A machine vision-based remote controllable metal multi-stage heat treatment device and method

Abstract

The present application relates to the technical field of heat treatment, in particular to a remote controllable metal multi-stage heat treatment device and method based on machine vision, the device comprising a furnace, a visual monitoring unit, a quenching unit and a controller, the furnace comprising a furnace body, heating elements and temperature measuring elements for detecting the temperature of different regions, the visual monitoring unit collecting real-time microscopic images of the sample through an observation window; the quenching unit is located below the furnace and comprises a quenching pool and an automatic unloading mechanism. The controller is electrically connected to the above-mentioned components, used to obtain temperature parameters and adjust the heating power accordingly, and obtain microscopic images, identify the microscopic structure state and its evolution process, compare the microscopic structure state with the quenching threshold to determine the quenching condition, and trigger the quenching instruction when the condition is met. Therefore, the device and method of the present application realize the seamless connection of heating and quenching, and improve the data reproducibility and safety and reliability of heat treatment.

Description

Technical Field

[0001] This invention relates to the field of heat treatment technology, and more specifically, to a remotely controllable multi-stage heat treatment device and method for metals based on machine vision. Background Technology

[0002] In the heat treatment process of metallic materials, multi-stage homogenization annealing and quenching are the core means to control the microstructure of alloys, eliminate segregation, and improve mechanical properties. Especially for small-sized metal samples, the quenching cooling rate has a decisive influence on the formation of the final microstructure. Any temperature fluctuation or air cooling during the transfer of the sample at high temperature will directly lead to an irreversible transformation of the non-equilibrium structure, thus seriously affecting the accuracy of subsequent material performance evaluation and the reproducibility of experiments.

[0003] Currently, tubular resistance furnaces or traditional box-type resistance furnaces are commonly used as high-temperature heat treatment equipment for metal samples. These traditional devices employ a closed heating structure, where the heating process, holding point, and degree of phase transformation completion rely entirely on the operator's subjective judgment based on experience. This not only incurs high trial-and-error costs but also lacks in-situ microscopic monitoring capabilities. A more significant technical shortcoming is that the heating and quenching zones are independent. After heat treatment, the small samples must be manually or robotically removed from the furnace, transported, and quenched. The time difference during the transfer process leads to uncontrollable air cooling of the samples. Therefore, it is impossible to observe and automatically identify the microstructure and phase transformation process of the small samples in real time during heating. Furthermore, the lack of a closed-loop feedback mechanism that seamlessly connects the heat treatment and quenching processes spatially and logically results in a heavy reliance on manual intervention, delayed quenching, and distorted microstructure criteria in the experimental process. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a remotely controllable multi-stage heat treatment device and method for metals based on machine vision. This solves the technical challenges of existing equipment being unable to identify the phase transformation process of metal microstructures in situ, and the delayed connection between heating and quenching processes easily leading to air-cooling degradation of small samples.

[0005] On one hand, the present invention provides a remotely controllable multi-stage heat treatment device for metals based on machine vision, comprising: The furnace chamber includes a furnace body, a heating element disposed on the furnace body, and a temperature measuring element disposed within the furnace body. The heating element is used to heat a metal sample disposed within the furnace body, and the temperature measuring element is used to detect the temperature of different areas within the furnace body. The furnace body is provided with a high-temperature resistant, light-transmitting observation window. A visual monitoring unit is installed above the furnace body and is used to acquire real-time images of the sample through the light-transmitting observation window. The quenching unit is located below the furnace body and includes a quenching pool located below the furnace body and an automatic unloading mechanism for transferring the sample from the furnace body to the quenching pool. The controller is electrically connected to the heating element, the temperature measuring element, the visual monitoring unit, and the quenching unit respectively. The controller is used to: automatically adjust the output power of the heating element according to the temperature parameters detected by the temperature measuring element and the set heat treatment process parameters, and determine whether to send a quenching trigger command to the quenching unit according to the microscopic image collected by the visual monitoring unit.

[0006] Preferably, the furnace body includes a furnace body, an upper cover, and a lower cover. The top and bottom of the furnace body are respectively provided with an upper cover opening and a lower cover opening. The light-transmitting observation window is provided on the upper cover, and the upper cover is hinged to the furnace body and used to open or close the upper cover opening. The lower cover is hinged to the furnace body and used to open or close the lower cover opening. The temperature measuring element includes multiple temperature measuring thermocouples, which are respectively arranged at the bottom of the upper cover, the top of the lower cover, the periphery of the upper cover opening, and the periphery of the lower cover opening.

[0007] Preferably, the visual monitoring unit includes a high-temperature resistant microscopic vision module mounted on the upper end of the cover; a ring-shaped cold light source is arranged around the lens of the microscopic vision module; the microscopic vision module is used to align with the sample through the light-transmitting observation window to acquire a microscopic image of the sample. And / or, the automatic unloading mechanism includes a linear drive, the drive end of which is connected to the lower cover via a hinged support, for driving the lower cover to flip up and down through telescopic movement to open or close the lower cover opening.

[0008] Preferably, the lower cover has a boss on the side facing the interior of the chamber, and the boss has a guide groove for placing the sample. The guide groove is used to guide the sample to slide down into the quenching tank when the automatic unloading mechanism drives the lower cover to flip down.

[0009] Preferably, it further includes a wireless communication unit, which is communicatively connected to the controller and is used to interact with the cloud server and the local terminal to receive remote start / stop control commands and modification commands for the heat treatment process parameters, and to remotely push the temperature parameters detected by the temperature measuring element, the microscopic images collected by the visual monitoring unit, and the judgment results of the controller to the cloud server and the local terminal.

