A method for detecting safety of a hook of a liquid molten metal hoist trolley in a steel plant

Abstract

This invention provides a safety detection method for the hook of an overhead crane used for transporting molten metal in a steel plant, belonging to the metallurgical field. The method includes: S101, acquiring visual signals of the hook engagement or disengagement process between the hook and the ladle trunnion; S102, acquiring the crane weight signal and comparing the current weight signal with the average of the previous three weight signals. If the current weight signal is greater than the average, it indicates that the ladle has begun engagement, proceeding to S103; if the current weight signal is less than the average, it indicates that the ladle has begun disengagement, proceeding to S103; otherwise, returning to S101; S103, using a cloud-edge system architecture, the equipment in the visual signal processing cabinet processes the acquired visual signals through a deployed model to determine whether the engagement or disengagement process is safe and whether the ladle's posture after hoisting is safe; S104, if unsafe, the equipment in the visual signal processing cabinet issues an alarm signal. Using this invention, the safety of the engagement or disengagement process can be checked in a timely manner, and an alarm can be triggered promptly.

Description

Technical Field

[0001] This invention relates to the field of metallurgy, and in particular to a method for safety testing of the hooks of overhead cranes used for transporting molten metal in steel plants. Background Technology

[0002] During the production and storage of molten metal, major safety accidents such as tipping, falling, collisions, and leaks of storage containers like steel ladles and iron ladles are prone to occur. In particular, improper alignment or ineffective disengagement of the overhead crane hooks from the trunnions of the steel or iron ladles can lead to serious safety accidents.

[0003] The safety regulations for hoisting molten metal at high temperatures include the following requirements:

[0004] 4.1 Metal smelting enterprises should establish and improve safety management systems and safe operating procedures for the hoisting of molten metal at high temperatures, and strictly implement them.

[0005] 7.11 When hoisting containers filled with molten metal, a trial lift should be performed first. Before the trial lift, the crane operator and ground signalman should confirm that both sides of the hook are securely attached.

[0006] 7.12 When hoisting heavy tanks or slag tanks, ensure that the lugs on both sides are securely fastened before notifying the crane operator to begin hoisting.

[0007] Based on this standard, a safety inspection method for the hooks of overhead cranes used for transporting molten metal in steel plants will be developed, which will help prevent major safety accidents caused by the overturning of molten metal containers such as ladles and tin ladles, and ensure safe production.

[0008] In actual production, there are numerous locations where overhead cranes lift molten metal containers such as steel ladles and iron ladles. Deploying model inference on a centralized cloud platform may lead to problems such as communication delays and untimely alarms. Summary of the Invention

[0009] This invention provides a safety detection method for the hook of an overhead crane used for transporting molten metal in a steel plant. This method can promptly check the safety of the hooking or detaching process and issue an alarm in a timely manner, preventing major safety accidents caused by improper alignment or ineffective detachment during the hooking or detachment process. The method includes:

[0010] S101, acquire visual signals of the hooking or unhooking process between the left and right hooks of the overhead crane and the left and right trunnions of the ladle, and send the acquired visual signals to the visual signal processing equipment cabinet.

[0011] S102: Obtain the crane weight signal and compare the current weight signal with the average of the previous three weight signals. If the current weight signal is greater than the average, it means that the ladle has started to be hooked up, and proceed to S103. If the current weight signal is less than the average, it means that the ladle has started to be unhooked, and proceed to S103. Otherwise, return to S101.

[0012] S103 adopts a cloud-edge system architecture. The devices in the visual signal processing cabinet process the acquired visual signals through the deployed YOLACT one-stage segmentation model, the detection model for determining the safety of hook attachment or detachment, and the ladle posture detection model to determine whether the attachment or detachment process is safe and whether the ladle posture is safe after being lifted. The model upgrade management is carried out in the background cloud.

[0013] S104 If the hanging or detaching process is unsafe, or the posture of the ladle after being lifted is unsafe, the equipment in the visual signal processing cabinet will issue an alarm signal and transmit the alarm signal to the alarm indicator light.

[0014] Furthermore, the visual signal processing equipment cabinet includes: a switch, an embedded machine vision side device 1, an embedded machine vision side device 2, and an embedded machine vision side device 3;

[0015] The process of acquiring visual signals of the engagement or disengagement of the left and right hooks of the overhead crane with the left and right trunnions of the ladle, and sending the acquired visual signals to the visual signal processing equipment cabinet, includes:

[0016] During the connection process, the left hook 1 of the overhead crane is connected to the left trunnion 1 of the ladle, and the right hook 2 of the overhead crane is connected to the right trunnion 2 of the ladle.

