A high-magnetic-induction oriented silicon steel plate strip shot blasting descaling unit based on supervised learning
Patent Information
- Application Number
- CN202521795576.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Utility models(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-22
- Publication Date
- 2026-09-08
- Estimated Expiration
- 2035-08-22
AI Technical Summary
传统方法难以精确控制这些参数，导致除鳞效果不稳定，甚至损伤硅钢板基体，影响其磁性能和机械性能
[0023] 1) This utility model studies the fracture mechanism of iron oxide scale on the surface of wide hot-rolled plate under external impact by combining finite element simulation with interference theory analysis, and uses supervised learning algorithm to explore the relationship between shot parameters and descaling effect, so as to achieve precise control of shot peening parameters and improve descaling efficiency and accuracy.
Smart Images
Figure CN224725671U_ABST
Abstract
Description
Technical Field
[0001] This utility model relates to the field of surface treatment technology for oriented silicon steel sheets and strips, specifically to a shot peening and descaling unit for oriented silicon steel sheets and strips based on supervised learning. Background Technology
[0002] Grain-oriented silicon steel is an important soft magnetic material widely used in electrical equipment such as transformers and motors. Hot-rolled grain-oriented silicon steel sheets and strips typically have a dense layer of iron oxide scale on their surface, mainly composed of ferric oxide, magnetite, and some iron oxide structures. This scale affects the magnetic properties, surface quality, and subsequent processing performance of the silicon steel sheet, thus requiring its removal. Traditional shot peening descaling technology uses high-speed shot to impact the surface of the hot-rolled sheet, causing the iron oxide scale to fracture and detach. However, traditional shot peening descaling technology suffers from problems such as difficulty in precisely controlling the descaling effect, easy damage to the substrate, and lack of intelligent control. Parameters such as the shot material, size, projection angle, projection speed, and projection pulse duration have a significant impact on the descaling effect. Traditional methods struggle to precisely control these parameters, leading to unstable descaling results and even damage to the silicon steel substrate, affecting its magnetic and mechanical properties. Therefore, there is an urgent need for intelligent control methods that can automatically adjust shot peening parameters based on different iron oxide scale characteristics. Utility Model Content
[0003] The technical problem to be solved by this utility model is to overcome the shortcomings of the existing technology and provide a shot peening and descaling unit for oriented silicon steel plates and strips based on supervised learning, which can achieve rapid and accurate removal of iron oxide scale and avoid damage to the substrate.
[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:
[0005] A shot peening and descaling unit for high-magnetic-induction grain-oriented silicon steel strip based on supervised learning includes a centrifugal shot peening unit, a shot peening control box, an intelligent control system, and sensor devices. The centrifugal shot peening unit is interconnected with the intelligent control system through the shot peening control box, and the sensor devices collect shot peening parameters in real time and transmit them to the intelligent control system. The intelligent control system includes a supervised learning algorithm module, a finite element simulation module, a fault diagnosis module, and an automatic adjustment module, wherein:
[0006] The supervised learning algorithm module receives sensor data and interacts with the finite element simulation module to establish a mapping model between projectile parameters and descaling effect;
[0007] The finite element simulation module provides theoretical support by simulating the cracking mechanism of iron oxide scale;
[0008] The fault diagnosis module monitors the equipment status in real time and triggers alarms;
[0009] The automatic adjustment module dynamically adjusts the shot peening parameters based on the model output.
[0010] The above technical solution describes a centrifugal shot peening unit comprising three shot peening units connected in series. Each unit includes a shot peening cleaning chamber, a shot peening assembly, a bucket elevator, a screw conveyor, and a dust removal system. The shot peening assembly consists of a centrifugal shot peening device and a variable frequency motor, with a shot propulsion speed of 50-100 m / s and an angle of 30°-90°. The dust removal system is connected to the shot peening host via pipelines and includes a dust collector and a variable frequency fan.
[0011] In the above technical solution, the shot material is stainless steel shot with a diameter of 0.2-1.0mm. The first unit uses 0.2-0.5mm fine shot for pretreatment, and the last unit uses 0.8-1.0mm coarse shot for enhanced cleaning.
[0012] In the above technical solution, the shot peening control box integrates a PLC controller, a data acquisition card, and an industrial computer. The PLC controller dynamically adjusts the motor speed and shot flow rate through a PID algorithm, and the data acquisition card converts the signals from the weighing sensor and the laser velocimeter into digital quantities.
