The detection method includes:

[0049] S11. Obtain an image sample set of the welding shear, wherein the image sample set is a collection of images of the shear outlet of the welding shear when the welding shear is in a normal working state and a shearing fault state.

[0050] Specifically, taking the galvanizing production line as an example, the welding machine shear is a device that the welder uses to shear the head and tail of the strip before welding. After the welding machine shear completes the shearing, the welder welds the head and tail of the two rolls of strip steel to ensure the continuous production of the galvanizing production line. After the head and tail of the strip are sheared, there is waste material. The shear outlet is used to discharge the waste material. A camera terminal can be installed on one side of the shear outlet to capture videos or images of the welding machine shear in normal working state and shearing failure state, and generate an image sample set based on the video or image.

[0051] It should be noted that the image of the welding shear in the shear fault state does not include the shear waste image, please refer to Figure 2 The image does not include the waste after the strip is sheared. After the strip is sheared, the waste is discharged normally from the shear outlet. Therefore, this type of image is defined as an image of the welding machine shear in a shearing failure state. Figure 3 The image contains waste after the strip is sheared. After the strip is sheared, the waste may be adsorbed in the welding shear due to the magnetism of the shear blade. This type of image is defined as an image of the welding shear in normal working condition.

[0052] In practical applications, since most videos or images are shot with a fixed camera position and the image samples are obtained in a single form, the convolutional neural network model training based on such image samples cannot achieve more accurate convergence, resulting in misjudgment of the convolutional neural network model in subsequent applications. Based on this, in a specific implementation, step S11 includes sub-steps S11-1 to S11-3, which are specifically described as follows:

[0053] S11-1. Collect multiple first images of different sizes when the welding machine is in normal working state, and multiple second images of different sizes when the shear is in shear fault state. When implementing image collection, a high-definition camera is added to the driving side area of ​​the welding machine. The high-tech camera is connected to a preset data processing terminal. The data processing terminal can be a PC (Personal Computer) device or a server. The hardware composition of the PC device can be configured based on actual needs and can meet the training and operation of the convolutional neural network model. The operating status of the high-definition camera can be checked through the data processing terminal, and the quantity data of the scrap steel bars discharged can be collected. The images taken by the high-definition camera collect the first image of the normal state and the second image of the abnormal state respectively.

[0054] S11-2, performing random preprocessing on the plurality of first images and the plurality of second images respectively to increase the number of first images and second images, wherein the preprocessing includes rotation, cropping, scaling and flipping. The random preprocessing of images can increase the diversity of image samples, thereby reducing the risk of overfitting in subsequent convolutional neural network model training. At the same time, adding images of different sizes to the data set makes the convolutional neural network model more adaptable to subsequent images to be identified.

[0055] S11-3, according to the set composed of the plurality of first images and the plurality of second images after preprocessing, obtain an image sample set. After randomly preprocessing the plurality of groups of first images, more images of the welding shear in a normal shearing state can be obtained; similarly, more images of the welding shear in a faulty shearing state can also be obtained, and all images are combined into an image sample set, and after obtaining the image sample set, enter step S12.

[0056] S12. Training a preset convolutional neural network model according to the image sample set until the convolutional neural network model accurately identifies image samples in the image sample set that are in a shearing fault state.

[0057] Specifically, a convolutional neural network model is constructed based on the pytorch framework. The image sample set can be divided into a training set and a test set. The convolutional neural network model is trained by the training set. After the training is completed, the convolutional neural network model is tested by the test set. The accuracy of the convolutional neural network model is determined based on the test results. When the accuracy is greater than a preset threshold, it is determined that the convolutional neural network model has completed training.