[0017] During the disengagement process, the left hook 1 disengages from the left trunnion 1 of the steel ladle, and the right hook 2 disengages from the right trunnion 2 of the steel ladle.

[0018] The monitoring camera 1 located on the left side of the ladle acquires the visual signal of the hook on the left side of the crane and the trunnion on the left side of the ladle being engaged or disengaged. The signal is then forwarded to the embedded machine vision side device 1 via a switch to determine whether the engagement or disengagement process of the hook on the left side and the trunnion on the left side is safe.

[0019] The monitoring camera 2 located on the right side of the ladle acquires visual signals of the process of the hook on the right side of the crane and the trunnion on the right side of the ladle being engaged or disengaged, and forwards them to the embedded machine vision side device 2 through the switch to determine whether the process of the hook on the right side and the trunnion on the right side is safe.

[0020] The monitoring camera 3, located directly in front of the ladle, captures real-time dynamic information about the ladle's edge during the hanging or detaching process. This information is then forwarded to the embedded machine vision edge device 3 via a switch to determine if the ladle is tilted.

[0021] Furthermore, the cloud-edge system architecture is adopted. The devices in the visual signal processing cabinet process the acquired visual signals through the deployed YOLACT one-stage segmentation model, the detection model for determining the safety of hook attachment or detachment, and the ladle posture detection model. The determination of whether the attachment or detachment process is safe and whether the ladle posture is safe after being lifted includes:

[0022] Adopting a cloud-edge system architecture, embedded machine vision edge devices 1 and 2 perform real-time segmentation of the hook and trunnion in the visual signal using the YOLACT one-stage segmentation model to obtain mask images of the hook and trunnion. Based on the obtained mask images of the hook and trunnion, the center point of the trunnion and the center line of the hook are calculated. A three-line longitudinal centering detection algorithm based on the center of the trunnion and the center line of the hook is used to determine whether the hooking or unhooking process is safe. Embedded machine vision edge device 3 executes a ladle posture judgment algorithm to determine whether the posture of the ladle after being lifted is safe.

[0023] Furthermore, based on the obtained mask image of the trunnion, the calculation of the trunnion center point includes:

[0024] With the trunnion baffle in place, the center point of the trunnion is obtained by least squares circle fitting based on the obtained trunnion mask image.

[0025] Without a trunnion baffle, the center point of the trunnion is obtained by distance transformation based on the mask image of the trunnion.

[0026] Furthermore, the calculation of the hook centerline based on the obtained mask image of the hook includes:

[0027] Determine the orientation of the hook;

[0028] If the hook faces left, fit a straight line to the left side of the hook connection segment, take several points at equal intervals within the connection segment of the straight line, and based on the several points at equal intervals, make several vertical line segments covering the mask to the right. Select three line segments with the closest length, take the average value b of the midpoints of the three line segments, and shift the fitted straight line to the right by b to obtain the center line of the hook.

[0029] If the hook faces to the right, fit a straight line to the right side of the hook connection segment. Take several points at equal intervals within the connection segment of the straight line. Based on the points taken at equal intervals, make several vertical line segments covering the mask to the left. Select three line segments with the closest lengths. Take the average value b of the midpoints of the three line segments. Shift the fitted straight line to the left by b to obtain the center line of the hook.

[0030] Furthermore, the method of using a three-line longitudinal centering detection algorithm based on the trunnion center and the hook center line to determine whether the engagement or disengagement process is safe includes:

[0031] Calculate the lateral distance θ from the center point P0 of the trunnion to the longitudinal center line L0 of the hook, θ = |x1 - x0|; where x1 is the value of the longitudinal center line L0 of the hook on the x-coordinate.

[0032] During the connection process, the crane hook and ladle trunnion are matched properly by combining the crane weight change process and the relationship between θ and the set threshold θ0.

[0033] During the disengagement process, the crane hook and ladle trunnion are correctly disengaged by combining the crane weight change process and the relationship between θ and the set threshold θ1.

[0034] Furthermore, during the connection process, the weight change of the overhead crane and the relationship between θ and θ0 satisfy the following:

[0035] 1) The overhead crane hook and the ladle trunnion are not connected, and the overhead crane hook is not in place.

[0036] Conditions: θ >= θ0 and the crane weight signal α = 0;

[0037] 2) The crane hook and ladle trunnion are not connected, although the crane hook is basically in place.

[0038] Conditions: θ < θ0 and the crane weight signal α = 0;

[0039] 3) The crane hook and ladle trunnion are being attached.

[0040] Conditions: θ < θ0 and the crane weight signal α > α0 and α > 0, where α0 is the mean of the first three crane weight signals;

[0041] 4) The crane hook and the ladle trunnion were successfully connected.