[0013] In the above technical solution, the supervised learning algorithm module adopts a deep neural network architecture. The input layer includes projectile velocity, flow rate, and surface image feature values of the plate and strip. The output layer generates optimized parameters for projectile angle, projectile velocity, and pulse duration. The hidden layer corrects the weights using stress distribution data provided by the finite element simulation module.
[0014] In the above technical solution, the automatic adjustment module implements dual closed-loop control. The inner loop adjusts the projectile speed deviation by ±2m / s in real time through PLC, while the outer loop iteratively updates the neural network model parameters every 5 minutes based on the surface cleanliness detection results.
[0015] The above technical solution also includes a data storage module, which records shot peening intensity, equipment vibration spectrum, and historical fault codes in a time-series database, and constructs a predictive maintenance model based on the random forest algorithm.
[0016] The above technical solution includes a shot velocity sensor, a shot flow sensor, and a strip surface image sensor. The shot velocity sensor uses a laser velocimeter or a high-speed camera. The strip surface image sensor uses a CCD camera or a line scan camera, and identifies the grayscale differences in the residual iron oxide scale area through a supervised learning algorithm, achieving a detection accuracy of 0.1 mm. 2 .
[0017] In the above technical solution, the finite element simulation module establishes a two-layer material model of iron oxide scale-matrix, and combines it with the Johnson-Cook constitutive equation to simulate the propagation characteristics of interface stress waves under different projectile incident angles, and outputs the critical peeling energy threshold for the supervised learning module to call.
[0018] A shot peening and descaling process for high magnetic induction oriented silicon steel strip based on supervised learning includes the following steps:
[0019] a) Strip conveying stage: Hot-rolled plates pass through three-stage tandem shot peening chambers at a speed of 0.5-3 m / s;
[0020] b) Parameter optimization stage: Based on the surface morphology after processing by the first unit, the supervised learning algorithm module dynamically adjusts the projectile angle of the last unit to 60°±5°, the projectile velocity to within ±8% of the baseline value of 75-90m / s, and the pulse duration to within ±15% of the baseline value of 0.6-1.2 seconds.
[0021] c) Quality closed-loop control stage: Online detection is performed using a CCD camera or a line array camera. When the iron oxide scale coverage is >0.5%, the shot flow rate is increased by 10%-15%, and the cycle continues until the surface roughness Ra of the substrate is ≤1.6μm.
[0022] After adopting the above technical solution, this utility model has the following positive effects:
[0023] 1) This utility model studies the fracture mechanism of iron oxide scale on the surface of wide hot-rolled plate under external impact by combining finite element simulation with interference theory analysis, and uses supervised learning algorithm to explore the relationship between shot parameters and descaling effect, so as to achieve precise control of shot peening parameters and improve descaling efficiency and accuracy.
[0024] 2) This utility model optimizes the projectile parameters and controls the impact force of the projectile to avoid damaging the silicon steel plate matrix and ensure its magnetic and mechanical properties.
[0025] 3) This utility model introduces a supervised learning algorithm to achieve intelligent control of the shot peening and descaling process, automatically adjusting shot peening parameters according to different iron oxide scale characteristics, thereby improving production efficiency and product quality.
[0026] 4) This utility model reduces manual intervention and lowers production costs through intelligent control and automated operation.
[0027] 5) This utility model monitors the equipment status in real time through a fault diagnosis module, promptly detects and handles faults, and improves the reliability and stability of the equipment. Attached Figure Description
[0028] To make the content of this utility model easier to understand, the present utility model will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein...
[0029] Figure 1 This is a schematic diagram of the overall structural design of this utility model;
[0030] Figure 2 This is a schematic diagram of the overall structural relationship of this utility model;
[0031] Figure 3 Schematic diagram of a single centrifugal shot peening unit
[0032] Figure 4 This is a flowchart illustrating the descaling process of this utility model;
[0033] Figure 5 A flowchart for supervised learning-based modeling;
[0034] Attached reference numerals: 1. Centrifugal shot peening unit; 1-1. Shot peening assembly; 1-2. Dust removal system; 2. Shot peening control box; 3. Intelligent control system; 4. Sensor device; Detailed Implementation
[0035] (Example 1)
[0036] See Figures 1-5 This utility model relates to a shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning. It includes a centrifugal shot peening unit 1, a shot peening control box 2, an intelligent control system 3, and a sensor device 4. The centrifugal shot peening unit 1 is interconnected with the intelligent control system 3 via the shot peening control box 2. The sensor device 4 collects shot peening parameters in real time and transmits them to the intelligent control system 3. The intelligent control system 3 includes a supervised learning algorithm module, a finite element simulation module, a fault diagnosis module, and an automatic adjustment module. Specifically: the supervised learning algorithm module receives sensor data and interacts with the finite element simulation module to establish a mapping model between shot parameters and descaling effect; the finite element simulation module provides theoretical support by simulating the fracture mechanism of iron oxide scale; the fault diagnosis module monitors the equipment status in real time and triggers alarms; and the automatic adjustment module dynamically adjusts the shot peening parameters according to the model output.
