Hardware and method for automatically measuring hardness indentation

Abstract

The invention relates to the field of hardness measurement, and discloses hardware and a method for automatically measuring hardness indentations, and the method comprises the following steps: S1, test piece detection; pressure is applied to the test piece through a durometer, and an image sensor is carried on the durometer; an objective lens switching and lighting system of the microscope is controlled by a circuit board, a proper objective lens and lighting brightness are selected according to the size of the indentation and the measurement requirement, and an image sensor is triggered to carry out image acquisition; s3, image processing: processing the image by using a pre-trained convolutional neural network; and S4, hardness calculation: calculating the hardness of the test piece by using a Vickers or Brinell hardness formula. The hardness tester is closely integrated with the circuit board carrying the NPU and the image acquisition system, so that an automatic hardness detection and indentation measurement process is realized, the requirements on hardness detection of a large number of test pieces in industrial production can be quickly met from test piece placement, hardness test to indentation image acquisition, analysis and hardness calculation, the production efficiency and the productivity are effectively improved, and the production cost is reduced. And the detection work can be completed with low cost and high quality without a traditional computer system.

Description

Technical Field

[0001] The present invention relates to the technical field of hardness measurement, and specifically provides a hardware and method for automatically measuring hardness indentations. Background Art

[0002] In the field of materials science and engineering, hardness, as an important mechanical property index of materials, plays a key role in material quality control, product design, and performance evaluation. Traditional hardness measurement methods mainly rely on manual operation of hardness testers to conduct indentation tests, and determine the hardness value by manually reading and calculating the indentation size.

[0003] In Vickers, Brinell, and Knoop hardness tests, the operator needs to manually place the specimen on the hardness tester workbench. After adjusting the position, the hardness tester is started to apply a specific test force, so that the indenter forms an indentation on the specimen surface. Subsequently, tools such as a microscope are used to manually observe the indentation, and measuring tools such as a scale are used to manually measure key dimensions such as the diameter or depth of the indentation, and then the hardness value is calculated according to the corresponding hardness calculation formula. This traditional measurement method has many significant disadvantages:

[0004] These methods usually use image acquisition devices to obtain indentation images, and rely on computer software for image analysis, recognition, and data processing. However, a large amount of time and computer resources are consumed in the calculation process, resulting in a waste of overall resources. Especially when processing high-resolution indentation images, the calculation speed significantly decreases, and real-time or rapid hardness measurement cannot be achieved. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a hardware and method for automatically measuring hardness indentations, which solves the problem of low hardness measurement efficiency.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for automatically measuring hardness indentations includes the following steps:

[0007] S1. Specimen detection; apply pressure to the specimen through a hardness tester. An image sensor with a pixel resolution of 5 million pixels is mounted on the hardness tester, and the image sensor is directly connected to a circuit board, and an NPU is mounted on the circuit board;

[0008] S2. Image uploading; use the circuit board to control the objective lens switching and lighting system of the microscope, select a suitable objective lens and lighting brightness according to the indentation size and measurement requirements, and trigger the image sensor to perform image acquisition;

[0009] S3. Image processing, use a pre-trained convolutional neural network (CNN) to process the image;

[0010] S4. Hardness calculation, using the Vickers or Brinell hardness formula to calculate the hardness of the specimen.

[0011] Preferably, in S1, there is also a power management module on the circuit board to provide stable power for the image sensor and NPU. In addition, the measurement results are transmitted to external devices through the communication interface.

[0012] Preferably, in step S2, the image sensor converts the optical signal into an electrical signal and transmits it to the NPU during the acquisition process.

[0013] Preferably, in step S3, the neural network model convolutional kernel is used to perform weighted summation operations on local regions of the image to generate a feature map.

[0014] Preferably, in step S3, the neural network model pooling layer is used to reduce the dimension of the feature map, reduce the number of parameters and computational complexity, while retaining important feature information. Common pooling operations include max pooling and average pooling. By performing a pooling operation, the size of the feature map is reduced, thereby reducing the amount of data transmitted to the next layer while retaining important features.

[0015] Preferably, in step S3, the stacking of the neural network model convolutional and pooling layers is used to gradually extract higher-level features, integrate the extracted features, and output the final indentation recognition result.