[0042] Conditions: θ < θ0 and the crane weight signal α = α0 and α > 0.

[0043] Furthermore, during the disengagement process, the change in the crane's weight, and the relationship between θ and θ1, satisfy the following:

[0044] 1) The crane hook and the ladle trunnion are not detached.

[0045] Conditions: θ <= θ1 and the crane weight signal α = α0 and α > 0;

[0046] 2) The crane hook and the ladle trunnion begin to separate.

[0047] Conditions: θ < θ1 and the crane weight signal α <= α0 and α > 0; where α0 is the mean of the first three crane weight signals;

[0048] 3) The crane hook was not completely detached from the ladle trunnion, causing the ladle to fall to the ground and no longer exerting any force on the hook.

[0049] Conditions: θ < θ1 and the crane weight signal α = 0;

[0050] 4) The crane hook is completely detached from the ladle trunnion.

[0051] Conditions: θ>θ1 and the crane weight signal α=0.

[0052] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0053] 1) Three cameras are used, corresponding to the left, right and front views of the steel ladle respectively, to ensure that the hanging and detaching process is all-round and without blind spots;

[0054] 2) Based on the segmentation mask of the crane hook and ladle trunnion, a robust algorithm is then used to calculate the center line of the ladle trunnion and the center line of the crane hook to ensure the stability of the detection process;

[0055] 3) To achieve efficient operation of the model on side devices, the YOLACT one-stage segmentation model is adopted. Compared with the two-stage detection model, the one-stage segmentation model has the advantages of small computational load, small model size and fast inference speed.

[0056] 4) It adopts a flexible cloud-edge architecture, with the model deployed on low-cost and small-sized embedded machine vision edge devices. It can perform inference locally, alarms can be triggered locally, and model upgrade management is carried out in the background cloud. It is suitable for safety monitoring systems and can meet the actual needs of timely local alarms in industrial production. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a schematic diagram of the structure of the safety detection system for the hook of the overhead crane used for transporting molten metal in a steel plant, provided in an embodiment of the present invention.

[0059] Figure 2 This is a schematic diagram of the structure of the left hook 1 provided in an embodiment of the present invention;

[0060] Figure 3 This is a schematic diagram of the structure of the right hook 2 provided in an embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of the main view of a steel ladle provided in an embodiment of the present invention;

[0062] Figure 5 A flowchart illustrating the working process of the hook provided in an embodiment of the present invention;

[0063] Figure 6 A schematic diagram of a mask image for a hook and trunnion provided in an embodiment of the present invention;

[0064] Figure 7 This is a schematic diagram of the calculation process for the center point of the ladle trunnion provided in an embodiment of the present invention;

[0065] Figure 8 This is a schematic diagram of the hook centerline calculation process provided in an embodiment of the present invention;

[0066] Figure 9 This is a schematic diagram showing that the crane hook and the ladle trunnion are not connected and the crane hook is not in place, as provided in an embodiment of the present invention.

[0067] Figure 10 This is a schematic diagram showing that the crane hook and ladle trunnion are not yet connected, but the crane hook is basically in place, according to an embodiment of the present invention.

[0068] Figure 11 This is a schematic diagram of the crane hook and ladle trunnion being connected according to an embodiment of the present invention;

[0069] Figure 12 This is a schematic diagram illustrating the successful connection between the overhead crane hook and the ladle trunnion provided in an embodiment of the present invention.

[0070] Figure 13 This is a flowchart illustrating the uncoupling process provided in an embodiment of the present invention.

[0071] Figure 14 This is a schematic diagram showing that the crane hook and ladle trunnion have not detached, as provided in an embodiment of the present invention.

[0072] Figure 15 This is a schematic diagram showing the detachment of the overhead crane hook from the ladle trunnion provided in an embodiment of the present invention.

[0073] Figure 16 This is a schematic diagram illustrating the relationship between the overhead crane hook and the ladle trunnion in an embodiment of the present invention, where the ladle falls to the ground and there is no longer any force exerted on the hook.

[0074] Figure 17 This is a schematic diagram showing the complete detachment of the overhead crane hook from the ladle trunnion, as provided in an embodiment of the present invention. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0076] Figure 1This invention relates to a safety detection system for the hooks of overhead cranes used for transporting molten metal in steel plants, and corresponds to the safety detection method for the hooks of overhead cranes used for transporting molten metal in steel plants provided in this embodiment. To better understand this invention, the main equipment is briefly described below:

[0077] 1) Steel ladle

[0078] A steel ladle, also known as a molten steel ladle, is a container used in steel mills or foundries to receive and transfer molten metal. An iron ladle is also a container for receiving and transferring molten metal. The difference is that a steel ladle is used to transfer molten steel, while an iron ladle is used to transfer molten iron. Iron and steel ladles have similar physical structures, both having two trunnions and requiring an overhead crane for transport.