[0010] On the other hand, the present invention also provides a machine vision-based multi-stage heat treatment method for metals, using the above-mentioned heat treatment apparatus, comprising the following steps: The metal sample is placed inside the furnace chamber, the heat treatment process parameters are set, and the heating element is activated to heat the sample in a stepped manner. During the heating process, the temperature parameters of different areas inside the furnace body are collected in real time, and the output power of the heating element is automatically adjusted according to the temperature parameters and the set heat treatment process parameters. The microscopic images of the sample are acquired in real time, and a quenching trigger command is generated based on the temperature parameters and the microscopic images. When the quenching trigger command is generated, the automatic unloading mechanism responds to the quenching trigger command, automatically opens the unloading channel, and allows the sample to slide into the quenching pool located below the furnace body, thus completing the automatic quenching.

[0011] Preferably, determining whether to generate a quenching trigger command based on the temperature parameter and the microscopic image includes: Image feature recognition is performed on the microscopic images to identify the state of the microstructure of the sample and its evolution over time in real time. When the temperature parameter reaches the set temperature standard and the state of the microstructure meets the visual judgment standard, the quenching trigger command is generated.

[0012] Preferably, the automatic adjustment of the output power of the heating element according to the temperature parameter and the set heat treatment process parameters includes: The real-time weighted temperature inside the furnace body is determined based on the multi-sensor weighted fusion algorithm and the weighting of the location of the multiple temperature measuring elements. The temperature gradient compensation amount is calculated based on the temperature difference of the multiple temperature measuring elements, and the thermal inertia compensation amount is calculated based on the mass, specific heat capacity and current temperature difference of the sample; wherein, the current temperature difference is the difference between the current temperature inside the furnace body and the set temperature; The output power adjustment of the heating element is calculated based on the real-time weighted temperature, the temperature gradient compensation amount, the thermal inertia compensation amount, and the adaptive PID control algorithm with feedforward compensation, so that the current temperature difference is controlled within the preset temperature difference range.

[0013] Preferably, the temperature compliance condition includes: the real-time weighted temperature reaches the preset target process temperature; the visual judgment compliance condition includes: the percentage of pixel area occupied by the characteristic phase of the microstructure in the microscopic image is lower than a preset percentage threshold.

[0014] Preferably, the step of performing image feature recognition on the microscopic image to identify the microstructure state of the sample and its evolution over time includes: The surface morphology features, color change features, and grain contour evolution features of the sample are extracted from the microscopic images. A convolutional neural network is used to extract the spatial distribution features in the microscopic image corresponding to the surface morphology features, the color change features, and the grain contour features, in order to identify the characteristic phases of the microstructure; The spatial distribution characteristics are learned by using a gated loop unit to study the temporal evolution of the heat treatment process, so as to comprehensively determine whether the sample is in a defective state of surface oxidation or grain boundary coarsening, or whether the characteristic phase has reached the critical point of phase transformation.

[0015] The beneficial technical effects of this invention are as follows: This invention provides a remotely controllable multi-stage heat treatment device for metals based on machine vision. By opening a light-transmitting observation window on the upper cover of the furnace body and installing a vision monitoring unit above the furnace body, the controller can acquire real-time microscopic images of the sample at high temperatures during the heat treatment process. Based on these images, the controller can identify the microstructure state of the sample and its evolution over time, thus solving the technical problems of existing closed heat treatment equipment that cannot observe the evolution of metal microstructure in situ and that phase transformation nodes rely entirely on subjective judgment based on human experience. Simultaneously, a quenching unit is located below the furnace body and includes a quenching pool and an automatic unloading mechanism for transferring the sample from the furnace body to the quenching pool. When the controller determines that the sample has reached the quenching conditions, it sends a quenching trigger command to the quenching unit, and the automatic unloading mechanism quickly transfers the sample into the quenching pool. This achieves seamless integration of the heating and quenching processes, eliminating the problem of irreversible microstructure degradation caused by air cooling during sample transfer from the furnace to the quenching pool in traditional heat treatment equipment. In addition, the controller can not only automatically adjust the heating power of the heating element according to the temperature parameters and the set heat treatment process parameters to ensure the temperature accuracy of the heat treatment process, but also automatically trigger the quenching command according to the recognition results of the microscopic image. This enables the entire multi-stage heat treatment process to be completed without manual intervention, from heating and holding to automatic quenching, thus improving the intelligence level, safety and reliability of metal multi-stage heat treatment experiments and the reproducibility of experimental results.

[0016] This invention provides a remotely controllable multi-stage heat treatment method for metals based on machine vision. By establishing a dual judgment standard of temperature parameters and microscopic images, it provides a reliable logical guarantee for the accurate judgment of the quenching timing in metal heat treatment. The temperature parameter represents whether the current physical-thermodynamic environment of the sample meets the preset process temperature requirements, providing the energy basis for the full evolution of the microstructure. The state of the microstructure and its evolution over time directly reflect the degree of dissolution of characteristic phases and the uniformity of the structure within the sample. Both mutually verify each other; only when the physical-thermodynamic conditions and the microstructure evolution state simultaneously meet the judgment criteria will the controller generate a quenching trigger command. This dual-criteria mechanism fundamentally avoids the misjudgment of quenching timing caused by relying solely on manual experience or single-index control in traditional heat treatment processes, ensuring that the quenching action is accurately executed within the optimal process window. Simultaneously, the deep integration of temperature feedback and visual recognition, relying on the controller's automated decision-making to replace subjective human intervention, improves the reproducibility of experimental data and the accuracy of microstructure evaluation.

[0017] Therefore, the multi-stage heat treatment apparatus and method of the present invention realizes automation from heating, heat preservation, visual monitoring, intelligent judgment to automatic quenching, thereby improving the accuracy, safety and reproducibility of multi-stage heat treatment experiments on metals. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a remotely controllable multi-stage heat treatment device for metal based on machine vision, according to one embodiment of the present invention. Figure 2 This is a schematic diagram of a remotely controllable multi-stage metal heat treatment device based on machine vision, according to another embodiment of the present invention. Figure 3 This is a schematic diagram of a remotely controllable multi-stage metal heat treatment device based on machine vision, without the fixed frame, in one embodiment of the present invention. Figure 4 This is a flowchart of a machine vision-based multi-stage heat treatment method for metals in one embodiment of the present invention.

