Abnormality detection method and device for sintering machine trolley

Abstract

The present application relates to the technical field of sintering machine trolley status monitoring, and discloses a method and device for detecting abnormalities in sintering machine trolleys. The method comprises: obtaining the real-time status of the trolley; determining whether the trolley has any abnormality based on the real-time status; if the trolley has no abnormality, continuing to obtain the real-time status of the trolley; if the trolley has any abnormality, issuing a corresponding abnormality prompt based on the abnormality of the real-time status. The present application breaks away from the traditional detection method of relying on posts or inspectors to conduct trolley inspections and observations, and promptly detects trolley faults and issues alarms, thereby improving the sintering production efficiency and the quality of sintered minerals.

Description

Technical Field

[0001] The present application relates to the technical field of sintering machine trolley status monitoring, and in particular, to a sintering machine trolley abnormality detection method and device. Background Art

[0002] Since there is no effective detection method applicable to sintering trolleys at present, the management of sintering machine trolleys in the industry is relatively extensive, and basically relies on post or mechanical inspectors to spend a lot of labor to conduct inspections and observations. Even if a lot of labor is spent, it is difficult to timely discover individual abnormal conditions, such as loose trolley wheels, deviation, offset of railings, broken and broken grate bars, etc. Even if workers and inspectors conduct uninterrupted inspections, some faults are difficult to find. Once these faults develop further, it is easy to cause emergency shutdown of equipment, seriously affecting the smooth progress of production and even threatening the personal safety of the post. Therefore, the proposal and implementation of the present invention will be of great significance to reducing the number of on-site operators and labor intensity, guiding equipment maintenance personnel to carry out targeted inspections and maintenance, avoiding safety accidents, improving personnel safety factors, timely discovering trolley faults and alarming them, and improving sintering production operation rate and sintered mineral quality. Summary of the Invention

[0003] The purpose of this application is to provide a method and device for detecting abnormalities in a sintering machine trolley, which automatically monitors the trolley in real time to see if there are any abnormal conditions, and forms an abnormal fault detection and diagnosis system that is effective for sintering trolleys, so as to detect abnormal conditions in time and issue abnormal prompts.

[0004] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by practice of the present application.

[0005] According to one aspect of an embodiment of the present application, a method for detecting an abnormality of a sintering machine trolley is provided, the detection method comprising: obtaining the real-time status of the trolley; determining whether the trolley has an abnormality based on the real-time status; if the trolley has no abnormality, continuing to obtain the real-time status of the trolley; if the trolley has an abnormality, issuing a corresponding abnormality prompt based on the abnormality of the real-time status.

[0006] In some embodiments, the abnormal situation includes a stuck wheel state. In the method of issuing a corresponding abnormal prompt based on the abnormal situation of the real-time state if the trolley has an abnormal situation, the method includes: acquiring a wheel image of the trolley once at a set interval; calculating the moving speed of the wheel based on two images acquired at adjacent times; acquiring the actual speed of the trolley; comparing the actual speed with the moving speed, and if the moving speed is less than the actual speed, it is determined to be a stuck wheel state, and a stuck wheel state prompt is issued.

[0007] In some embodiments, the abnormal condition also includes a wheel axle-steering state, and a sensor is set between the left wheel and the right wheel of the trolley. If the trolley has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition of the real-time state. The method also includes: obtaining a first actual distance between the left wheel and the sensor, and obtaining a first ideal distance between the left wheel and the sensor; obtaining a second actual distance between the right wheel and the sensor, and obtaining a second ideal distance between the right wheel and the sensor; calculating a first difference between the first actual distance and the first ideal distance; calculating a second difference between the second actual distance and the second ideal distance; if both the first difference and the second difference are not 0, and the sum of the first difference and the second difference is 0, it is determined to be a wheel axle-steering state, and a wheel axle-steering state prompt is issued.

[0008] In some embodiments, the abnormal condition also includes a loose wheel state. In the method of issuing a corresponding abnormal prompt based on the abnormal condition of the real-time state if the trolley has an abnormal condition, the method also includes: if the first preset threshold value < the absolute value of the first difference < the second preset threshold value, and the second difference value is 0, it is determined that the left wheel is in a loose wheel state, and a wheel loose state prompt is issued; if the first preset threshold value < the absolute value of the second difference < the second preset threshold value, and the first difference value is 0, it is determined that the right wheel is in a loose wheel state, and a wheel loose state prompt is issued.

[0009] In some embodiments, the abnormal situation also includes a wheel off state. In the method of issuing a corresponding abnormal prompt based on the abnormal situation of the real-time state if the trolley has an abnormal situation, the method also includes: if the second preset threshold value is less than the first difference absolute value, it is determined that the left wheel is in a wheel off state, and a wheel off state prompt is issued; if the second preset threshold value is less than the second difference absolute value, it is determined that the right wheel is in a wheel off state, and a wheel off state prompt is issued.