[0058] Since the training process of the convolutional neural network model is based on a large number of image samples, it is limited by the number of samples and the training time, which may lead to poor convergence of the convolutional neural network model, or the training and verification consumes more energy of the staff. Based on this, in a specific implementation, step S12 includes sub-steps S12-1 to S12-3, which are specifically described as follows:

[0059] S12-1. Divide the image sample set into a training set and a test set according to a preset ratio, and configure initial parameters for the convolutional neural network model. The preset ratio can be set based on actual needs, for example, setting the ratio of the training set to the test set to 7:3. The image samples in the image sample set can be converted to the same size, for example, 640mm×640mm. The initial parameter configuration of the convolutional neural network model includes:

[0060] 1. The first layer of convolution of the convolutional neural network model was changed from the original 6×6 convolution to 3×3 convolution. Referring to the ELAN design concept, the C3 module was replaced with the C2f module, and the depth of the module was adjusted from 3, 6, 9, 3 of the C3 module to 3, 6, 6, 3 of C2f.

[0061] 2. The detection head uses anchor-free+Decoupled-head, which is replaced by a decoupled head structure to separate the classification task and the regression task. This decoupled head no longer has the previous objectness (target foreground / background) branch, but is directly decoupled into two paths, and its regression branch uses the integral form representation proposed in Distribution Focal Loss. At the same time, Anchor-Based is replaced by Anchor-Free.

[0062] 3. Loss (or loss function) calculation includes 2 branches: classification and regression branches, without the previous objectness branch. The sigmoid function is used to calculate the probability of each category, and BCE Loss (Binary Cross Entropy) is used to calculate the global category loss. VFL Loss + CIOU Loss is used as the regression loss. The 3 Losses can be weighted with a certain weight ratio of 7.5:0.5:1.5. The main purpose is to balance the importance of different loss functions in model training. After comparison, this ratio is more conducive to convergence accuracy during the optimization process of the convolutional neural network model.

[0063] S12-2. The convolutional neural network model with the initial parameter configuration is trained according to the training set, and the training ratio of the misclassified image samples is adjusted according to the number of training times. A fixed position can be selected in the image samples of the training set to mark the scrap steel bars of the image samples. The marking tool uses labelImg, and the labels are divided into two types: normal state and shear fault state. After the labeling is completed, the image samples of the training set are input into the convolutional neural network model for training.

[0064] When the number of training times is less than the first threshold, the number of input samples for recognition errors is increased; as the number of training times increases, when the number of training times is greater than the second threshold, the number of input samples for recognition errors is reduced. The above scheme is implemented based on the Task-Aligned Assigner. The box matching strategy is changed from static matching to the matching method of the task-aligned assigner. The distribution ratio of positive and negative samples is dynamically adjusted during the training process. In the early stage of training, the convolutional neural network model may find it difficult to distinguish the shearing fault state, so it is necessary to pay more attention to those samples that are easily misclassified and give them greater weights. As the training progresses, the convolutional neural network model gradually becomes more powerful and can better detect anomalies. Therefore, the weight of difficult samples should be gradually reduced, while the weight of easy-to-classify samples should be increased.

[0065] The training strategy adds the last 10 epochs and turns off the Mosaic operation, which can effectively improve the training accuracy. The training parameters used include batch size of 16 and optimizer of SGD. The total number of training epochs is increased from 300 to 500.

[0066] S12-3, test the trained convolutional neural network model according to the test set, and determine that the convolutional neural network model has completed training when the test result meets the preset target. The preset target can be freely set based on actual needs. For example, the recognition accuracy of the shear fault state is set to be greater than 95%. When the test result meets the preset target, it means that the convolutional neural network model can accurately identify the shear fault state, and determine that the convolutional neural network model has completed training. After the convolutional neural network model completes training, enter step S13.

[0067] S13. When the welding shear performs a shearing task, the current image of the shear outlet is input into the trained convolutional neural network model, wherein the shearing task is to perform head and tail shearing of the strip to be welded.

[0068] Specifically, whether the welding shear performs the shearing task can be determined based on the movement instruction or stroke of the welding frame. The welding frame is used to pull the strip steel to the target position for implementing the shearing task. Therefore, a travel switch can be installed on the welding machine. After the movement of the welding frame triggers the travel switch, it is determined that the welding shear is in the state of performing the shearing task. The current image of the shearing outlet is input into the convolutional neural network model to determine whether a shearing failure occurs during the execution of the shearing task.