[0019] Explanation of reference numerals in the attached figures: 11-Furnace body; 111-Upper cover; 112-Light-transmitting observation window; 113-Lower cover; 114-Upper cover opening; 115-Lower cover opening; 1131-Boss; 1132-Guide groove; 12-Temperature measuring element; 21-Quenching pool; 22-Automatic unloading mechanism; 3-Controller; 4-Microscopic vision module; 01-Fixing frame. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0021] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0022] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0023] See Figures 1 to 3 As shown in the figure, a multi-stage metal heat treatment device based on machine vision according to an embodiment of the present invention includes a furnace, a vision monitoring unit, a quenching unit and a controller 3. The furnace includes a furnace body 11, heating elements mounted on the furnace body 11, and temperature measuring elements 12 located within the furnace body 11. The heating elements are used to heat the metal sample located within the furnace body 11, and the temperature measuring elements 12 are used to detect the temperature of different areas within the furnace body 11. The furnace body 11 has a high-temperature resistant, light-transmitting observation window 112. A visual monitoring unit is installed above the furnace body 11 and is used to acquire real-time images of the sample through the light-transmitting observation window 112. A quenching unit is located below the furnace body 11 and includes a quenching pool 21 located below the furnace body 11 and an automatic unloading mechanism 22 for transferring the sample from the furnace body 11 to the quenching pool 21. A controller 3 is electrically connected to the heating elements, temperature measuring elements 12, visual monitoring unit, and quenching unit. The controller 3 is used to: automatically adjust the output power of the heating elements according to the temperature parameters detected by the temperature measuring elements 12 and the set heat treatment process parameters, and determine whether to send a quenching trigger command to the quenching unit based on the microscopic images acquired by the visual monitoring unit.

[0024] Specifically, the heating element can be a resistance wire, which is uniformly embedded in the four side walls of the furnace body 11, the fixed inner bottom plate of the furnace body 11, and the inner side of the top cover 111. To improve heating efficiency and temperature conduction uniformity, the resistance wire is embedded in a ceramic skeleton located inside the insulation layer. The controller 3 divides the resistance wire arranged inside the furnace body 11 into multiple independent heating zones in terms of circuit control. For example, based on the spatial distribution characteristics of the furnace body along the height direction (up and down) and longitudinal and transverse directions (front and back / left and right), the furnace body 11 can be divided into multiple independent heating power control loops.

[0025] The furnace body 11 and the quenching tank 21 are fixedly installed on the site (such as the ground or test bench) by a fixing frame 01 to maintain the precise vertical relative position between the furnace body 11 and the quenching tank 12, ensuring that the sample can accurately slide into the quenching tank 21 when the automatic unloading mechanism 22 drives the lower cover 113 to open. The controller 3 is integrated into the control cabinet, which is arranged on one side of the fixing frame 01 or adjacent to the furnace body 11. The controller 3 is electrically connected to the heating element, temperature measuring element 12, visual monitoring unit and automatic unloading mechanism 22 of the quenching unit inside the furnace through cables (such as power lines and signal transmission lines) to send control commands to each component and receive their feedback signals, so as to realize the centralized control and coordinated operation of the entire metal multi-stage heat treatment device.

[0026] It should be noted that in this embodiment, by opening a light-transmitting observation window 112 on the upper cover 111 of the furnace body 11 and installing a visual monitoring unit above the furnace body 11, the controller 3 can acquire microscopic images of the sample at high temperatures in real time during the heat treatment process. Based on these microscopic images, the controller can identify the microstructure state of the sample and its evolution over time, thus solving the technical problem that existing closed heat treatment equipment cannot observe the evolution of metal microstructure in situ and that phase transformation nodes rely entirely on subjective judgment based on human experience. Simultaneously, the quenching unit is located below the furnace body 11 and includes a quenching pool 21 and an automatic unloading mechanism 22 for transferring the sample from the furnace body 11 to the quenching pool 21. When the controller 3 determines that the sample has reached the quenching conditions, it sends a quenching trigger command to the quenching unit, and the automatic unloading mechanism 22 quickly transfers the sample to the quenching pool 21. This achieves seamless integration of the heating and quenching processes, eliminating the problem of irreversible degradation of the microstructure caused by air cooling during the transfer of the sample from the furnace to the quenching pool in traditional heat treatment equipment. In addition, the controller 3 has both temperature closed-loop regulation and visual closed-loop control functions. It can not only automatically adjust the heating power of the heating element according to the temperature parameters and the set heat treatment process parameters to ensure the temperature accuracy of the heat treatment process, but also automatically trigger the quenching command according to the recognition results of the microscopic image. This allows the entire multi-stage heat treatment process to be completed without manual intervention, from heating and holding to automatic quenching, thus improving the intelligence level, safety and reliability of metal multi-stage heat treatment experiments and the reproducibility of experimental results.

[0027] In one embodiment of the present invention, the furnace body 11 includes a furnace body, an upper cover 111 and a lower cover 113. The top and bottom of the furnace body are respectively provided with an upper cover opening 114 and a lower cover opening 115. A light-transmitting observation window 112 is provided on the upper cover 111, and the upper cover 111 is hinged to the furnace body and is used to open or close the upper cover opening 114. The lower cover 113 is hinged to the furnace body and is used to open or close the lower cover opening 115. The temperature measuring element 12 includes a plurality of temperature measuring thermocouples, which are respectively arranged at the bottom of the upper cover 111, the top of the lower cover 113, the periphery of the upper cover opening 114 and the periphery of the lower cover opening 115.