[0010] In some embodiments, the abnormal situation also includes a wheel-off state. In the method of issuing a corresponding abnormal prompt based on the abnormal situation of the real-time state if the trolley has an abnormal situation, the method also includes: obtaining a wheel image of the trolley and performing image recognition on the wheel image; if the wheel cannot be identified in the wheel image, it is determined to be a wheel-off state, and a wheel-off state prompt is issued.

[0011] In some embodiments, before acquiring the wheel image of the trolley and performing image recognition on the wheel image, the method further includes: acquiring multiple wheel images of the trolley with wheels present, and generating a first training set; acquiring multiple wheel images of the trolley with wheels not present, and generating a second training set; and identifying whether a wheel exists in the wheel image based on the first training set and the second training set.

[0012] In some embodiments, before obtaining the real-time status of the trolley, the method further includes: obtaining the identity card of each trolley and the corresponding position of each identity card; recording the order of the identity cards according to the corresponding positions of the identity cards for locating each trolley.

[0013] In some embodiments, when determining whether the trolley has an abnormality based on the real-time status, if the trolley has no abnormality, continuing to obtain the real-time status of the trolley, and if the trolley has an abnormality, issuing a corresponding abnormal prompt based on the abnormality of the real-time status, the method also includes: obtaining the abnormal identity identification plate of the trolley with the abnormality; querying the corresponding position of the abnormal identity identification plate based on the abnormal identity identification plate, and searching for the trolley for maintenance based on the corresponding position of the abnormal identity identification plate.

[0014] According to one aspect of an embodiment of the present application, an abnormality detection device for a sintering machine trolley is provided, and the detection device includes: a status detection module, which is used to obtain the real-time status of the trolley; an abnormality prompt module, which is used to determine whether the trolley has an abnormality based on the real-time status; if the trolley has no abnormality, the real-time status of the trolley is continued to be obtained; if the trolley has an abnormality, a corresponding abnormality prompt is issued according to the abnormality of the real-time status.

[0015] Compared with the existing technology, the technical solution of this application has the following significant benefits: it automatically monitors the trolley in real time for abnormal conditions, forming an effective abnormal fault detection and diagnosis system suitable for sintering trolleys, promptly discovering abnormal conditions, issuing abnormality prompts, and carrying out timely maintenance. This application breaks away from the traditional detection method of relying on post or inspectors to conduct trolley inspections and observations, and promptly detects trolley faults and issues alarms, thereby improving the sintering production efficiency and the quality of sintered ore.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other features and advantages of the present application will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings.

[0018] Figure 1 shows a flow chart according to an embodiment of the present application;

[0019] Figure 2 A schematic diagram of an abnormality detection device for a sintering machine trolley according to one embodiment of the present application is shown. DETAILED DESCRIPTION

[0020] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0021] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner.In the following description, many specific details are provided so as to provide a full understanding of the embodiments of the present application. However, it will be appreciated by those skilled in the art that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps etc. can be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0022] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0023] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, while others may be combined or partially combined. Therefore, the actual execution order may vary depending on the actual situation.

[0024] In order to make those skilled in the art better understand this application, Figure 1 The details of this application are described in detail.

[0025] According to some embodiments, the present application provides a method for detecting an abnormality of a sintering machine trolley, the detection method comprising:

[0026] Step 101, obtaining the real-time status of the machine vehicle;

[0027] Step 102: determine whether the trolley has any abnormality according to the real-time status. If the trolley has no abnormality, continue to obtain the real-time status of the trolley. If the trolley has an abnormality, issue a corresponding abnormality prompt according to the abnormality of the real-time status.

[0028] Based on the above embodiment, in step 101, the real-time status includes the wheel status and the trolley operation status.

[0029] In step 102, abnormal conditions include wheel jamming, wheel axle deflection, and trolley deviation. Each condition is acquired using a different method. The system acquires the real-time status of each condition in real time, compares it with the corresponding standard status, and ultimately determines whether it is abnormal. If so, an abnormality alert is issued to facilitate timely maintenance.

[0030] According to some embodiments, the abnormal condition includes a stuck wheel state. In step 102, if the trolley has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition in the real-time state. The method includes:

[0031] The wheel image of the machine vehicle is acquired once at a set interval;

[0032] The moving speed of the wheel is calculated based on two images acquired adjacently in time;

[0033] Get the actual speed of the machine vehicle;

[0034] The actual speed is compared with the moving speed. If the moving speed is less than the actual speed, it is determined to be a stuck wheel state, and a stuck wheel state prompt is issued.