[0079] 2) Overhead crane hook

[0080] Hook 1: This hook operates on the left side of the ladle, such as... Figure 2 As shown. Its function is to be attached to the left trunnion 1 of the ladle to lift and transport the ladle.

[0081] Hook 2: This hook operates on the right side of the ladle, such as... Figure 3 As shown. Its function is to be attached to the right trunnion 2 of the ladle to lift and transport the ladle.

[0082] 3) Surveillance cameras

[0083] Monitoring camera 1: Located on the left side of the ladle, directly facing the left trunnion 1 of the ladle. Its function is to dynamically capture in real time the visual signals of the process of the left hook 1 and the left trunnion 1 being engaged (also called hooking) or disengaged (also called unhooking).

[0084] Monitoring camera 2: Located on the right side of the ladle, directly facing the right trunnion 2 of the ladle. Its function is to dynamically capture visual signals of the engagement or disengagement process between the right hook 2 and the right trunnion 2 in real time.

[0085] Monitoring camera 3: Positioned directly in front of the ladle, its function is to dynamically capture real-time information about the ladle's edge from the front, in order to further determine whether the ladle is tilted. The main view of the ladle captured by monitoring camera 3 is shown below. Figure 4 As shown.

[0086] 4) Visual signal processing equipment cabinet

[0087] The visual signal processing equipment cabinet is a single electrical cabinet containing a switch, embedded machine vision edge device 1 (hereinafter referred to as "Embedded Device 1"), embedded machine vision edge device 2 (hereinafter referred to as "Embedded Device 2"), and embedded machine vision edge device 3 (hereinafter referred to as "Embedded Device 3"), and externally connected to an alarm indicator light. Embedded machine vision edge devices 1 and 2 are equipped with a detection model for determining the safety of hook engagement or disengagement (this model uses a three-line longitudinal centering detection algorithm based on the trunnion center and hook centerline to determine whether the engagement or disengagement process is safe). Embedded machine vision edge device 3 is equipped with a ladle posture detection model (this model uses a ladle posture judgment algorithm to determine whether the posture of the ladle after lifting is safe). The model calculation results are transmitted to the host computer, and alarms are triggered as needed, while alarm logs are also transmitted to the host computer.

[0088] Switch: Its function is to forward the visual signals captured by the surveillance cameras. The dynamic visual signals captured by surveillance cameras 1, 2, and 3 are connected to the switch of the visual signal processing equipment via their respective output network cables, and then forwarded to the corresponding embedded machine vision side-side devices.

[0089] Embedded machine vision edge device 1: Its function is to process the visual signals captured by the monitoring camera 1 via the switch during the hooking or unhooking process of the left hook 1 and the left trunnion 1, and to determine whether the hooking or unhooking process of the left hook is safe based on the visual signals, and to issue an alarm as needed.

[0090] Embedded machine vision edge device 2: Its function is to process the visual signals captured by the monitoring camera 2 via the switch during the hooking or unhooking process of the right hook 2 and the right trunnion 2, and to determine whether the hooking or unhooking process of the right hook is safe based on the visual signals, and to issue an alarm as needed.

[0091] Embedded machine vision edge device 3: Its function is to process the visual signal of the steel ladle forward edge information captured by the monitoring camera 3 via the switch, and to determine whether the steel ladle is placed horizontally or safely based on the visual signal, and to issue an alarm as needed.

[0092] Alarm indicator light: Located outside the visual signal processing equipment cabinet, its function is to receive alarm signals generated by the internal equipment of the visual signal processing equipment cabinet through analysis and calculation, and to flash alarms in different ways according to the alarm signals.

[0093] 5) Host computer

[0094] The host computer is connected to the visual signal processing equipment cabinet via a network cable. The host computer, acting as the cloud, is typically a cloud computing server. Its function is to receive the results of edge computing analysis from the devices inside the visual signal processing equipment cabinet and display them visually in the management system on the host computer. The background cloud manages different versions and scenarios of models, supports incremental training, version management, and upgrade distribution, and can also collect alarm logs during operation for statistical analysis.

[0095] Before or after hoisting the ladle, the overhead crane hook needs to be engaged or disengaged from the ladle trunnion. To better understand this invention, the processes of engaging and disengaging the hook are described separately.