[0028] It should be noted that in this embodiment, the visual monitoring unit is fixedly installed on the upper cover 111, allowing the visual monitoring unit to open and close synchronously with the upper cover 111. When the operator needs to open the upper cover 111 to handle samples or clean the inside of the furnace, the visual monitoring unit and its matching annular cold light source will flip upwards along with the upper cover 111, completely exiting the upper opening area of ​​the furnace. This avoids the spatial obstruction caused by the fixed visual equipment interfering with the furnace opening during the opening operation, ensuring sufficient operating space and avoiding the risk of the lens being damaged by bumps during frequent operations.

[0029] Multiple thermocouples spatially cover the upper, lower, and surrounding areas of the furnace body 11. Unlike traditional heat treatment equipment that relies on single-point or single-area thermocouples, which cannot accurately reflect the three-dimensional temperature field inside the furnace body and lead to large temperature control feedback deviations and uneven sample heating, this system fully integrates thermal state information from different spatial regions. This improves the accuracy and representativeness of the temperature feedback signal, providing reliable basic data support for subsequent high-precision closed-loop temperature control. It avoids inconsistent sample thermodynamic responses caused by local temperature measurement deviations, and provides hardware and underlying algorithm guarantees for the precise execution of multi-stage heat treatment processes.

[0030] In one embodiment of the present invention, the visual monitoring unit includes a high-temperature resistant microscopic vision module 4 mounted on the upper end of the cover 111; a ring-shaped cold light source is arranged around the lens of the microscopic vision module 4; the microscopic vision module 4 is used to align with the sample through the light-transmitting observation window 112 to acquire a microscopic image of the sample.

[0031] It should be noted that in this embodiment, the microscopic vision module 4 can be a water-cooled microscope. By arranging a ring-shaped cold light source around the lens of the microscopic vision module 4, the emitting surface can be uniformly oriented towards the light-transmitting observation window 112. This provides shadowless and high-brightness illumination for the sample surface inside the furnace body 11 under high-temperature conditions, eliminating local glare or shadow blind spots caused by high-temperature radiation. This ensures that the microscopic vision module 4 can clearly and stably acquire microscopic images of the sample through the high-temperature resistant light-transmitting observation window 112. More importantly, it provides the controller 3 with high-quality, consistent image data that is not affected by external stray light, based on the microscopic image recognition of the microstructure state and its evolution process. This ensures the reliability of image recognition and automatic quenching judgment logic from the source of visual acquisition, realizing a closed-loop monitoring system for heat treatment that is directly visible and quantifiable.

[0032] In one embodiment of the present invention, the automatic unloading mechanism 22 includes a linear drive, the drive end of which is connected to the lower cover 113 via a hinged support, for driving the lower cover 113 to flip up and down through telescopic movement to open or close the lower cover opening 1115.

[0033] Specifically, the linear actuator can be an electric push rod, a pneumatic cylinder, or a hydraulic cylinder. The cylinder end (fixed end) of the linear actuator is hinged to a fixed side bracket on one side of the furnace body 11, while its telescopic rod end (driving end) is connected to the lower cover 113 via a hinged support. When the controller 3 issues a quenching trigger command, the linear actuator is immediately powered or pneumatically supplied. Its driving end extends and pushes the lower cover 113 downward through the support, causing the lower cover 113 to flip downward and open with the hinge connecting it to the furnace body 11 as its axis. As the tilt angle of the lower cover 113 gradually increases, the metal sample, which was originally placed horizontally inside the furnace body 11, loses its horizontal support. Under its own gravity, it slides downward along the inclined surface of the lower cover 113 against friction and eventually detaches from the edge of the lower cover 113, falling directly into the quenching pool 21 directly below, completing the quenching transfer.

[0034] This embodiment employs a gravity-assisted unloading structure with a linear actuator and a hinged support to drive the lower cover to flip downwards and open. The unloading drive mechanism is directly integrated into the lower cover 113 below the furnace body 11. This compact structure reduces the displacement and time required for the sample to travel from the furnace interior to the quenching tank 21. The linear actuator responds quickly upon receiving the quenching trigger command, ensuring the sample completes the transfer from the high-temperature environment to the low-temperature quenching medium in a short time. This preserves the sample's true microstructure at high temperatures, avoiding irreversible changes in the non-equilibrium microstructure caused by excessively long transfer times. This improves the accuracy of microstructure determination and the reliability of experimental results in metal heat treatment experiments.

[0035] In one embodiment of the present invention, the lower cover 113 is provided with a boss 1131 on the side facing the interior of the chamber. The boss 1131 is provided with a guide groove 1132 for placing the sample. The guide groove 1132 is used to guide the sample to slide down into the quenching tank 21 when the automatic unloading mechanism 22 drives the lower cover 113 to flip downward.

[0036] It should be noted that this embodiment provides a positioning and orientation guidance structure for small samples during the heating and holding stage and the quenching and sliding stage by setting a boss 1131 and a guide groove 1132 on the side of the lower cover 113 facing the interior of the furnace. During the heating and multi-stage holding process, the sample is contained in the guide groove 1132, and the groove wall and bottom of the guide groove 1132 form a limit, which can prevent the sample from drifting due to airflow disturbance, mechanical vibration or thermal expansion and contraction of the sample itself during long-term high-temperature heat treatment. This ensures that the sample is always stably located in the clear focusing field of view and effective heating area of ​​the visual monitoring unit, ensuring the continuity and consistency of microscopic image acquisition and the uniformity of heat treatment. During the quenching stage, when the lower cover 113 is driven to flip downward by the linear drive, the groove wall of the guide groove 1132 forms a relatively fixed sliding guide surface, overcoming the risk of random rolling or throwing caused by the sample's center of gravity offset or the change of the tilt angle of the lower cover 113. It forces the sample to slide quickly away from the lower cover 113 and into the quenching tank 21 in a predetermined direction. This guiding structure not only shortens the time for the sample to be transferred from the furnace to the quenching pool 21, but also ensures that the sample's sliding trajectory, water entry posture and impact angle are highly consistent in each experiment. This minimizes the difference in cooling rate caused by manual transfer or mechanical throwing, and ensures the authenticity of the microstructure after quenching and the reproducibility of the experimental results.