[0035] Based on the above embodiment, the determination of wheel stuck is mainly based on obtaining wheel speed based on machine vision and image processing technology, including image acquisition, preprocessing, feature extraction and feature matching. The image acquisition system acquires a continuous frame sequence; preprocessing mainly corrects image distortion and eliminates the influence of lighting changes and shadows; wheel area positioning extracts the wheel rotation area and removes interference from the surrounding background; feature matching is a key step in speed measurement. By matching the markers of two adjacent frames and calculating the size of the angle based on the perspective relationship, the moving speed of the wheel is calculated based on the angle. By comparing the moving speed with the actual speed of the trolley, it can be determined whether the wheel is stuck. In the stuck state, the moving speed of the wheel is less than the actual speed of the trolley. If the moving speed of the wheel is greater than the actual speed of the trolley, it means that the wheel is slipping.

[0036] Wheel image acquisition is the first step in the operation of this system. The wheel spoke image is extracted from the acquired image for the measurement of wheel speed. Wheel image acquisition includes three steps: camera installation, parameter adjustment, and image acquisition and online transmission.

[0037] Image preprocessing, in image processing, refers to the processing performed on the input image before feature extraction, segmentation, and matching. Wheel image preprocessing primarily involves correcting camera distortion and addressing lighting effects. It also includes basic preprocessing steps such as normalization, smoothing, and enhancement.

[0038] Image distortion correction is achieved by calling OpenCV library functions. The two sides of the image before correction are severely deformed and distorted. By comparing the corrected image with the monitored area before correction, a more ideal correction effect can be achieved while ensuring no color distortion.

[0039] Distortion correction principle:

[0040] For radial distortion correction, the first three terms of Taylor expansion are generally used to fit the radial distortion.

[0041] x corrected =x(1+k1r 2 +k2r 4 +k3r 6 )

[0042] y corrected =y(1+k1r 2 +k2r 4 +k3r 6 ).

[0043] Tangential distortion correction: Use the following formula to fit the tangential distortion.

[0044] x corrected =x+[2p1xy+p2(r 2 +2x 2 )]

[0045] y corrected =y+[2p1xy+p2(r 2 +2x 2 )].

[0046] To eliminate distortion, five distortion parameters are needed: k1, k2, p1, p2, and k3.

[0047] You also need to know the internal and external parameters of the camera. The internal parameters include information such as focal length (f x ,f y ), optical center (c x ,cy ), etc. This is also called the camera matrix. It depends entirely on the camera itself and only needs to be calculated once, so it is known and used from now on. It can be represented by the following 3×3 matrix:

[0048]

[0049] The extrinsic parameters correspond to the rotation and translation vectors, which transform the coordinates of the 3D point into the coordinate system.

[0050] Distortion correction steps:

[0051] (1) Take a distorted image of the checkerboard calibration plate. The checkerboard can provide the position of a point in 3D space in the 2D image.

[0052] (2) Given the number of inner corner points of the chessboard in the horizontal and vertical directions, identify the position of each inner corner point.

[0053] (3) Using the OpenCV library function, the camera intrinsic parameter matrix and distortion parameters are calculated using the positions of each inner corner point.

[0054] (4) Using the camera intrinsic parameter matrix, distortion parameters and image size information, calculate the mapping equation from the distorted image to the undistorted image.

[0055] (5) Use the remapping equation to convert the distorted image into an undistorted image.

[0056] Image contrast processing

[0057] This invention primarily employs three image enhancement methods: histogram equalization, high-lift filtering, and bilateral filtering. Contrast-limited local histogram equalization is effective for uneven illumination and effectively enhances image contrast, but it also increases noise. High-lift filtering effectively enhances image edge features but also increases noise. Bilateral filtering effectively suppresses noise without blurring edges.

[0058] Histogram equalization principle:

[0059] Let r k ,k=0,1,2,...,L-1, represents the grayscale of an L-level grayscale digital image f(x,y). The histogram is defined as:

[0060]

[0061] Where n k The grayscale in f is r k The number of pixels, M and N are the number of rows and columns of the image, respectively. For all values ​​of k, p(r k ) always sums to 1. In fact, p(r k) is an estimate of the probability of the gray levels appearing in the image.

[0062] By expanding the histogram of a low-contrast image with a relatively concentrated distribution of histogram containers to the entire grayscale and distributing it evenly, a high-contrast image can be obtained. This is the essence of histogram equalization.

[0063] The principle of high boost filtering:

[0064] High-lift filtering consists of three steps: blurring the original image, subtracting the blurred image from the original image, and adding the template to the original image.