[0096] 1) Hooking process, such as Figure 5 As shown, the specific steps may include:

[0097] Step 1: Acquiring Visual Signals. During the connection process, the left hook 1 of the overhead crane is connected to the left trunnion 1 of the ladle, and the right hook 2 of the overhead crane is connected to the right trunnion 2 of the ladle. The visual signal of the connection process between the left hook and the left trunnion is acquired by the monitoring camera 1 located on the left side of the ladle and forwarded to the embedded machine vision edge device 1 through the switch to determine whether the connection process between the left hook and the left trunnion is safe. The visual signal of the connection process between the right hook and the right trunnion is acquired by the monitoring camera 2 located on the right side of the ladle and forwarded to the embedded machine vision edge device 2 through the switch to determine whether the connection process between the right hook and the right trunnion is safe. The monitoring camera 3 located directly in front of the ladle captures the edge information of the ladle in real time from the front of the ladle during the connection process and forwards it to the embedded machine vision edge device 3 through the switch to determine whether the ladle is tilted.

[0098] Step 2: Weight signal determination. Obtain the crane weight signal and compare the current weight signal with the average of the previous three weight signals. If the current weight signal is greater than the average, it indicates that the ladle is being attached, and proceed to Step 3; otherwise, return to Step 1.

[0099] Step 3: Signal Preprocessing. To improve the precision and accuracy of image processing in Step 4, the images in the real-time visual signals transmitted to the embedded machine vision side devices 1, 2, and 3 can be preprocessed to adapt to harsh working conditions. The preprocessing operations include brightness correction and noise reduction.

[0100] Step 4: Model Algorithm Processing. The embedded machine vision edge device 1 uses the YOLACT one-stage segmentation model to perform real-time segmentation of the hook and trunnion in the visual signal, segmenting the mask images of the hook and trunnion from the input image, such as... Figure 6As shown. Then, depending on the actual situation of whether or not there is a trunnion baffle, different methods are used to calculate the trunnion center point; depending on the hook orientation, different methods are used to calculate the hook center line; finally, a three-line longitudinal center positioning detection algorithm based on the trunnion center and the hook center line is used to determine whether the hanging process is safe. The embedded machine vision side device 2 performs similar tasks to the embedded machine vision side device 1, but performs the aforementioned tasks on the other side of the ladle; the embedded machine vision side device 3 executes the ladle posture judgment algorithm to assist in judging whether the posture of the ladle after being lifted is safe.

[0101] like Figure 7 As shown, the algorithm for calculating the center point of the ladle trunnion is divided into two types: one with a trunnion baffle and the other with a trunnion baffle.

[0102] With the trunnion baffle in place, the center point of the trunnion is obtained by least squares circle fitting based on the obtained trunnion mask image, and the required point set is extracted from the edge contour.

[0103] Without a trunnion baffle, the center point of the trunnion is obtained by distance transformation based on the mask image of the trunnion.

[0104] like Figure 8 As shown, the prerequisite for implementing the crane hook centerline extraction algorithm is to determine whether the hook is facing left or right.

[0105] If the hook faces left, fit a straight line to the left side of the hook connection segment, then take several points at equal intervals within the connection segment of the straight line. Based on the points taken at equal intervals, then make several vertical line segments covering the mask to the right. Select three line segments with the closest length, take the average value b of the midpoints of the three line segments, and finally translate the fitted straight line to the right by b to obtain the center line of the hook.

[0106] If the hook faces to the right, fit a straight line to the right side of the hook connection segment, then take several points at equal intervals within the connection segment of the straight line. Based on the points taken at equal intervals, then make several vertical line segments covering the mask to the left. Select three line segments with the closest length, take the average value b of the midpoints of the three line segments, and finally translate the fitted straight line to the left by b to obtain the center line of the hook.

[0107] In this embodiment, the safety judgment process for hook attachment is as follows:

[0108] As can be seen from step 2, when the crane weight signal is compared with the average value α0 of the previous three weight signals, if it is greater than the average value α0, it indicates that there is a hooking operation between the crane hook and the ladle trunnion. Calculate the lateral distance θ from the center point P0(x0,y0) of the trunnion to the longitudinal center line L0(x=x1) of the hook, where θ=|x1-x0|.

[0109] During the connection process, the crane hook and ladle trunnion are matched properly by combining the crane weight change process and the relationship between θ and the set threshold θ0. To ensure connection safety, θ0 is required to take a small value.