[0037] In one embodiment of the present invention, a wireless communication unit is further included. The wireless communication unit is communicatively connected to the controller 3 and is used to interact with the cloud server and the local terminal to receive remote start / stop control commands and modification commands for heat treatment process parameters, and to remotely push the temperature parameters detected by the temperature measuring element 12, the microscopic images collected by the visual monitoring unit, and the judgment results of the controller 3 to the cloud server and the local terminal.

[0038] Specifically, the wireless communication unit integrates a WiFi module or a 4G / 5G wireless communication module, establishing a bidirectional data link with the controller 3 via an industrial standard communication interface. The wireless communication unit is used to upload data such as the temperature data calculated in real-time by the controller 3, the microscopic image frames acquired by the visual monitoring unit, the recognition results generated by the controller 3, the power adjustment parameters of the heating element, and the quenching trigger command records to a cloud server for cloud storage and backup in real time via a wireless network. Simultaneously, the wireless communication unit is also used to receive control commands from the cloud server or a local terminal APP. These control commands include online modification commands for heat treatment process parameters (such as multi-stage stepped heating temperature and time), remote adjustment commands for quenching judgment thresholds, and remote start / stop control commands for the equipment. Through the wireless communication unit, operators can remotely monitor and control the device in real-time from any location far from the experimental site using the cloud server or a local terminal APP.

[0039] Furthermore, the controller 3 also incorporates a built-in full-process data storage and traceability unit, used to automatically record the temperature data of the temperature measuring element 12, the microscopic image frames acquired by the visual monitoring unit, the identification results and adjustment parameters of the controller 3, and generate a unique experiment number to achieve data association storage and full-process traceability. The controller 3 also includes a safety monitoring module, which operates independently of the main control logic. It compares the feedback value of the temperature measuring element 12 with the preset over-temperature threshold in real time, and continuously monitors the signal integrity of each temperature measuring element 12 and the feedback of the furnace airtightness sensor. Once over-temperature, thermocouple failure (loss or short circuit of a thermocouple signal), or airtightness abnormality (such as a sudden drop in protective gas pressure) occurs, the safety monitoring module will directly bypass the conventional control process and send a hardware-level interrupt signal to the main control circuit within milliseconds, forcibly cutting off the heating power supply and simultaneously triggering the audible and visual alarm device.

[0040] See Figure 4 As shown, embodiments of the present invention also provide a remotely controllable metal heat treatment method based on machine vision, using the above-described heat treatment apparatus, comprising the following steps: S1: Place the metal sample inside the furnace body 11, set the heat treatment process parameters, and start the heating element to heat the sample in a stepped manner. S2: During the heating process, the temperature parameters of multiple temperature measuring elements 12 in different areas of the furnace body 11 are collected in real time, and the output power of the heating elements is automatically adjusted according to the temperature parameters and the set heat treatment process parameters. S3: Real-time acquisition of microscopic images of the sample, and determination of whether to generate a quenching trigger command based on temperature parameters and microscopic images; S4: When a quenching trigger command is generated, the automatic unloading mechanism 22 responds to the quenching trigger command and automatically opens the unloading channel, allowing the sample to slide into the quenching pool 21 located below the furnace body 11, thus completing the automatic quenching.

[0041] It should be noted that the stepped heating in step S1 avoids the risk of thermal stress concentration and overheating within the material during the heating process, providing a basic technological guarantee for the uniform evolution of the microstructure. Step S2 performs real-time temperature acquisition and closed-loop feedback adjustment for different areas within the furnace body 11, ensuring the uniformity and stability of the three-dimensional thermal field during heating, and providing a reliable thermodynamic environment for clear acquisition of microscopic images. Step S3 uses real-time temperature parameters and real-time microscopic images for comprehensive decision-making, constructing a dual verification logic of thermodynamic physical conditions and microstructure evolution state. Only when the temperature parameters and microscopic images simultaneously meet the corresponding judgment criteria will the controller trigger the generation of a quenching trigger command, fundamentally eliminating the extreme risks of overheating or insufficient quenching caused by misjudgment of a single parameter. In step S4, the automatic unloading mechanism 22 responds to the quenching command and rapidly transfers the sample to the quenching pool 21 located directly below the furnace body 11 by gravity sliding. This completely eliminates the air-cooling degradation phenomenon that easily occurs in small samples in traditional transfer processes, ensuring the authenticity of the quenched structure and the reproducibility of experimental results. This overall process seamlessly connects multi-stage stepped heating, three-dimensional thermal field control, visual recognition, and automatic quenching, realizing unmanned, precise, and safe operation of the entire heat treatment process, and improving the process reliability and data consistency of multi-stage heat treatment experiments for metals.

[0042] In one embodiment of the present invention, determining whether to generate a quenching trigger command based on temperature parameters and microscopic images includes: S31: Perform image feature recognition on microscopic images to identify the state of the microstructure of the sample and its evolution over time in real time. S32: When the temperature parameter reaches the set temperature standard and the microstructure state meets the visual judgment standard, the quenching trigger command is generated.