[0065] There are many algorithms for blurring images. The simplest method is to use mean smoothing. Mean smoothing replaces the central element with the average value of all pixels in the area covered by the convolution box. The following is a 3×3 normalized convolution box:

[0066]

[0067] If Gaussian smoothing is used, we only need to replace the convolution kernel with a Gaussian kernel, which is sampled from a Gaussian bell curve. Here is a 3×3 Gaussian kernel with σ=1:

[0068]

[0069] The same algorithms that blur an image can also sharpen it. The essence of a smoothing algorithm is to separate the low-frequency components (smooth grayscale transitions) from the original image. Subtracting the low-frequency components from the original image yields the high-frequency components (edge ​​details and noise). If the high-frequency components are added to the original image with a certain weight, the resulting image has enhanced edge details.

[0070] Bilateral filtering works by using both spatial Gaussian weighting and grayscale similarity Gaussian weighting. The spatial Gaussian function ensures that only pixels in the neighborhood contribute to the center point, while the grayscale similarity Gaussian function ensures that only pixels with grayscale values ​​close to the center pixel are used for the blurring operation. This method ensures that boundaries are not blurred, as grayscale values ​​vary significantly at boundaries.

[0071] Image enhancement:

[0072] (1) Use contrast-limited local histogram equalization to eliminate the effects of uneven illumination and improve the contrast of the image;

[0073] (2) Then use high-lift filtering to enhance image details;

[0074] (3) Use bilateral filtering to remove noise;

[0075] (4) Use histogram equalization to improve the contrast of the image.

[0076] Image feature extraction, wheel edge extraction.

[0077] Edge detection is performed based on the Canny algorithm. Canny edge detection is currently the most advanced edge detection algorithm. It is based on three goals: low error rate, edge points should be well located, and the located edges must be as close to the true edges as possible.

[0078] Edge detection steps:

[0079] (1) Smoothing and denoising. Edge detection is easily affected by noise. Gaussian filtering has good smoothing performance and can retain image details to the greatest extent while removing noise.

[0080] (2) Detect all possible edges in the image. The most obvious feature of an edge is its gradient, so the Sobel operator can be used to calculate the gradients in the horizontal and vertical directions (Gx and Gy). Based on the two gradient images (Gx and Gy), the gradient and direction of the boundary are found using the following formula:

[0081]

[0082] (3) Non-maximum suppression. Each pixel is checked to see if its gradient is the largest among the surrounding points with the same gradient direction. This step finds points that are more likely to be edges through comparison.

[0083] (4) Hysteresis threshold. Use two thresholds, high and low, to identify edges. All edges above the high threshold are retained, and all edges below the low threshold are discarded. Edges between the high and low thresholds are considered candidate edges.

[0084] For matching area positioning, we use the Hough gradient method to optimize the standard Hough circle detection. Its principle is that the center of a circle is the intersection of all normal lines around the circle.

[0085] Regional positioning steps:

[0086] (1) Edge detection, such as Canny edge detection. Then, for each non-zero point in the edge image, consider its local gradient, that is, the gradient obtained by calculating the Sobel first-order derivative in the x and y directions using the Sobel function.

[0087] (2) Determine the candidate circle center. Through the gradient, we traverse each point in the accumulator from a minimum to a maximum value along the line represented by this slope, while recording the location of each non-zero pixel in the contour image.

[0088] (3) Determine the optimal radius. Select the best value from the minimum distance to the maximum radius as the radius of the circle.

[0089] (4) Determine the center of the circle.

[0090] Image feature matching, wheel rotation speed and trolley running speed determination.

[0091] The rotation angle is calculated by matching landmarks in two adjacent frames based on an affine relationship. This is crucial for speed detection. The quality of the speed detection algorithm directly determines the system's functionality. Low detection accuracy can lead to missed detections, directly impacting the safe operation of the vehicle. High false alarm rates can result in unnecessary labor. Speed ​​measurement is primarily based on ORB feature matching.

[0092] ORB feature matching speed measurement steps:

[0093] (1) Construct an image scale pyramid and use the Fast algorithm to detect corner points with scale-invariant features.

[0094] (2) Use machine learning and non-maximum suppression to screen the optimal key points.

[0095] (3) Use the first-order grayscale moment to calculate the direction of the key point.

[0096] (4) Use the improved Brief descriptor to describe the features of key points, including their orientation.

[0097] (5) Brute force matching of feature points of adjacent frames.

[0098] (6) Calculate the best-fit perspective transformation matrix from the starting frame to the target frame.

[0099] (7) Decompose the perspective transformation matrix and calculate the wheel rotation angle, horizontal displacement and scaling coefficient between adjacent frames.