[0037] Centrifugal shot peening unit 1 comprises three shot peening units connected in series. Each unit includes a shot peening cleaning chamber, a shot peening assembly 1-1, a bucket elevator, a screw conveyor, and a dust collection system. The shot peening assembly 1-1 consists of a centrifugal shot peening unit and a variable frequency motor, with a shot propulsion speed of 50-100 m / s and an angle of 30°-90°. The shot material is stainless steel shot with a diameter of 0.2-1.0 mm. The first unit uses 0.2-0.5 mm fine shot for pretreatment, and the last unit uses 0.8-1.0 mm coarse shot for enhanced cleaning. The dust collection system 1-2 is connected to the shot peening unit via pipeline and includes a dust collector and a variable frequency fan. The shot peening control box 2 integrates a PLC controller, a data acquisition card, and an industrial computer. The PLC controller dynamically adjusts the motor speed and shot flow rate through a PID algorithm, and the data acquisition card converts signals from the weighing sensor and laser velocimeter into digital quantities.
[0038] The supervised learning algorithm module employs a deep neural network architecture. The input layer includes projectile velocity, flow rate, and surface image features of the strip. The output layer generates optimized parameters for launch angle, launch velocity, and pulse duration. The hidden layer uses stress distribution data provided by the finite element simulation module to correct the weights. The automatic adjustment module implements dual closed-loop control. The inner loop adjusts the launch velocity deviation by ±2m / s in real time via PLC, while the outer loop iteratively updates the neural network model parameters every 5 minutes based on surface cleanliness detection results.
[0039] Preferably, the present invention also includes a data storage module, which records shot peening intensity, equipment vibration spectrum, and historical fault codes in a time-series database, and constructs a predictive maintenance model based on the random forest algorithm.
[0040] Sensor device 4 includes a projectile velocity sensor, a projectile flow sensor, and a strip surface image sensor; the projectile velocity sensor uses a laser velocimeter or a high-speed camera; the strip surface image sensor uses a CCD camera or a line scan camera, and identifies the grayscale differences in the residual iron oxide scale area through a supervised learning algorithm, with a detection accuracy of 0.1 mm. 2 The finite element simulation module establishes a two-layer material model of iron oxide scale and matrix, and combines the Johnson-Cook constitutive equation to simulate the propagation characteristics of interfacial stress waves under different projectile incident angles, outputting the critical peeling energy threshold for use by the supervised learning module.
[0041] This invention optimizes model training to achieve close coordination between the automatic adjustment module, control box, and shot peening device. Based on the output of the supervised learning algorithm module, it automatically adjusts the operating parameters of the shot peening device. The fault diagnosis module interacts with the data storage module and control box to monitor the equipment's operating status in real time, issuing alarms when faults are detected to ensure the stability and safety of equipment operation. The data storage module stores historical and real-time data, providing data support for the supervised learning algorithm module and the fault diagnosis module.
[0042] This utility model relates to a shot peening and descaling process for a high magnetic induction oriented silicon steel sheet and strip based on supervised learning, comprising the following steps:
[0043] 1) Place the hot-rolled plate on the steel belt conveyor system and transport it to the three-stage tandem shot peening chamber. The running speed of the hot-rolled plate is 0.5-3m / s.
[0044] 2) Start the centrifugal shot peening unit and set the initial shot peening parameters, including shot material, shot diameter, shot velocity, shot angle, and shot pulse duration;
[0045] 3) Activate the sensor device to monitor various parameters during the shot peening process in real time;
[0046] 4) Start the shot peening control box to transmit the data collected by the sensor device to the intelligent control system;
[0047] 5) The supervised learning model in the intelligent control system is trained based on historical and real-time data to obtain a mapping relationship model between projectile parameters and descaling effect;
[0048] 6) The supervised learning algorithm module uses the trained model to optimize shot peening parameters, including shot material, shot diameter, projectile velocity, projectile angle, and projectile pulse duration. Based on the surface morphology after the first unit is processed, the supervised learning algorithm module dynamically adjusts the projectile angle of the last unit to 60°±5°, the projectile velocity to within ±8% of the baseline value of 75-90m / s, and the pulse duration to within ±15% of the baseline value of 0.6-1.2 seconds.