[0043] It should be noted that this embodiment provides a reliable logical guarantee for the accurate judgment of the quenching timing in metal heat treatment by establishing a dual judgment standard of temperature parameters and microscopic images. The temperature parameter represents whether the current physical and thermodynamic environment of the sample meets the preset process temperature requirements, providing the energy basis for the full evolution of the microstructure. The state of the microstructure and its evolution over time directly reflect the degree of dissolution of characteristic phases (such as undissolved second phases, eutectic phases, etc.) and the uniformity of the microstructure. Both are mutually corroborative and indispensable. Only when both the physical and thermodynamic conditions and the microstructure evolution state simultaneously meet the judgment criteria will the controller generate a quenching trigger command. This dual-criteria mechanism fundamentally avoids the misjudgment of quenching timing caused by relying solely on manual experience or single-index control in traditional heat treatment processes (such as premature quenching due to the temperature being reached but the microstructure not yet fully evolved, or late quenching due to temperature fluctuations but the microstructure being overheated), ensuring that the quenching action is executed accurately within the optimal process window. Meanwhile, this embodiment deeply integrates temperature feedback and visual recognition, relying on the controller's automated decision-making to replace subjective human intervention, which can improve the reproducibility of experimental data and the accuracy of microstructure evaluation, and also effectively eliminate the safety hazards caused by manual supervision during long-term high-temperature annealing experiments, truly realizing unmanned, precise and intelligent closed-loop control of the entire process of multi-stage heat treatment of metals.

[0044] In one embodiment of the present invention, the automatic adjustment of the output power of the heating element according to the temperature parameter and the set heat treatment process parameters includes: S21: Determine the real-time weighted temperature inside the furnace body 11 by assigning weights based on the multi-sensor weighted fusion algorithm and the location of multiple temperature measuring elements 12. S22: Calculate the temperature gradient compensation amount based on the temperature difference of multiple temperature measuring elements 12, and calculate the thermal inertia compensation amount based on the sample mass, specific heat capacity and current temperature difference; wherein, the current temperature difference is the difference between the current temperature inside the furnace body 11 and the set temperature. S23: Calculate the output power adjustment of the heating element based on the real-time weighted temperature, temperature gradient compensation, thermal inertia compensation, and the adaptive PID (proportional-integral-derivative) control algorithm that integrates feedforward compensation, so that the current temperature difference is controlled within the preset temperature difference range.

[0045] Specifically, firstly, controller 3 calculates the real-time weighted temperature (denoted as ) obtained in step S21. Set the target process temperature The current temperature deviation is calculated. This weighted temperature-based deviation e will be directly used as the basis for the subsequent adaptive PID control loop (i.e. , This is a proportionality coefficient used to provide an instantaneous power regulation response based on the current temperature deviation e; This is the integral coefficient, used to eliminate steady-state cumulative deviation during the temperature control process; The input variable (which is the differential coefficient used to suppress sudden temperature changes) is used to ensure the accuracy of macroscopic feedback regulation.

[0046] Secondly, regarding the thermal inertia compensation amount, controller 3 also bases it on the real-time weighted temperature. Thermodynamic feedforward prediction is carried out. Specifically, controller 3 combines the sample mass m and the material specific heat capacity. Based on this, the system thermal efficiency is set. and the predicted compensation time window Combined with thermal inertia compensation coefficient Accurately calculate the dynamic compensation amount for thermal inertia: Traditional PID controllers only respond with a lag after detecting a temperature difference, which can easily lead to temperature overshoot. In contrast, the thermal inertia compensation of this invention is based on the advance prediction of weighted temperature, which can effectively shorten the heating time and suppress overshoot.

[0047] Furthermore, regarding the temperature gradient compensation amount, the controller 3 directly reads the actual temperature values ​​of each area of ​​the multiple temperature measuring elements 12 distributed on the bottom of the upper cover 111, the top of the lower cover 113, the periphery of the upper cover opening 114, and the periphery of the lower cover opening 115 of the furnace body 11. The controller 3 extracts the highest temperature from the current temperature measurement point inside the furnace. and the lowest temperature at the temperature measurement point Calculate the current furnace temperature gradient For example, when the upper zone of the furnace is detected to be... The lower area is When controller 3 determines that there is thermal lag in the lower region, it will direct the gradient compensation power to the heating element in the lower region to actively eliminate the non-uniformity of the spatial thermal field. The formula for calculating the temperature gradient compensation is: ;in This is the gradient compensation coefficient.

[0048] Finally, in step S23, controller 3 will combine the basic adaptive PID control quantity and the external conventional feedforward compensation quantity. Dynamic compensation for thermal inertia and temperature gradient compensation amount The four components are combined and superimposed to form the final composite control model. The specific formula for synthesizing the control quantities is as follows: .in, This is the standard feedforward compensation amount. Among them... It is a model that integrates an adaptive PID control algorithm with feedforward compensation.

[0049] For example, ; ;at this time However, the furnace wall has not yet reached the target temperature, the internal temperature of the sample lags behind, and the furnace body 11 has insufficient heat storage. If only relying on PID (Detecting Low Temperature - Increasing Power - Temperature Exceeding - Decreasing Power) is used, temperature overshoot, slow heating, and temperature fluctuations are likely to occur. Therefore, it is necessary to predict in advance how much heat is needed to raise the furnace and material to the target temperature. Using the above method, the sample is an aluminum alloy, and its mass... Specific heat capacity is Then, the additional heat required to reach the target temperature can be calculated using thermodynamic formulas. Based on this heat prediction, controller 3, according to By increasing the output power of the heating element in advance, the lag caused by thermal inertia is actively compensated before a significant temperature overshoot. In actual control, controller 3 will combine the aforementioned adaptive PID control quantity and conventional feedforward compensation quantity. Temperature gradient compensation amount and dynamic compensation for thermal inertia By superimposing these values, the final composite output in the complete formula is generated. This allows for multiple precise compensations during the multi-stage heating process.

[0050] Therefore, it can eliminate temperature overshoot caused by thermal inertia and three-dimensional temperature difference caused by uneven spatial distribution during metal heat treatment, and control the preset temperature difference range inside the furnace body 11 between -0.5℃ and +0.5℃.

[0051] In one embodiment of the present invention, image feature recognition is performed on the microscopic image to identify the microstructure state of the sample and its evolution over time in real time, including: S311: Extracting surface morphology features, color change features, and grain contour evolution features of samples from microscopic images; S312: A convolutional neural network is used to extract the spatial distribution features in the microscopic image corresponding to the surface morphology features, the color change features, and the grain contour features, in order to identify the characteristic phases of the microstructure; S313: A gated loop unit is used to learn the temporal evolution of the spatial distribution characteristics during the heat treatment process, so as to comprehensively determine whether the sample is in a defective state of surface oxidation or grain boundary coarsening, or whether the characteristic phase has reached the critical point of phase transformation.