[0100] (8) Use the scaling factor to verify the results.

[0101] (9) Based on the above measured wheel rotation angle, horizontal displacement and sampling period, the wheel rotation speed and horizontal movement speed are calculated.

[0102] The wheel speed may differ from the value calculated based on the rolling assumption due to the condition of the bearings, but the horizontal movement speed must be the same as the actual speed of the trolley. Therefore, the calculated movement speed result is verified with the actual speed reference value of the trolley.

[0103] According to some embodiments, the abnormal condition also includes a wheel axle misalignment state. In step 102, a sensor is provided between the left wheel and the right wheel of the trolley. If the trolley has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition of the real-time state. The method further includes:

[0104] Acquire a first actual distance between the left wheel and the sensor, and acquire a first ideal distance between the left wheel and the sensor;

[0105] obtaining a second actual distance between the right wheel and the sensor, and obtaining a second ideal distance between the right wheel and the sensor;

[0106] calculating a first difference between the first actual distance and the first ideal distance;

[0107] calculating a second difference between the second actual distance and the second ideal distance;

[0108] If both the first difference and the second difference are not 0, and the sum of the first difference and the second difference is 0, it is determined to be a wheel axle-contact state, and a wheel axle-contact state prompt is issued.

[0109] According to some embodiments, the abnormal condition further includes a loose wheel state. In step 102, if the trolley has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition in the real-time state. The method further includes:

[0110] If the first preset threshold value is less than the absolute value of the first difference value and less than the second preset threshold value, and the second difference value is 0, it is determined that the left wheel is in a loose state, and a loose wheel state prompt is issued;

[0111] If the first preset threshold value is less than the absolute value of the second difference value and is less than the second preset threshold value, and the first difference value is 0, it is determined that the right wheel is in a loose state, and a loose wheel state prompt is issued.

[0112] Based on the above embodiment, the first preset threshold is a value greater than or equal to 0, and both the first preset threshold and the second preset threshold can be set according to actual needs.

[0113] According to some embodiments, the abnormal condition also includes a wheel fall state. In step 102, if the trolley has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition of the real-time state. The method further includes:

[0114] If the second preset threshold is less than the absolute value of the first difference, it is determined that the left wheel is in a wheel-off state, and a wheel-off state prompt is issued;

[0115] If the second preset threshold value is less than the second absolute value of the difference, it is determined that the right wheel is in a wheel-off state, and a wheel-off state prompt is issued.

[0116] Furthermore, the abnormal situation also includes a wheel fall state. In step 102, if the vehicle has an abnormal situation, a corresponding abnormal prompt is issued according to the abnormal situation of the real-time state. The method further includes:

[0117] Acquire a wheel image of a trolley and perform image recognition on the wheel image;

[0118] If no wheel can be identified in the wheel image, it is determined to be in a wheel-off state, and a wheel-off state prompt is issued.

[0119] Based on the above embodiment, before acquiring the wheel image of the trolley and performing image recognition on the wheel image, the method further includes:

[0120] Acquire multiple wheel images of wheels on a vehicle and generate a first training set;

[0121] Acquire multiple wheel images of the vehicle without wheels and generate a second training set;

[0122] Whether a wheel exists in the wheel image is identified based on the first training set and the second training set.

[0123] Based on machine vision, with the help of wheel edge extraction and matching area positioning algorithm, if the image recognition cannot identify the outer contour of the wheel within the time range, it is determined that the wheel has fallen off.

[0124] According to some embodiments, before obtaining the real-time status of the trolley in step 101, the method further includes:

[0125] Obtain the identification plate of each vehicle and the corresponding position of each identification plate;

[0126] The sequence of the identification plates is recorded according to the corresponding positions of the identification plates, so as to locate each trolley.

[0127] Based on the above embodiment, in step 102, it is determined whether the machine vehicle has any abnormality according to the real-time status. If the machine vehicle has no abnormality, the real-time status of the machine vehicle is continuously acquired. If the machine vehicle has an abnormality, a corresponding abnormality prompt is issued according to the abnormality of the real-time status. The method further includes:

[0128] Obtain the abnormal identification plate of the machine vehicle with abnormal conditions;

[0129] The corresponding position of the abnormal identity identification plate is inquired according to the abnormal identity identification plate, and the machine vehicle is searched for maintenance according to the corresponding position of the abnormal identity identification plate.

[0130] Furthermore, after determining whether the trolley has an abnormality according to the real-time status, and if the trolley has no abnormality, continuing to obtain the real-time status of the trolley, and if the trolley has an abnormality, issuing a corresponding abnormality prompt according to the abnormality of the real-time status, the method further includes:

[0131] Obtain the abnormal identification plate of the machine vehicle with abnormal conditions;

[0132] The corresponding position of the abnormal identity identification plate is inquired according to the abnormal identity identification plate, and the machine vehicle is searched for maintenance according to the corresponding position of the abnormal identity identification plate.