[0110] In this embodiment, during the coupling process, the weight change of the overhead crane and the relationship between θ and θ0 satisfy the following:

[0111] 1) The overhead crane hook and ladle trunnion are not connected, or the overhead crane hook is not in place. Figure 9 As shown

[0112] Conditions: θ >= θ0 and weight signal α = 0

[0113] 2) The crane hook and ladle trunnion are not connected, although the crane hook is basically in place. Figure 10 Conditions shown: θ < θ0 and weight signal α = 0

[0114] 3) The crane hook and ladle trunnion are being connected, such as... Figure 11 As shown

[0115] Conditions: θ < θ0 and weight signal α > α0 and α > 0

[0116] 4) The crane hook and the ladle trunnion are successfully connected, such as... Figure 12 As shown

[0117] Conditions: θ < θ0 and the weight signal α = α0 and α > 0

[0118] Step 5: Hazard Alarm. If the hanging process is unsafe, or the ladle is not in a safe position after being lifted, the vision signal processing equipment cabinet will issue an alarm signal. This alarm signal can be issued by any one of the embedded machine vision side devices 1, 2, or 3, and will be transmitted to the alarm indicator light to report the potential major dangers during the hanging process and prompt manual intervention; otherwise, return to Step 1.

[0119] 2) The uncoupling process, such as Figure 13 As shown, the specific steps may include:

[0120] Step 1: Acquire visual signals. During the disengagement process, the left hook 1 disengages from the left trunnion 1 of the ladle, and the right hook 2 disengages from the right trunnion 2 of the ladle. The monitoring camera 1 located on the left side of the ladle acquires the visual signal of the disengagement process of the left hook of the crane and the left trunnion of the ladle, and forwards it to the embedded machine vision edge device 1 through the switch to determine whether the disengagement process of the left hook and the left trunnion is safe. The monitoring camera 2 located on the right side of the ladle acquires the visual signal of the disengagement process of the right hook of the crane and the right trunnion of the ladle, and forwards it to the embedded machine vision edge device 2 through the switch to determine whether the disengagement process of the right hook and the right trunnion is safe. The monitoring camera 3 located directly in front of the ladle captures the edge information of the ladle in real time from the front of the ladle during the disengagement process, and forwards it to the embedded machine vision edge device 3 through the switch to determine whether the ladle is tilted.

[0121] Step 2: Weight signal judgment. Obtain the crane weight signal and compare the current weight signal with the average of the previous three weight signals. If it is less than the preset threshold, it means that the ladle has started to unhook, and proceed to step 3; otherwise, return to step 1.

[0122] Step 3: Signal Preprocessing. To improve the precision and accuracy of image processing in Step 4, the images in the real-time visual signals transmitted to the embedded machine vision side devices 1, 2, and 3 can be preprocessed to adapt to harsh working conditions. The preprocessing operations include brightness correction and noise reduction.

[0123] Step 4: Model Algorithm Processing. Embedded machine vision edge device 1 uses the YOLACT one-stage segmentation model to segment the hook and trunnion in the visual signal in real time, obtaining mask images of the hook and trunnion. Then, depending on the actual situation of whether or not there is a trunnion baffle, different methods are used to calculate the trunnion center point; depending on the hook orientation, different methods are used to calculate the hook center line; finally, a three-line longitudinal center positioning detection algorithm based on the trunnion center and the hook center line is used to determine whether the disengagement process is safe; Embedded machine vision edge device 2 performs similar tasks to embedded machine vision edge device 1, but performs the aforementioned tasks on the other side of the ladle; Embedded machine vision edge device 3 executes a ladle posture judgment algorithm to assist in determining whether the posture of the ladle after lifting is safe.

[0124] In this embodiment, the safety judgment process for hook detachment is as follows:

[0125] According to step 2, when the crane weight signal is compared with the average value α0 of the previous three weight signals, if it is less than the average value α0, it indicates that the crane hook and the ladle trunnion are disconnected. Calculate the lateral distance θ from the trunnion center point P0(x0,y0) to the longitudinal center line L0 (x=x1) of the hook, where θ=|x1-x0|.

[0126] During the disengagement process, the crane hook and ladle trunnion are correctly disengaged by considering the changes in the crane's weight and the relationship between θ and the set threshold θ1. To ensure complete disengagement, θ1 is required to have a relatively large value.

[0127] In this embodiment, during the disengagement process, the change in the crane's weight and the relationship between θ and θ1 satisfy the following:

[0128] 1) The crane hook and the ladle trunnion are not detached, such as Figure 14 As shown

[0129] Conditions: θ <= θ1 and the weight signal α = α0 and α > 0

[0130] 2) The crane hook and the ladle trunnion begin to detach, such as Figure 15 As shown

[0131] Conditions: θ < θ1 and the weight signal α <= α0 and α > 0

[0132] 3) The crane hook and the ladle trunnion were not completely disengaged, the ladle fell to the ground, and there was no longer any force between it and the hook. Figure 16 As shown

[0133] Conditions: θ < θ1 and weight signal α = 0

[0134] 4) The crane hook is completely detached from the ladle trunnion, such as Figure 17 As shown

[0135] Conditions: θ > θ1 and weight signal α = 0

[0136] Step 5: Hazard Alarm. If the detachment process is unsafe, or the ladle is not in a safe position after being lifted, the vision signal processing equipment cabinet will issue an alarm signal. This alarm signal can be issued by any one of the embedded machine vision side devices 1, 2, or 3, and will be transmitted to the alarm indicator light to report the potential danger during the detachment process and prompt manual intervention; otherwise, return to Step 1.