[0052] Specifically, in the recognition process of steps S311 to S313 above, the joint working principle of the convolutional neural network and the gated recurrent unit is as follows: Convolutional Neural Networks (CNNs) are the core module for extracting spatial features from microscopic images. Their working principle is not based on direct visual observation, but rather on using a small window called the convolution kernel to slide across the pixel matrix of the microscopic image, performing mathematical convolution operations on each local region to extract the core feature responses in the image. For example, let a local pixel region of a microscopic image be... After passing through a pre-defined convolution kernel During the sliding calculation, the convolution kernel is multiplied and summed with the corresponding elements of the local region of the image, resulting in a local feature response value of -60 for that region. It's important to emphasize that in practical engineering applications, the matrix values ​​in the convolution kernel are not manually set, but are automatically adjusted and optimized through machine learning and training on a large number of microscopic image samples. Through this mathematical convolution mapping, the CNN can automatically find the brightness differences, grain boundaries, morphology, size, and spatial distribution patterns corresponding to the characteristic phase (for aluminum alloy samples) from the grayscale pixels of the microscopic image. After the CNN extracts the spatial features of a single frame of microscopic image, the information from a single frame is insufficient to determine the dynamic evolution of the tissue and is easily affected by high-temperature heat waves. At this point, the gated recurrent unit (GRU) intervenes. The GRU receives the spatial feature outputs of the CNN from multiple consecutive time steps (i.e., multiple frames of microscopic images) and uses its unique gating mechanism to learn the temporal pattern of how the microstructure evolves over time. The output of the GRU ultimately answers two core questions: what is the current spatial state of the tissue, and how does the tissue evolve with increasing heat preservation time? Based on this, controller 3, according to the time-series learning and evolution of GRU, enters the judgment stage. Controller 3 calculates the pixel area ratio (area fraction) of the characteristic phase in the acquired microscopic image and compares this pixel area ratio with a preset target value. When the judgment result is that the area fraction < the target value, controller 3 determines that the characteristic phase of the current microstructure has been fully dissolved and the sample has reached the qualified state of the phase transition critical point.

[0053] Subsequently, controller 3 sends a stop heating command and a quenching trigger command to the corresponding execution drive module. Upon receiving the quenching trigger command, the drive module controls the heating element to stop heating and simultaneously drives the automatic unloading mechanism to slide the sample along the unloading channel into the quenching tank to complete the automatic quenching. The entire process described above, from image acquisition, feature extraction, temporal evolution law learning, logical judgment to hardware execution, is completed automatically under the unified coordination of controller 3, requiring no manual intervention.

[0054] In one embodiment of the present invention, the temperature compliance condition includes: the real-time weighted temperature reaches the preset target process temperature; the visual judgment compliance condition includes: the percentage of pixel area occupied by the characteristic phase of the microstructure in the microscopic image is lower than a preset percentage threshold. Specifically, the microscopic image is the input data carrier acquired by the visual monitoring unit, the microstructure is the objective physical structure inside the sample reflected in the microscopic image, and the characteristic phase is the key target object in the microstructure that is accurately identified and located by the controller 3 (through CNN) and used to characterize the phase transformation process of heat treatment. The controller 3 calculates the pixel area ratio of the characteristic phase in the microscopic image to accurately determine the actual phase transformation compliance status of the current microstructure.

[0055] It should be noted that this embodiment provides a highly reliable automatic quenching determination mechanism for multi-stage heat treatment of metals by setting up a dual verification logic of temperature compliance and visual compliance. The temperature compliance condition ensures that the thermodynamic environment of the sample meets the preset heat treatment process requirements, providing a physical basis for the full evolution of the microstructure. The visual compliance condition directly verifies whether the microstructure inside the sample (such as the second phase of an aluminum alloy sample) has fully dissolved and reached the phase transformation critical point by accurately quantifying the pixel area ratio of characteristic phases in the microscopic image. Only when the above-mentioned physical thermodynamic conditions (temperature compliance) and microstructure state (visual compliance) are simultaneously satisfied does the controller 3 finally determine that the quenching time has arrived. This dual verification logic of physical conditions and microstructure state mutually verifying each other solves the problem of misjudgment of quenching time caused by traditional equipment relying solely on manual experience or single-factor control, and eliminates experimental deviations caused by premature quenching (insufficient microstructure evolution) or delayed quenching (overheating of the microstructure). Meanwhile, since the quenching trigger command is issued only when the above two conditions are met simultaneously, this ensures that the quenching transfer operation of the sample is performed only within the optimal process window, further guaranteeing the authenticity of the microstructure and the reproducibility of the experimental results of the small sample during the rapid quenching process, thereby improving the intelligent control accuracy and operational reliability of the metal multi-stage heat treatment device.

[0056] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A remotely controllable multi-stage heat treatment device for metals based on machine vision, characterized in that, include: The furnace includes a furnace body (11), a heating element disposed on the furnace body (11), and a temperature measuring element (12) disposed inside the furnace body (11). The heating element is used to heat the metal sample disposed inside the furnace body (11), and the temperature measuring element (12) is used to detect the temperature of different areas inside the furnace body (11). The furnace body (11) is provided with a high-temperature resistant light-transmitting observation window (112). A visual monitoring unit is installed above the furnace body (11) and is used to acquire real-time images of the sample through the light-transmitting observation window (112). The quenching unit is located below the furnace body (11) and includes a quenching pool (21) located below the furnace body (11) and an automatic unloading mechanism (22) for transferring the sample from the furnace body (11) to the quenching pool (21). The controller (3) is electrically connected to the heating element, the temperature measuring element (12), the visual monitoring unit and the quenching unit respectively; the controller (3) is used to: automatically adjust the output power of the heating element according to the temperature parameters detected by the temperature measuring element (12) and the set heat treatment process parameters, and determine whether to send a quenching trigger command to the quenching unit according to the microscopic image collected by the visual monitoring unit.