[0133] Furthermore, abnormal conditions also include the state of the trolley running off track. A single-point displacement sensor is added to the lower railing of the trolley to illuminate the continuous distance data information of the lower trolley railing. Along the trolley's running direction, multiple consecutive trolley railing distance detection values ​​are read over a period of time. The polynomial fitting algorithm is used for fitting, and then the fitting results are used to calculate the continuous slope values. If the value continuously follows a slope direction, the proportion of the slope greater than the target value is judged. If the proportion exceeds the set maximum allowable range, it means that the trolley has deviated.

[0134] In essence, polynomial fitting is also a linear model, and its mathematical expression is:

[0135]

[0136] Where M is the highest degree of the polynomial, j represents the power of M, and x is the coefficient.

[0137] The number of samples is, for each sample, the corresponding output is, using the sum of the squares of the errors as the loss function, then the loss function can be expressed as:

[0138]

[0139] Adding 1 / 2 before the loss function here is just for the convenience of subsequent derivation and does not affect the final result.

[0140] From the above analysis, we can know that polynomial fitting is actually two processes: 1. Generate polynomial features on the original feature vector to obtain new features; 2. Perform linear regression on the new features.

[0141] Furthermore, abnormal conditions also include the offset state of the baffle. Cameras and modeling light sources are mainly placed on both sides of the trolley track to accurately measure the relative displacement of each baffle, and timely alarms are issued when serious displacement is found.

[0142] Single side railing deviation monitoring and alarm:

[0143] <1> The current detection value is compared with the target value and exceeds the specified range;

[0144] <2> The current detection value is compared with the average value of the guardrail itself over a period of time and exceeds the specified range;

[0145] Overall offset monitoring and alarm of the guardrail:

[0146] The offset of both side guardrails exceeds the range, one side is offset in the positive direction and the other side is offset in the negative direction.

[0147] Furthermore, abnormal conditions also include grate bar damage. The trolley railing is the equipment that carries the sintering raw materials. Under the action of high-temperature rapid cooling and heating, it often becomes damaged or falls off, resulting in an incomplete grate bed. There are also cases where sintering materials adhere to the grate bars, causing large-area gap blockages. Uneven grate bar gaps will affect air permeability and the quality of sintered ore. The falling of grate bars will cause the sintering materials to spill, and severe uneven air permeability will result in increased sintering return ore and scrap rate. By installing a high-definition industrial camera and lighting source on the machine head, continuous image capture of the grate bed and grate bars is performed, and the images are transmitted back to the image processing server in the control room for online analysis of grate plate damage.

[0148] Key control logic and algorithm implementation:

[0149] Step 1: Locate the image of the grate bars to be captured.

[0150] Step 2: If the grate image appears in the specified area, preprocess the image of the grate area.

[0151] Adaptive binarization is performed, and then the grate image is segmented according to the number of grate rows. The image binarization data is returned as the basis for grate bar falling off detection.

[0152] The binarization algorithm compares the input pixel value I with a value C and determines the output value based on the comparison result. In adaptive binarization, the comparison value C is different for each pixel. The comparison value C is calculated by subtracting the difference delta from the block range centered on the pixel.

[0153] There are two common methods for calculating C:

[0154] 1. Subtract the delta from the mean (using box filter, the performance will be very good)

[0155] 2. Gaussian distribution weighted sum minus the difference delta (using Gaussian filter GaussionBlur)

[0156] APIs provided by OpenCV:

[0157] void adaptiveThreshold(InputArray src,OutputArray dst,doublemaxValue,int adaptiveMethod,int thresholdType,int blockSize,double C)

[0158] Perform an opening operation on each row of grate images, then process the image through image morphological dilation operation, and find the contour of the connected area through contour detection;

[0159] Image morphological operations are a series of operations that manipulate images based on shape. Morphological operations produce an output image by applying structuring elements to an image. They alter the shape of an object, such as erosion, which makes it thinner, and dilation, which makes it fatter. A series of operations, including dilation and erosion, and their combined effects are called image morphological operations. The most basic morphological operations are erosion and dilation.

[0160] Expansion principle:

[0161] Dilation: Find local maximum;

[0162] a. Define a convolution kernel B. The kernel can be of any shape and size and has a single defined reference point - the anchor point. Usually, it is a square or disk with a reference point. The kernel can be called a template or mask.

