CN-121982666-A - Automatic slag raking method and system for induction furnace and computer storage medium

CN121982666ACN 121982666 ACN121982666 ACN 121982666ACN-121982666-A

Abstract

The invention belongs to the field of automatic control, and particularly relates to an automatic slag raking method and system for an induction furnace and a computer storage medium; the automatic slag raking method comprises the steps of controlling a robot arm to carry an image acquisition device to a detection point and obtain a fusion image at an induction furnace mouth, wherein the fusion image is obtained through image registration fusion based on visible light and a thermal image, identifying and dividing a furnace mouth frame, a slag liquid mask area and a slag mask area based on a semantic division model according to the fusion image, calculating the average liquid level height and the average slag thickness based on the position coordinates of the furnace mouth frame in the image and the first average depth of field and the second average depth of field corresponding to the slag mask area if the slag liquid mask area is identified, and calculating the first furnace feeding height and the first initial rake height according to the average liquid level height and the average slag thickness if the calculated average liquid level height is within a preset range and the average slag thickness is larger than a preset thickness threshold.

Inventors

TAN SHENGHU
MA SHUANG
WANG JUNCAN

Assignees

北京瓦特曼智能科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251218

Claims (10)

1. An automatic slag raking method for an induction furnace, comprising: the control robot arm carries the image acquisition device to the detection point and acquires a fusion image at the induction furnace mouth, wherein the fusion image is obtained by image registration fusion based on visible light and thermal images; Identifying and dividing a furnace mouth frame, a slag liquid mask area and a slag mask area based on a semantic division model according to the fusion image; if the slag liquid mask area is identified, calculating the average height of the liquid level and the average thickness of slag based on the position coordinates of the furnace mouth frame in the image, the first average depth of field corresponding to the obtained slag liquid mask area and the second average depth of field corresponding to the slag mask area; If the calculated average height of the liquid level is within a preset range and the average thickness of the slag is larger than a preset thickness threshold, calculating a first furnace feeding height and a first initial raking height according to the average height of the liquid level and the average thickness of the slag; The method comprises the steps of controlling a machine arm to carry a slag raking tool to enter a first circulation mode, enabling the machine arm to extend into a preset depth of a furnace mouth at a first furnace inlet height during each circulation, enabling the machine arm to descend to a first target height for raking, returning the machine arm to the detection point after a preset first circulation time, and enabling the first target height to be reduced in an equal amount according to a first difference value from the first initial raking height according to the circulation time.
2. The automatic slag raking method for an induction furnace according to claim 1, wherein the fused image is obtained by image registration fusion based on visible light and thermal images, comprising: Extracting features of the visible light and the thermal image respectively, and carrying out rapid feature matching by using rapid approximate nearest neighbor search to obtain matching point pairs; And calculating transformation parameters according to the matching point pairs, mapping the thermal image to a visible light image coordinate system, and reconstructing a fusion image based on Laplacian pyramid fusion.
3. The automatic slag raking method for an induction furnace according to claim 1, wherein the identifying and dividing the furnace mouth frame, the slag liquid mask region, and the slag mask region based on the semantic division model from the fused image includes: Providing a U-Net++ network architecture, and taking the encoder of the U-Net++ as a double branch, namely an RGB branch and a thermal image branch; At the i layer and the j nested jump joint, carrying out attention weighted fusion on RGB modal characteristics and thermal image modal characteristics of the thermal image, and transmitting the fused characteristics to the next layer; and inputting the fused features into a decoder based on the jump connection structure of the U-Net to generate a slag liquid mask region and a slag mask region segmentation result.
4. The automatic slag raking method for an induction furnace according to claim 1, wherein the obtaining a first average depth of field corresponding to a slag liquid mask area and a second average depth of field corresponding to the slag mask area based on the position coordinates of the furnace mouth frame in the image, and calculating the average height of the liquid surface and the average thickness of the slag, comprises: Extracting a mask area and corresponding position coordinates in the furnace mouth frame, and cutting a target image area according to the furnace mouth frame coordinates; Extracting a first average depth of field corresponding to the slag liquid mask and a second average depth of field corresponding to the slag mask based on the target image area; calculating to obtain the average height of the liquid level based on the first average depth of field and the position coordinates; And calculating to obtain the average thickness of the slag based on the average height of the liquid level, the second average depth of field and the position coordinates.
5. The automatic slag raking method for an induction furnace according to claim 1, the automatic slag raking method for the induction furnace is characterized by further comprising the following steps: If the slag liquid mask area is not identified, calculating the current height of slag based on the position coordinates of the furnace mouth frame in the image and the second average depth of field corresponding to the acquired slag mask area; Controlling the robot arm to carry the slag raking tool to enter a second circulation mode, calculating a second initial raking height based on the current slag height, and controlling the robot arm to carry the slag raking tool to extend into a preset depth of a furnace mouth from the detection point each time and descend to a second target height for raking, wherein the second target height is reduced in sequence from the second initial raking height according to a second difference value according to circulation times; and in each cycle, when the robot arm carries a slag raking tool to return to the detection point, controlling the image acquisition device to acquire the fusion image again until the slag liquid mask area is identified.
6. The automatic slag raking method for an induction furnace of claim 5, further comprising: And stopping raking slag and prompting the first information if the calculated average height of the liquid level is not in the preset range.
7. The automatic slag raking method for an induction furnace of claim 6, further comprising: If the calculated average height of the liquid level is within a preset range and the average thickness of the slag is smaller than or equal to a preset thickness threshold value, calculating a ratio value of the slag liquid mask area and the slag mask area, and judging whether the ratio value is larger than the preset ratio threshold value; Stopping raking slag and entering a standby state to wait for a next slag raking signal if the ratio value is greater than a preset ratio threshold value; If the ratio value is smaller than or equal to a preset ratio threshold value, calculating a third initial raking height based on the average liquid level height; and controlling the mechanical arm to carry the slag raking tool into a third circulation mode, extending to a preset depth of the furnace mouth from the detection point every time, and descending to a third initial raking height to rake the slag until the ratio value is larger than a preset ratio threshold value if the ratio value is larger than the preset ratio threshold value.
8. An automatic slag raking system, comprising: A multi-axis robotic arm; The slag raking tool is connected to the tail end of the multi-axis mechanical arm; The image acquisition device comprises at least one visible light probe and an infrared probe; A controller electrically connected to the multi-axis robot arm, the image acquisition device, and capable of being implemented as the automatic slag raking method for an induction furnace of any one of claims 1 to 7.
9. The automatic slag raking system of claim 8, wherein the slag raking tool comprises a fixed seat and a rake rod part, a plurality of mounting bayonets are arranged on the fixed seat, and one end of the rake rod part can be connected with different mounting bayonets to switch different lengths; The periphery of the fixed seat is provided with a mounting bracket, and the image acquisition device is fixedly connected to the periphery of the harrow rod part through the mounting bracket; the fixed seat deviates from the one end of harrow pole portion is provided with the balancing weight, the balancing weight is relative the center pin asymmetry setting of fixed seat.
10. A computer storage medium, characterized in that the computer readable medium has stored therein a computer program, wherein the computer program is arranged to perform the automatic slag raking method for an induction furnace according to any one of claims 1 to 7 when run.