[0137] In summary, this invention provides a safety detection method for overhead crane hooks used in the hoisting of molten metal in steel plants. This method addresses two major safety accidents that can lead to the tipping of ladles or tundishes: one where the overhead crane hook is not properly engaged with the ladle trunnion during hoisting, and the other where the overhead crane hook is not disengaged from the trunnion sufficiently during lowering. The method includes at least the following features:

[0138] Beneficial effects:

[0139] 1) Three cameras are used, corresponding to the left, right and front views of the steel ladle respectively, to ensure that the hanging and detaching process is all-round and without blind spots;

[0140] 2) Based on the segmentation mask of the crane hook and ladle trunnion, a robust algorithm is then used to calculate the center line of the ladle trunnion and the center line of the crane hook to ensure the stability of the detection process;

[0141] 3) To achieve efficient operation of the model on side devices, the YOLACT one-stage segmentation model is adopted. Compared with the two-stage detection model, the one-stage segmentation model has the advantages of small computational load, small model size and fast inference speed.

[0142] 4) It adopts a flexible cloud-edge architecture, with the model deployed on low-cost and small-sized embedded machine vision edge devices. It can perform inference locally, alarms can be triggered locally, and model upgrade management is carried out in the background cloud. It is suitable for safety monitoring systems and can meet the actual needs of timely local alarms in industrial production.

[0143] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0144] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for safety inspection of the hook of an overhead crane used for transporting molten metal in a steel plant, characterized in that, include: S101, acquire visual signals of the hooking or unhooking process between the left and right hooks of the overhead crane and the left and right trunnions of the ladle, and send the acquired visual signals to the visual signal processing equipment cabinet. S102: Obtain the crane weight signal and compare the current weight signal with the average of the previous three weight signals. If the current weight signal is greater than the average, it means that the ladle has started to be hooked up, and proceed to S103. If the current weight signal is less than the average, it means that the ladle has started to be unhooked, and proceed to S103. Otherwise, return to S101. S103 adopts a cloud-edge system architecture. The devices in the visual signal processing cabinet process the acquired visual signals through the deployed YOLACT one-stage segmentation model, the detection model for determining the safety of hook attachment or detachment, and the ladle posture detection model to determine whether the attachment or detachment process is safe and whether the ladle posture is safe after being lifted. The model upgrade management is carried out in the background cloud. S104 If the hanging or detaching process is unsafe, or the posture of the steel ladle after being lifted is unsafe, the equipment in the visual signal processing cabinet will issue an alarm signal and transmit the alarm signal to the alarm indicator light for alarm. The visual signal processing equipment cabinet includes: a switch, an embedded machine vision side device 1, an embedded machine vision side device 2, and an embedded machine vision side device 3; The process of acquiring visual signals of the engagement or disengagement of the left and right hooks of the overhead crane with the left and right trunnions of the ladle, and sending the acquired visual signals to the visual signal processing equipment cabinet, includes: During the connection process, the left hook of the overhead crane is connected to the left trunnion of the ladle, and the right hook of the overhead crane is connected to the right trunnion of the ladle. During the disengagement process, the left hook disengages from the left trunnion of the ladle, and the right hook disengages from the right trunnion of the ladle. The visual signal of the connection or disconnection process between the left hook of the crane and the left trunnion of the ladle is obtained by the monitoring camera located on the left side of the ladle, and forwarded to the embedded machine vision side device 1 through the switch to determine whether the connection or disconnection process between the left hook and the left trunnion is safe. The visual signal of the hook on the right side of the crane and the trunnion on the right side of the ladle is obtained by the monitoring camera located on the right side of the ladle, and forwarded to the embedded machine vision side device 2 through the switch to determine whether the hook on the right side and the trunnion on the right side are safe. The monitoring camera located directly in front of the ladle captures real-time dynamic information about the ladle's edge during the hanging or detaching process from the front of the ladle. This information is then forwarded to the embedded machine vision side device 3 via a switch to determine whether the ladle is tilted. The cloud-edge system architecture is adopted, and the equipment in the visual signal processing cabinet processes the acquired visual signals through the deployed YOLACT one-stage segmentation model, the detection model for determining the safety of hook attachment or detachment, and the ladle posture detection model. The determination of whether the attachment or detachment process is safe and whether the ladle's posture after being lifted is safe includes: Adopting a cloud-edge system architecture, embedded machine vision edge devices one and two use the YOLACT one-stage segmentation model to segment the hooks and trunnions in the visual signal in real time, obtaining mask images of the hooks and trunnions. Based on the obtained mask images of the hooks and trunnions, the center point of the trunnion and the center line of the hook are calculated. A three-line longitudinal centering detection algorithm based on the center of the trunnion and the center line of the hook is used to determine whether the hooking or unhooking process is safe. Embedded machine vision edge device three executes a ladle posture judgment algorithm to determine whether the posture of the ladle after it is lifted is safe.