[0016] Preferably, in S4, calculate the diagonal length d of the indentation. The formula is where w and h are the width and height of the indentation. Use a function to calculate the area of the indentation. According to the diagonal length d and the applied force F, use the formula to calculate the Vickers hardness.

[0017] Preferably, in S4, calculate the average diameter D of the Brinell indentation. where S is the circular area of the recognized Brinell indentation. The circular radius r is obtained by S / π, and the circular diameter D is obtained by 2r. Then, through the Brinell hardness calculation formula of 0.102×F / D2, according to the average diameter D of the Brinell indentation and the applied force F, calculate the Brinell hardness.

[0018] A hardware for automatically measuring hardness indentation, characterized by comprising a hardness tester and an image sensor mounted on the hardness tester, and the image sensor is directly connected to the circuit board carrying the NPU.

[0019] The present invention provides a hardware and method for automatically measuring hardness indentation. It has the following beneficial effects:

[0020] By closely integrating a hardness tester with a circuit board equipped with an NPU and an image acquisition system, the present invention realizes an automated hardness detection and indentation measurement process. From specimen placement, hardness testing to indentation image acquisition, analysis, and hardness calculation, the entire process requires little manual intervention, greatly shortening the measurement time for a single specimen. At the same time, when performing indentation measurement, there is no need to add an additional computer, significantly reducing resource waste. It can quickly meet the demand for hardness detection of a large number of specimens in industrial production, effectively improving production efficiency and productivity. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0023] Embodiment:

[0024] Please refer to the attached Figure 1 , the embodiments of the present invention provide a hardware and method for automatically measuring hardness indentations, including the following steps:

[0025] S1. Place the specimen on the platform of the hardness tester, and perform hardness detection on the specimen through the hardness tester. The hardness tester mainly includes an indenter (for applying pressure), a loading mechanism (for controlling pressure application), a measurement system (for detecting indentations), a support structure (for fixing the sample), a display and recording device (for presenting results), a control system (for controlling operation steps), and a calibration device (for ensuring accuracy). After the detection, use a microscopic camera to take a picture of the indentation. Select a high-resolution CCD / CMOS image sensor and directly connect it to the circuit board. The pixel resolution of the image sensor is 5 million pixels (2592×1944), which can resolve the details of the indentation edge at an appropriate magnification, and directly connect the image sensor to the circuit board. The circuit board is equipped with an NPU (neural network processor), and the NPU is responsible for processing the signals collected by the image sensor, accelerating the operation of the neural network algorithm at the hardware level, and being used for tasks such as feature extraction and size measurement of images. There also needs to be a power management module on the circuit board to provide stable power for the image sensor, NPU, and other electronic components. In addition, transmit the measurement results to an external device (such as a display screen) or a storage device through a communication interface, and USB interface, Bluetooth interface, or Wi-Fi interface, etc. can be used.

[0026] S2. Image upload. Use the circuit board to control the objective lens switching and lighting system of the microscope. Select the appropriate objective lens and lighting brightness according to the indentation size and measurement requirements, and trigger the image sensor to collect images. The image sensor converts the optical signal into an electrical signal and transmits it to the NPU.

[0027] S3. Image preprocessing; The NPU uses a pre-trained convolutional neural network (CNN) model to process the image. The CNN model divides the image into several rows and columns in the horizontal and vertical directions, and analyzes the pixel changes in these rows and columns to identify the edge of the indentation. The CNN model will detect the gradient change of each pixel and determine the boundary of the indentation through the gradient change. After the collected image is transmitted to the circuit board equipped with the NPU, the NPU starts to run the pre-trained convolutional neural network model to process the image. First, the image is divided into numerous rows and columns in the horizontal and vertical directions, for example, divided into 200 rows and 200 columns, to construct a fine pixel analysis grid. The model analyzes the pixel changes in these rows and columns point by point to identify the edge of the indentation. When a significant mutation in the pixel value is detected, it is determined that this is the edge of the indentation. To further improve the accuracy of edge detection, the model will continuously repeat the above identification operation, detect and correct the edge pixels of the indentation multiple times. Each time of repeated detection, the model will dynamically adjust the analysis accuracy and range according to the previous detection result, and gradually identify the real edge of the indentation.