[0069] It should be noted that the identification of shearing faults through the convolutional neural network model requires a large amount of computing resources, and the data processing terminal running the convolutional neural network model has a good hardware configuration to implement real-time processing. The embodiment of the present invention implements image processing when the welding shear performs a shearing task, which can save a lot of computing resources and effectively reduce the operating energy consumption of the data processing terminal, making the solution have good applicability.

[0070] In practical applications, installing hardware such as a travel switch on a welding shear requires rewiring, and there is a risk of hardware damage, resulting in insufficient reliability of shear fault detection. Based on this, in a specific implementation, step S13 includes sub-steps S13-1 to S13-2, which are specifically described as follows:

[0071] S13-1, judging whether the welding machine shears have executed the shearing task according to the current image of the shearing outlet. Whether the shearing task is executed can be determined based on the change characteristics of the welding rack and other components in the current image, for example, whether the welding rack moves to the target position based on adjacent image frames. When the comparison results of adjacent image frames can determine that the welding rack moves to the target position, it means that the welding machine shears are executing the shearing task; otherwise, it is determined that the welding machine shears have not executed the shearing task.

[0072] Exemplarily, the determination of whether the welding shear performs the shearing task is as follows:

[0073] The first step is to extract the target area in the current image of the shear outlet according to the preset pixel frame. The hardware equipment can be installed in the operating room on the side of the welding machine shear, and the display is facing the outside of the operating room for easier observation by the welding machine operator. To set the target area based on the observation results, please continue to refer to Figure 3 , area A in the figure is the target area.

[0074] The second step is to perform grayscale processing and binarization processing on the target area in turn to obtain the number of white pixels in the target area. Grayscale processing is to convert a color image into a grayscale image, and binarization processing is to convert a grayscale image into a black and white image. A black and white image only has black pixels and white pixels. By measuring the white pixels of a black and white image, the number of white pixels can be obtained.

[0075] In the third step, when the number of white pixels is greater than the preset number threshold, it means that the welding rack moves to the driving side and stops, and it is determined that the welding shear performs the shearing task. The number threshold can be set freely. When the detection is performed based on the movement of the welding rack, the number threshold is set to 4200. When it is determined based on the number threshold that the welding shear performs the shearing task, step S13-2 is entered.

[0076] S13-2, if yes, then all current images in the preset time period are input into the trained convolutional neural network model. The preset time period is derived based on the time when the welding shear performs the shearing task, and represents the overall duration of the shearing fault detection. Still taking the welding frame as the target feature for detection as an example, after entering the detection, the range of the continuous detection time represented by the preset time period is set to 30 seconds. Through on-site observation and testing, the welding shear will only shear after the detection device reaches the signal point, and it only takes 2-3 seconds from the completion of shearing to the scrap steel bar falling into the waste hopper. Only in this time period may there be an abnormal situation of scrap steel bar removal. Considering that there may be errors in the calculation of the detection start signal point and the on-site operation in time, the time is set to 30 seconds. After the detection signal is determined, the current time point time_now is recorded, and the time point time_end after 30 seconds is calculated and expressed. When the time point time_ss in the detection process is greater than time_end, this detection is jumped out.

[0077] S14. According to the output result of the trained convolutional neural network model, it is determined whether the welding shear is in a normal working state or a shearing fault state.

[0078] Specifically, the output result can directly indicate whether there is a shearing fault state. When it is indicated as a shearing fault state, it is directly determined based on the output result whether the state of the welding shear is a normal working state or a shearing fault state.

[0079] Of course, the output result can also be used to characterize the number of scrap steel bars discharged by the shearing task. When the number of scrap discharged from the shearing outlet in the preset period is 2, it means that all the scraps have been discharged, and it is determined that the welding shear is in a normal working state; on the contrary, when the number of scrap discharged from the shearing outlet in the preset period is less than 2, it means that the scraps may not have been discharged, or partially discharged, and it is determined that the welding shear is in a shearing failure state.

[0080] The method of representing the shear fault state can be selected according to actual needs, such as alarming through an alarm or displaying a pop-up window on a display. Figure 4 , Figure 4 This is a schematic diagram of the pop-up window display under abnormal working conditions. The pop-up window display is performed according to the waste discharge quantity represented by the output result, and the display time of the pop-up window display is kept not less than the preset time threshold. The output results with shearing failures can be prompted by color filling to remind the on-site staff.