2. The method for safety detection of the hook of an overhead crane for transporting molten metal in a steel plant according to claim 1, characterized in that, Based on the obtained mask image of the trunnion, the center point of the trunnion is calculated as follows: With the trunnion baffle in place, the center point of the trunnion is obtained by least squares circle fitting based on the obtained trunnion mask image. Without a trunnion baffle, the center point of the trunnion is obtained by distance transformation based on the mask image of the trunnion.

3. The method for safety detection of hooks on overhead cranes used for transporting molten metal in steel plants according to claim 1, characterized in that, The calculation of the hook centerline based on the obtained mask image includes: Determine the orientation of the hook; If the hook faces left, fit a straight line to the left side of the hook connection segment, take several points at equal intervals within the connection segment of the straight line, and based on the several points at equal intervals, make several vertical line segments covering the mask to the right. Select three line segments with the closest length, take the average value b of the midpoints of the three line segments, and shift the fitted straight line to the right by b to obtain the center line of the hook. If the hook faces to the right, fit a straight line to the right side of the hook connection segment. Take several points at equal intervals within the connection segment of the straight line. Based on the points taken at equal intervals, make several vertical line segments covering the mask to the left. Select three line segments with the closest lengths. Take the average value b of the midpoints of the three line segments. Shift the fitted straight line to the left by b to obtain the center line of the hook.

4. The method for safety detection of the hook of an overhead crane for transporting molten metal in a steel plant according to claim 1, characterized in that, The method of using a three-line longitudinal centering detection algorithm based on the trunnion center and the hook centerline to determine whether the engagement or disengagement process is safe includes: Calculate the lateral distance θ from the center point P0 of the trunnion to the longitudinal center line L0 of the hook, θ=|x1-x0|; where x1 is the value of the longitudinal center line L0 of the hook on the x-coordinate; During the connection process, the crane hook and ladle trunnion are matched properly by combining the crane weight change process and the relationship between θ and the set threshold θ0. During the disengagement process, the crane hook and ladle trunnion are correctly disengaged by combining the crane weight change process and the relationship between θ and the set threshold θ1.

5. The method for safety inspection of the hook of an overhead crane for transporting molten metal in a steel plant according to claim 4, characterized in that, During the connection process, the weight change of the overhead crane and the relationship between θ and θ0 satisfy the following: 1) The overhead crane hook and the ladle trunnion are not connected, and the overhead crane hook is not in place. Conditions: θ>=θ0 and the crane weight signal α=0; 2) The crane hook and ladle trunnion are not connected, although the crane hook is basically in place. Conditions: θ < θ0 and the crane weight signal α = 0; 3) The crane hook and ladle trunnion are being attached. Conditions: θ < θ0 and the crane weight signal α >= α0 and α > 0, where, α0 is the average of the first three overhead crane weight signals; 4) The crane hook and the ladle trunnion were successfully connected. Conditions: θ < θ0 and the crane weight signal α = α0 and α > 0.

6. The method for safety inspection of the hook of an overhead crane for transporting molten metal in a steel plant according to claim 4, characterized in that, During the disengagement process, the change in the crane's weight and the relationship between θ and θ1 satisfy the following: 1) The crane hook and the ladle trunnion are not detached. Conditions: θ <= θ1 and the crane weight signal α = α0 and α > 0; 2) The crane hook and the ladle trunnion begin to separate. Conditions: θ < θ1 and the crane weight signal α <= α0 and α > 0; where α0 is the mean of the first three crane weight signals; 3) The crane hook was not completely detached from the ladle trunnion, causing the ladle to fall to the ground and no longer exerting any force on the hook. Conditions: θ < θ1 and the crane weight signal α = 0; 4) The crane hook and the ladle trunnion are completely detached. Conditions: θ>θ1 and the crane weight signal α=0.

Images