2. The remotely controllable multi-stage heat treatment device for metals based on machine vision according to claim 1, characterized in that, The furnace body (11) includes a furnace body, an upper cover (111) and a lower cover (113). The top and bottom of the furnace body are respectively provided with an upper cover opening (114) and a lower cover opening (115). The light-transmitting observation window (112) is provided on the upper cover (111). The upper cover (111) is hinged to the furnace body and is used to open or close the upper cover opening (114). The lower cover (113) is hinged to the furnace body and is used to open or close the lower cover opening (115). The temperature measuring element (12) includes multiple temperature measuring thermocouples. The multiple temperature measuring thermocouples are respectively arranged at the bottom of the upper cover (111), the top of the lower cover (113), the periphery of the upper cover opening (114) and the periphery of the lower cover opening (115).

3. The remotely controllable multi-stage heat treatment device for metals based on machine vision according to claim 2, characterized in that, The visual monitoring unit includes a high-temperature resistant microscopic vision module (4) installed on the upper end of the cover (111); a ring-shaped cold light source is arranged around the lens of the microscopic vision module (4); the microscopic vision module (4) is used to align the sample through the light-transmitting observation window (112) to acquire a microscopic image of the sample. And / or, the automatic unloading mechanism (22) includes a linear drive, the drive end of which is connected to the lower cover (113) via a hinged support, for driving the lower cover (113) to flip up and down through telescopic movement to open or close the lower cover opening (115).

4. The remotely controllable multi-stage heat treatment device for metals based on machine vision according to claim 2, characterized in that, The lower cover (113) has a boss (1131) on one side facing the interior of the chamber. The boss (1131) has a guide groove (1132) for placing the sample. The guide groove (1132) is used to guide the sample to slide down into the quenching pool (21) when the automatic unloading mechanism (22) drives the lower cover (113) to flip downward.

5. The remotely controllable multi-stage heat treatment device for metals based on machine vision according to claim 1, characterized in that, It also includes a wireless communication unit, which is connected to the controller (3) for interacting with the cloud server and the local terminal to receive remote start / stop control commands and modification commands for the heat treatment process parameters, and remotely push the temperature parameters detected by the temperature measuring element (12), the microscopic images collected by the visual monitoring unit and the judgment results of the controller (3) to the cloud server and the local terminal.

6. A remotely controllable multi-stage heat treatment method for metals based on machine vision, using the heat treatment apparatus as described in any one of claims 1 to 5, characterized in that, Includes the following steps: The metal sample is placed inside the furnace body (11), the heat treatment process parameters are set, and the heating element is started to heat the sample in a stepped manner. During the heating process, the temperature parameters of different areas inside the furnace body (11) are collected in real time, and the output power of the heating element is automatically adjusted according to the temperature parameters and the set heat treatment process parameters. The microscopic images of the sample are acquired in real time, and a quenching trigger command is generated based on the temperature parameters and the microscopic images. When the quenching trigger command is generated, the automatic unloading mechanism (22) responds to the quenching trigger command and automatically opens the unloading channel, so that the sample slides into the quenching pool (21) located below the furnace body (11) to complete the automatic quenching.

7. The remotely controllable multi-stage heat treatment method for metals based on machine vision according to claim 6, characterized in that, The step of determining whether to generate a quenching trigger command based on the temperature parameter and the microscopic image includes: Image feature recognition is performed on the microscopic images to identify the state of the microstructure of the sample and its evolution over time in real time. When the temperature parameter reaches the set temperature standard and the state of the microstructure meets the visual judgment standard, the quenching trigger command is generated.

8. The remotely controllable multi-stage heat treatment method for metals based on machine vision according to claim 7, characterized in that, The automatic adjustment of the output power of the heating element based on the temperature parameters and the set heat treatment process parameters includes: The real-time weighted temperature inside the furnace body (11) is determined by assigning weights based on the multi-sensor weighted fusion algorithm and the location of multiple temperature measuring elements (12). The temperature gradient compensation amount is calculated based on the temperature difference of the multiple temperature measuring elements (12), and the thermal inertia compensation amount is calculated based on the mass, specific heat capacity and current temperature difference of the sample; wherein, the current temperature difference is the difference between the current temperature inside the furnace body (11) and the set temperature. The output power adjustment of the heating element is calculated based on the real-time weighted temperature, the temperature gradient compensation amount, the thermal inertia compensation amount, and the adaptive PID control algorithm with feedforward compensation, so that the current temperature difference is controlled within the preset temperature difference range.

9. The remotely controllable multi-stage heat treatment method for metals based on machine vision according to claim 8, characterized in that, The temperature compliance conditions include: the real-time weighted temperature reaches the preset target process temperature; the visual judgment compliance conditions include: the percentage of pixel area occupied by the characteristic phase of the microstructure in the microscopic image is lower than a preset percentage threshold.

10. The remotely controllable multi-stage heat treatment method for metals based on machine vision according to claim 9, characterized in that, The step of performing image feature recognition on the microscopic image to identify the microstructure state of the sample and its evolution over time includes: The surface morphology features, color change features, and grain contour evolution features of the sample are extracted from the microscopic images. A convolutional neural network is used to extract the spatial distribution features in the microscopic image corresponding to the surface morphology features, the color change features, and the grain contour features, in order to identify the characteristic phases of the microstructure; The spatial distribution characteristics are learned by using a gated loop unit to study the temporal evolution of the heat treatment process, so as to comprehensively determine whether the sample is in a defective state of surface oxidation or grain boundary coarsening, or whether the characteristic phase has reached the critical point of phase transformation.

Images