[0081] The steps of the test are as follows:

[0082] 1) Set the measurement result of white pixels in the target area as ROI_sum, detect signal points that meet ROI_sum>4200, and enter the shearing fault detection task.

[0083] 2) The start time of the current test, time_now, and the time point 30 seconds after the start of the test, time_end, are recorded.

[0084] Detection start time time_now = str(datetime.now()).split('.')[0]

[0085] Detection start time + 30s time_end = (datetime.strptime(time_now,'%Y-%m-%d%H:%M:%S')+timedelta(seconds=30)).strftime('%Y-%m-%d%H:%M:%S') / / The entire detection lasts for 30s. During the 30s between the start of the detection and the end of the detection, the scrap steel bars will be discharged.

[0086] 3) Detect each frame of the image, display and save the results.

[0087] 4) Record the current time point time_ss = str(datetime.now()).split('.')[0]

[0088] 5) Determine whether time_ss is greater than time_end. If so, the detection process stops and the loop is exited.

[0089] 6) If there are two small strips of scrap discharged within 30 seconds, the welding shear is in a normal state. If there are no small strips or only one small strip is discharged, it is determined that the welding shear is abnormal, that is, the shearing failure state. An alarm is required when an abnormal state of scrap steel strips is detected. Set flag = 0. When there is a normal small strip elimination state within the specified time, flag is 1, which is a normal state. The normal state is printed on the interface and this test is jumped out. Otherwise, the state is abnormal, the test results are saved and a pop-up window prompt is made. If only one scrap steel strip is detected, such as Figure 4 The pop-up window prompt lasts for 3 seconds.

[0090] The following embodiment of the present invention will generally describe the detection method of strip shearing failure. Figure 5 , Figure 5 This is a logical diagram of the detection method, including:

[0091] S501, obtaining a monitoring video captured by a fixed camera arranged on one side of the welding machine, and obtaining an image sample set based on an image frame of the monitoring video.

[0092] S502, image screening and annotation, obtaining a training set and a test set for training and testing a convolutional neural network model through an image sample set, and screening and annotating the training set and the test set.

[0093] S503: Perform cluster analysis on the training set and the test set.

[0094] S504: Train and test the convolutional neural network model according to the training set and the test set to obtain a convolutional neural network model that can accurately identify shear faults.

[0095] S505: Acquire a detection start signal based on the current image of the welding machine shear.

[0096] S506, determine whether the number of white pixels in the current image is greater than the number threshold 4200; when the number of white pixels is greater than the number threshold 4200, enter step S507, when the number of white pixels is less than or equal to the number threshold 4200, return to step S505.

[0097] S507: Output a prediction result based on the recognition result of the current image by the convolutional neural network model.

[0098] Based on the same technical concept as the detection method, the embodiment of the present invention also provides a detection device for strip shearing failure, see Figure 6 , Figure 6 It is a structural diagram of the detection device. The detection device includes:

[0099] An acquisition module 601 is used to acquire an image sample set of a welding shear, wherein the image sample set is an image set of a shear outlet of the welding shear when the welding shear is in a normal working state and a shearing fault state;

[0100] A training module 602 is used to train a preset convolutional neural network model according to the image sample set until the convolutional neural network model accurately identifies image samples in the image sample set that have a shearing fault state;

[0101] An input module 603 is used to input the current image of the shear outlet into the trained convolutional neural network model when the welding shear performs a shearing task, wherein the shearing task is to perform a head and tail shearing task on the strip to be welded;

[0102] The determination module 604 is used to determine whether the welding shear is in a normal working state or a shearing fault state according to the output result of the trained convolutional neural network model.

[0103] In an optional embodiment, the acquisition module includes:

[0104] The acquisition submodule is used to acquire a plurality of first images of different sizes when the welding machine shear is in a normal working state, and a plurality of second images of different sizes when the shear is in a shear fault state;

[0105] A preprocessing submodule, used for performing random preprocessing on the plurality of first images and the plurality of second images respectively, so as to increase the number of first images and second images, wherein the preprocessing includes rotation, cropping, scaling and flipping;

[0106] The first acquisition submodule is used to obtain an image sample set according to a set consisting of a plurality of first images and a plurality of second images after preprocessing.