Description

Automatic slag raking method and system for induction furnace and computer storage medium Technical Field The invention belongs to the field of automatic control, and particularly relates to an automatic slag raking method and system for an induction furnace and a computer storage medium. Background In the zinc liquid treatment link of a smelting plant, slag can be inevitably generated on the surface layer of the zinc liquid, and if the slag is not cleaned timely, the blanking process in the process and the exhaust inside the zinc liquid are affected. At present, aiming at cleaning of slag on the surface layer of zinc liquid, the traditional mode mainly relies on manual slag raking operation, as demonstrated in fig. 1, an operator holds a slag raking tool, judges the position of slag by self experience and hand feeling, manually cleans the slag on the liquid surface, a plurality of dangerous sources exist on the slag raking operation site, and the material splashing in the operation process can cause damage such as bruise, so that serious accidents even endanger lives, thereby belonging to one of extremely dangerous operation areas of smelting plants. In addition, because manual slag raking mainly relies on manual experience operation, the accuracy and consistency of slag raking tool positions are difficult to ensure. In the operation process, zinc liquid is easily carried out together due to improper operation, so that liquid level loss is caused, normal treatment and subsequent production flow of the zinc liquid are affected, production efficiency and product quality are reduced, more importantly, the tail end of a slag raking tool for manually raking slag often enters high-temperature slag liquid, tail end loss is serious, and the original material proportion in an induction furnace is affected. Disclosure of Invention Aiming at the defects of high safety risk, poor operation accuracy, easy liquid level loss, unbalanced material proportion, serious tool loss and the like of manual slag raking, the application provides an automatic slag raking method, an automatic slag raking system and a computer storage medium for an induction furnace, so as to realize safe, accurate and efficient full-automatic slag raking operation. The automatic slag raking method for the induction furnace comprises the steps of controlling a robot arm to carry an image acquisition device to a detection point and obtain a fusion image at an induction furnace mouth, wherein the fusion image is obtained through image registration fusion based on visible light and a thermal image, identifying and dividing a furnace mouth frame, a slag liquid mask area and a slag mask area based on a semantic dividing model according to the fusion image, calculating the average liquid level height and the average slag thickness based on the position coordinates of the furnace mouth frame in the image and the first average depth of field corresponding to the obtained slag liquid mask area and the second average depth of field corresponding to the obtained slag liquid mask area if the slag liquid mask area is identified, calculating the first feeding height and the first initial rake height according to the average liquid level height and the average slag thickness if the calculated average liquid level height is within a preset range and the average slag thickness is larger than a preset thickness threshold, and controlling the robot arm to carry slag tools to stretch into the preset depth of the furnace mouth at the first feeding height each time and descend to the preset target height, and sequentially lower the rake height to the first target height according to the first initial rake height, wherein the first rake height is sequentially lowered from the first target height to the first target value. The fusion image is obtained by image registration fusion based on visible light and thermal images, and comprises the steps of extracting features of the visible light and the thermal images respectively, carrying out rapid feature matching by utilizing rapid approximate nearest neighbor search to obtain matching point pairs, calculating transformation parameters according to the matching point pairs, mapping the thermal images to a visible light image coordinate system, and reconstructing the fusion image based on Laplacian pyramid fusion. In a further scheme of the application, a furnace mouth frame, a slag liquid mask area and a slag mask area are identified and segmented based on a semantic segmentation model according to a fusion image, and the method comprises the steps of providing a U-Net++ network architecture, enabling a U-Net++ encoder to be a double branch, namely an RGB branch and a thermal image branch, carrying out attention weighted fusion on RGB mode characteristics and thermal image mode characteristics of a thermal image at an ith layer and a jth nested jump joint, transmitting the fused characteristics to a next layer, inputting the fu