[0163] b. Convolve kernel B with image A and calculate the maximum value of the pixels in the area covered by kernel B;

[0164] c. Assign this maximum value to the pixel specified by the reference point;

[0165] Therefore, the highlight areas in the image gradually grow.

[0166] Corrosion principle:

[0167] Erosion: local minimum (the opposite of dilation);

[0168] a. Define a convolution kernel B. The kernel can be of any shape and size and has a single defined reference point - the anchor point. Usually, it is a square or disk with a reference point. The kernel can be called a template or mask.

[0169] b. Convolve kernel B with image A and calculate the minimum value of the pixels in the area covered by kernel B;

[0170] c. Assign this minimum value to the pixel specified by the reference point;

[0171] Therefore, the highlight areas in the image gradually decrease.

[0172] Open Operation: The process of first corroding and then expanding;

[0173] Function: Eliminate small objects; separate objects at thin points; smooth larger boundaries without significantly changing their area;

[0174] Closing Openration: dilation followed by corrosion.

[0175] Function: Eliminate small black holes (black spots);

[0176] Traverse all contours, fit the minimum outer rectangle, obtain the width, height and angle of the connected area, and judge the grate status through the pre-set alarm length threshold and alarm width threshold of the defect area.

[0177] The first category is the vacancies formed after the grate bars are damaged, and the vacancies are judged; the threshold condition is to set a grayscale threshold, and the pixel points with grayscale less than the grayscale threshold are the defect points, and the area formed by the defect points is the defect area; and the alarm length threshold and alarm width threshold of the defect area are set.

[0178] The second category is the alarm for grate deflection and looseness. The grate image is first preprocessed to calculate the grate tilt area angle α, and the tilt angle α is used for judgment. The threshold condition is the tilt angle threshold. When the angle α of the grate unit deflected relative to the set initial direction exceeds the tilt angle threshold, an alarm is issued.

[0179] The third category is the grate bar tilting alarm. It traverses all contours, fits the minimum enclosing rectangle, and fits the minimum enclosing triangle to obtain the width, height, and angle of the connected area. It then determines whether the grate bars are tilted based on the pre-set alarm length threshold and alarm width threshold of the tilted area and the image morphology of the alarm area.

[0180] The following describes an embodiment of a device of the present application, which can be used to execute the abnormality detection method of the sintering machine trolley in the above-mentioned embodiment of the present application.

[0181] Figure 2 A simplified diagram of an abnormality detection device 200 for a sintering machine trolley in one embodiment of the present application is shown. The abnormality detection device 200 for a sintering machine trolley includes:

[0182] Status detection module 201, used to obtain the real-time status of the trolley;

[0183] The abnormality prompt module 202 is used to determine whether the machine vehicle has any abnormality according to the real-time status. If the machine vehicle has no abnormality, the real-time status of the machine vehicle is continuously acquired. If the machine vehicle has an abnormality, a corresponding abnormality prompt is issued according to the abnormality of the real-time status.

[0184] Based on the above embodiment, in the state detection module 201, the real-time state includes the wheel state and the locomotive vehicle operation state, etc.

[0185] In the abnormality notification module 202, abnormal conditions include wheel jamming, wheel axle deflection, and vehicle deviation. Each condition is acquired using a different method. The system acquires the real-time status of each condition in real time, compares it with the corresponding standard status, and ultimately determines whether it is abnormal. If so, an abnormality notification is issued to facilitate timely maintenance.

[0186] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of this application and include common knowledge or customary techniques in the art that are not disclosed herein.

[0187] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for detecting abnormality of a sintering machine trolley, characterized in that: The detection method comprises: Get the real-time status of the machine; Determine whether the machine vehicle has an abnormality according to the real-time status; if the machine vehicle has no abnormality, continue to obtain the real-time status of the machine vehicle; if the machine vehicle has an abnormality, issue a corresponding abnormality prompt according to the abnormality of the real-time status; The abnormal condition includes a wheel axle misalignment state, a sensor is provided between the left wheel and the right wheel of the machine vehicle, and if the machine vehicle has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition of the real-time state, the method further includes: Acquire a first actual distance between the left wheel and the sensor, and acquire a first ideal distance between the left wheel and the sensor; obtaining a second actual distance between the right wheel and the sensor, and obtaining a second ideal distance between the right wheel and the sensor; calculating a first difference between the first actual distance and the first ideal distance; calculating a second difference between the second actual distance and the second ideal distance; If both the first difference and the second difference are not 0, and the sum of the first difference and the second difference is 0, it is determined that the wheel is in a misaligned state, and a wheel misaligned state prompt is issued; The abnormal condition also includes a loose wheel state. If the vehicle has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition in the real-time state. The method further includes: If the first preset threshold value is less than the absolute value of the first difference value and less than the second preset threshold value, and the second difference value is 0, it is determined that the left wheel is in a loose state, and a loose wheel state prompt is issued; If the first preset threshold value is less than the absolute value of the second difference value and is less than the second preset threshold value, and the first difference value is 0, it is determined that the right wheel is in a loose state, and a loose wheel state prompt is issued.