[0049] 7) The shot peening control box controls the operation of the centrifugal shot peening unit according to the optimized shot peening parameters;
[0050] 8) After shot peening, use a strip surface image sensor to detect the removal of iron oxide scale. When the iron oxide scale coverage is >0.5%, trigger the shot flow rate to increase by 10%-15%, and cycle until the surface roughness Ra of the substrate is ≤1.6μm.
[0051] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this utility model. It should be understood that the above descriptions are merely specific embodiments of this utility model and are not intended to limit this utility model. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this utility model should be included within the protection scope of this utility model.
Claims
1. A shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning, characterized in that: The system includes a centrifugal shot peening unit (1), a shot peening control box (2), an intelligent control system (3), and a sensor device (4). The centrifugal shot peening unit (1) is interconnected with the intelligent control system (3) through the shot peening control box (2). The sensor device (4) collects shot peening parameters in real time and transmits them to the intelligent control system (3). The intelligent control system (3) includes a supervised learning algorithm module, a finite element simulation module, a fault diagnosis module, and an automatic adjustment module, wherein: The supervised learning algorithm module receives sensor data and interacts with the finite element simulation module to establish a mapping model between projectile parameters and descaling effect; The finite element simulation module provides theoretical support by simulating the cracking mechanism of iron oxide scale; The fault diagnosis module monitors the equipment status in real time and triggers alarms; The automatic adjustment module dynamically adjusts the shot peening parameters based on the model output.
2. The shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning according to claim 1, characterized in that: The centrifugal shot peening unit (1) includes three shot peening units connected in series. Each unit includes a shot peening cleaning chamber, a shot peening assembly (1-1), a bucket elevator, a screw conveyor, and a dust removal system (1-2). The shot peening assembly (1-1) consists of a centrifugal shot peening machine and a variable frequency motor. The shot blasting speed is 50-100 m / s and the angle is 30°-90°. The dust removal system is connected to the shot peening host through a pipeline and includes a dust collector and a variable frequency fan.
3. The shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning according to claim 2, characterized in that: The shot is made of stainless steel and has a diameter of 0.2-1.0 mm. The first unit uses 0.2-0.5 mm fine shot for pretreatment, and the last unit uses 0.8-1.0 mm coarse shot for enhanced cleaning.
4. The shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning according to claim 1, characterized in that: The shot peening control box (2) integrates a PLC controller, a data acquisition card and an industrial computer. The PLC controller dynamically adjusts the motor speed and shot flow rate through a PID algorithm, and the data acquisition card converts the signals from the weighing sensor and the laser velocimeter into digital quantities.
5. The shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning according to claim 1, characterized in that: The supervised learning algorithm module adopts a deep neural network architecture. The input layer includes projectile velocity, flow rate, and surface image feature values of the plate and strip. The output layer generates optimized parameters for projectile angle, projectile velocity, and pulse duration. The hidden layer corrects the weights using stress distribution data provided by the finite element simulation module.
6. The shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning according to claim 1, characterized in that: The automatic adjustment module implements dual closed-loop control. The inner loop adjusts the ejection speed deviation by ±2m / s in real time via PLC, while the outer loop iteratively updates the neural network model parameters every 5 minutes based on the surface cleanliness detection results.
7. The shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning according to claim 1, characterized in that: It also includes a data storage module to record shot peening intensity, equipment vibration spectrum, and historical fault codes in a time-series database, and to build a predictive maintenance model based on the random forest algorithm.
8. The shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning according to claim 1, characterized in that: The sensor device (4) includes a projectile velocity sensor, a projectile flow sensor, and a strip surface image sensor; the projectile velocity sensor uses a laser velocimeter or a high-speed camera; the strip surface image sensor uses a CCD camera or a line array camera, and identifies the grayscale differences in the residual iron oxide scale area through a supervised learning algorithm, with a detection accuracy of 0.1 mm. 2 .
9. The shot peening and descaling unit for high magnetic induction oriented silicon steel plates and strips based on supervised learning according to claim 1, characterized in that: The finite element simulation module establishes a two-layer material model of iron oxide scale-matrix, and combines it with the Johnson-Cook constitutive equation to simulate the propagation characteristics of interfacial stress waves under different projectile incident angles, and outputs the critical peeling energy threshold for the supervised learning module to call.