[0107] In an optional embodiment, the training module includes:

[0108] A configuration submodule is used to divide the image sample set into a training set and a test set according to a preset ratio, and configure initial parameters for the convolutional neural network model;

[0109] An adjustment submodule is used to train the convolutional neural network model that has completed the initial parameter configuration according to the training set, and adjust the training ratio of the misclassified image samples according to the number of training times;

[0110] The first determination submodule is used to test the trained convolutional neural network model according to the test set, and determine that the convolutional neural network model has completed training when the test result meets the preset target.

[0111] In an optional embodiment, the input module includes:

[0112] The first judgment submodule is used to judge whether the welding shear performs the shearing task according to the current image of the shearing outlet;

[0113] The input submodule is used to input all current images of a preset time period into the trained convolutional neural network model when judging that the welding shear performs a shearing task based on the current image of the shearing outlet.

[0114] In an optional embodiment, the first judgment submodule includes:

[0115] An extraction unit, used to extract a target area in the current image of the shearing outlet according to a preset pixel frame;

[0116] An obtaining unit is used to perform grayscale processing and binarization processing on the target area in sequence to obtain the number of white pixels in the target area;

[0117] The determination unit is used to determine that the welding shear performs the shearing task when the number of white pixel points is greater than a preset number threshold.

[0118] In an optional embodiment, the determining module includes:

[0119] The second judgment submodule is used to judge whether the output result represents that the number of waste materials discharged from the shearing outlet in a preset time period is less than 2;

[0120] The second determination submodule is used to determine that the welding machine shear is in a normal working state when the waste discharge quantity is 2;

[0121] The third determination submodule is used to determine that the welding shear is in a shearing fault state when the waste discharge quantity is less than 2.

[0122] In an optional embodiment, the determining module further includes:

[0123] The display submodule is used to display a pop-up window according to the waste discharge quantity represented by the output result, and keep the display time of the pop-up window display not less than a preset time threshold.

[0124] Based on the same technical concept as the detection method, an embodiment of the present invention also provides an electronic device, including a processor and a memory, wherein the memory is coupled to the processor and stores instructions, and when the instructions are executed by the processor, the electronic device executes the steps of any one of the detection methods.

[0125] Based on the same technical concept as the detection method, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the detection methods are implemented.

[0126] The technical solution provided in the embodiments of the present invention has at least the following technical effects or advantages:

[0127] 1. Obtain an image sample set of the welding shears; train a preset convolutional neural network model based on the image sample set until the convolutional neural network model accurately identifies image samples in the image sample set that are in a shearing fault state; since the convolutional neural network model can accurately identify shearing faults, when the welding shears perform a shearing task, the current image of the shearing outlet is input into the trained convolutional neural network model; based on the output result of the trained convolutional neural network model, determine whether the welding shears is in a normal working state or a shearing fault state. This technical solution implements intelligent detection of shearing faults through artificial intelligence, has high detection efficiency, and at the same time reduces the error detection of manual naked eye observation in the existing technology, thereby taking into account the efficiency and accuracy of shearing fault detection.

[0128] 2. The detection scheme based on the embodiment of the present invention can reduce the one-time investment and construction costs of equipment. The industrial detection and control method based on deep learning can be widely used in strip steel production lines. The purchase of a complete set of intelligent detection systems ranges from 30,000 to 50,000 yuan. The above measures solve the production line shutdown failure caused by production personnel checking the abnormal derivatives of the discharge of small strips of waste, improve the quality and output of the production line, and reduce the failure of spare equipment by 1 time / month, an average of 1 hour / month. Increase production capacity by 180 tons / month and increase revenue by 18,000 / month. The total annual revenue is increased by 18,000 / month*12 months=216,000 / year.

[0129] 3. Improve production efficiency and reduce labor intensity. It is convenient for welding machine operators to observe the equipment status, ensure the safety of operator inspection, and reduce abnormal shutdown of production lines.