2. The method according to claim 1, characterized in that The abnormal condition includes a stuck wheel state. If the machine vehicle has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition of the real-time state. The method includes: The wheel image of the machine vehicle is acquired once at a set interval; The moving speed of the wheel is calculated based on two images acquired adjacently in time; Get the actual speed of the machine vehicle; The actual speed is compared with the moving speed. If the moving speed is less than the actual speed, it is determined to be a stuck wheel state, and a stuck wheel state prompt is issued.

3. The method according to claim 1, characterized in that The abnormal situation also includes a wheel falling state. If the trolley has an abnormal situation, a corresponding abnormal prompt is issued according to the abnormal situation of the real-time state. The method further includes: If the second preset threshold is less than the absolute value of the first difference, it is determined that the left wheel is in a wheel-off state, and a wheel-off state prompt is issued; If the second preset threshold value is less than the second absolute value of the difference, it is determined that the right wheel is in a wheel-off state, and a wheel-off state prompt is issued.

4. The method according to claim 1, wherein The abnormal situation also includes a wheel falling state. If the trolley has an abnormal situation, a corresponding abnormal prompt is issued according to the abnormal situation of the real-time state. The method further includes: Acquire a wheel image of a trolley and perform image recognition on the wheel image; If no wheel can be identified in the wheel image, it is determined to be in a wheel-off state, and a wheel-off state prompt is issued.

5. The method according to claim 4, characterized in that Before acquiring the wheel image of the trolley and performing image recognition on the wheel image, the method further includes: Acquire multiple wheel images of wheels on a vehicle and generate a first training set; Acquire multiple wheel images of the vehicle without wheels and generate a second training set; Whether a wheel exists in the wheel image is identified based on the first training set and the second training set.

6. The method according to claim 1, characterized in that Before obtaining the real-time status of the vehicle, the method further includes: Obtain the identification plate of each vehicle and the corresponding position of each identification plate; The sequence of the identification plates is recorded according to the corresponding positions of the identification plates, so as to locate each trolley.

7. The method according to claim 6, characterized in that After determining whether the machine vehicle has an abnormality according to the real-time status, and if the machine vehicle has no abnormality, continuing to obtain the real-time status of the machine vehicle, and if the machine vehicle has an abnormality, issuing a corresponding abnormality prompt according to the abnormality of the real-time status, the method further includes: Obtain the abnormal identification plate of the machine vehicle with abnormal conditions; The corresponding position of the abnormal identity identification plate is inquired according to the abnormal identity identification plate, and the machine vehicle is searched for maintenance according to the corresponding position of the abnormal identity identification plate.

8. An abnormality detection device for a sintering machine trolley, characterized in that: The detection device comprises: Status detection module, used to obtain the real-time status of the machine vehicle; An abnormality prompt module is used to determine whether the machine vehicle has an abnormality according to the real-time status. If the machine vehicle has no abnormality, the real-time status of the machine vehicle is continuously acquired. If the machine vehicle has an abnormality, a corresponding abnormality prompt is issued according to the abnormality of the real-time status. The abnormal condition includes a wheel axle state, and a sensor is set between the left wheel and the right wheel of the machine vehicle. If the machine vehicle has an abnormal condition, a corresponding abnormal prompt is issued according to the abnormal condition of the real-time state, including: Acquire a first actual distance between the left wheel and the sensor, and acquire a first ideal distance between the left wheel and the sensor; obtaining a second actual distance between the right wheel and the sensor, and obtaining a second ideal distance between the right wheel and the sensor; calculating a first difference between the first actual distance and the first ideal distance; calculating a second difference between the second actual distance and the second ideal distance; If both the first difference and the second difference are not 0, and the sum of the first difference and the second difference is 0, it is determined that the wheel is in a misaligned state, and a wheel misaligned state prompt is issued; The abnormal condition also includes a loose wheel state. If the vehicle has an abnormal condition, issuing a corresponding abnormal prompt according to the abnormal condition in the real-time state includes: If the first preset threshold value is less than the absolute value of the first difference value and less than the second preset threshold value, and the second difference value is 0, it is determined that the left wheel is in a loose state, and a loose wheel state prompt is issued; If the first preset threshold value is less than the absolute value of the second difference value and is less than the second preset threshold value, and the first difference value is 0, it is determined that the right wheel is in a loose state, and a loose wheel state prompt is issued.

Images