[0130] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0131] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (modules, systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0132] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0133] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0134] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0135] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for detecting strip shearing failure, characterized in that: The method comprises: Acquire an image sample set of the welding shear, wherein the image sample set is a set of images of the shearing outlet of the welding shear when the welding shear is in a normal working state and a shearing fault state; Training a preset convolutional neural network model according to the image sample set until the convolutional neural network model accurately identifies image samples in the image sample set that exist in the shearing fault state; When the welding shear performs a shearing task, the current image of the shear outlet is input into the trained convolutional neural network model, wherein the shearing task is to shear the head and tail of the strip to be welded; According to the output result of the trained convolutional neural network model, it is determined whether the welding shear is in the normal working state or the shear fault state.

2. The method for detecting strip shearing failure according to claim 1, characterized in that: When the welding shear performs a shearing task, the current image of the shearing outlet is input into a trained convolutional neural network model, including: Determining whether the welding shear performs the shearing task according to the current image of the shearing outlet; If so, all current images in the preset time period are input into the trained convolutional neural network model.

3. The method for detecting strip shearing failure according to claim 2, characterized in that: The step of judging whether the welding shear performs the shearing task according to the current image of the shearing outlet comprises: Extracting a target area in the current image of the shearing outlet according to a preset pixel frame; Performing grayscale processing and binarization processing on the target area in sequence to obtain the number of white pixels in the target area; When the number of white pixel points is greater than a preset number threshold, it is determined that the welding shear performs the shearing task.

4. The method for detecting strip shearing failure according to claim 1, characterized in that: Determining that the welding shear is in the normal working state or the shear fault state according to the output result of the trained convolutional neural network model includes: Determining whether the output result indicates that the amount of waste discharged from the shear outlet in a preset time period is less than 2; If not, it is determined that the welding shear is in the normal working state; If so, it is determined that the welding shear is in the shear fault state.

5. The method for detecting strip shearing failure according to claim 4, characterized in that: After determining that the welding shear is in the shear fault state, the method further includes: A pop-up window is displayed according to the waste discharge quantity represented by the output result, and the display time of the pop-up window is kept not less than a preset time threshold.

6. The method for detecting strip shearing failure according to claim 1, characterized in that: The step of obtaining a set of image samples of a welding machine shear comprises: Collecting a plurality of first images of different sizes when the welding shear is in a normal working state, and a plurality of second images of different sizes when the shear is in a fault state; Performing random preprocessing on the plurality of first images and the plurality of second images respectively to increase the number of the first images and the second images, wherein the preprocessing includes rotation, cropping, scaling and flipping; The image sample set is obtained according to a set consisting of the plurality of first images and the plurality of second images after preprocessing.

7. The method for detecting strip shearing failure according to claim 1, characterized in that: The method of training a preset convolutional neural network model according to the image sample set until the convolutional neural network model accurately identifies image samples in the image sample set that are in the shearing fault state comprises: Dividing the image sample set into a training set and a test set according to a preset ratio, and configuring initial parameters for the convolutional neural network model; The convolutional neural network model with initial parameter configuration is trained according to the training set, and the training ratio of misclassified image samples is adjusted according to the number of training times; The trained convolutional neural network model is tested according to the test set, and when the test result meets the preset target, it is determined that the convolutional neural network model has completed training.

8. A strip shearing fault detection device, characterized in that: The device comprises: An acquisition module, used for acquiring an image sample set of a welding shear, wherein the image sample set is an image set of a shearing outlet of the welding shear when the welding shear is in a normal working state and a shearing fault state; A training module, used for training a preset convolutional neural network model according to the image sample set, until the convolutional neural network model accurately identifies image samples in the image sample set that are in the shearing fault state; An input module, used for inputting the current image of the shear outlet into the trained convolutional neural network model when the welding shear performs a shearing task, wherein the shearing task is a task of shearing the head and tail of the strip to be welded; A determination module is used to determine whether the welding shear is in the normal working state or the shear fault state according to the output result of the trained convolutional neural network model.

9. An electronic device, characterized in that: The electronic device comprises a processor and a memory, wherein the memory is coupled to the processor and stores instructions, and when the instructions are executed by the processor, the electronic device executes the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

Images