[0028] S4. Hardness calculation; For example, Vickers hardness, calculate the diagonal length d of the indentation, and the formula is where w and h are the width and height of the indentation, use a function to calculate the area of the indentation. According to the diagonal length d and the applied force F, use the formula to calculate the Vickers hardness. The applied force comes from the hardness tester data. By using the integrated NPU (Neural Network Processing Unit), it can efficiently process image and model inference tasks, greatly shortening the calculation time. The NPU calculates the hardness value of the specimen through the hardness calculation formula based on the calculated indentation area or diameter and in combination with the load parameters of the hardness tester.

[0029] In one embodiment

[0030] S1. Steadily place the metal specimen to be detected on the working platform of the high-precision hardness tester. The platform is made of special shock-absorbing materials, which can effectively reduce the influence of external vibration on the detection result. The main structure of the hardness tester used in this embodiment includes:

[0031] Indenter: A diamond cone indenter is selected, with a cone angle of 136°, which meets the Vickers hardness test standard, or tungsten carbide steel balls with diameters of 10 / 5 / 2.5 / 1 mm, etc., which meet the Brinell hardness test standard. It can form clear and regular indentations on the surface of metal specimens when applying pressure.

[0032] Loading mechanism: A high-precision electric loading device is adopted, which can precisely control the pressure to gradually increase from 0 to the preset value. The loading accuracy can reach ±0.1 N, ensuring the stability and accuracy of pressure application.

[0033] Measurement system: It is equipped with a high-resolution microscopic imaging module, with a high-resolution CCD image sensor built-in. The pixel resolution is 5 million pixels (2592×1944), and it is directly connected to a special circuit board. This sensor can clearly resolve the details of the indentation edge at a magnification of 100 times, providing high-quality images for subsequent size measurement.

[0034] Calibration device: Regularly calibrate the hardness tester using a standard hardness block to ensure the accuracy of the detection results. During the calibration process, the control system will automatically collect the indentation data of the standard hardness block, compare it with the standard value, and automatically adjust the parameters of the hardness tester to ensure that the measurement error is within the allowable range.

[0035] In addition, a high-performance NPU (neural network processor) is mounted on the circuit board. Its powerful computing power can accelerate the operation of neural network algorithms at the hardware level. At the same time, an efficient power management module is integrated on the circuit board, using switching regulated power supply technology to provide stable 5V and 3.3V power for the image sensor, NPU and other electronic components. The measurement results are quickly transmitted to an external computer or storage device through the USB3.0 interface, facilitating further analysis and processing of the data.

[0036] S2. Image acquisition

[0037] Use the control system to control the objective lens switching and lighting system of the microscope. When detecting metal specimens in different hardness ranges, according to the size of the indentation and measurement requirements, automatically select appropriate objective lenses, such as 50x and 100x objective lenses. At the same time, by adjusting the brightness of the LED lighting system, ensure that the best lighting conditions are obtained at the indentation part. After the settings are completed, trigger the image sensor to perform image acquisition. The image sensor converts the received optical signal into an electrical signal and quickly transmits it to the NPU.

[0038] S3. Image preprocessing

[0039] The NPU processes the collected images by running a pre-trained convolutional neural network (CNN) model. First, the image is divided into 200 rows and 200 columns in the horizontal and vertical directions to construct a fine pixel analysis grid. The CNN model analyzes the pixel changes in these rows and columns point by point, and identifies the edges of the indentation by detecting the gradient changes of each pixel. When a significant mutation in the pixel value is detected, it is determined that this is the edge of the indentation. To further improve the accuracy of edge detection, the model will automatically repeat the above identification operation 3 - 5 times. Each time the detection is repeated, the model will dynamically adjust the analysis accuracy and range according to the previous detection results. For example, it will narrow the pixel analysis range near the edge and improve the analysis accuracy to gradually identify the true and accurate edge of the indentation.

[0040] S4. Hardness calculation

[0041] After accurately identifying the edges of the indentation, the diagonal length of the indentation is calculated through a specific algorithm. Then, a special function is used to calculate the area of the indentation. According to the diagonal length and the force applied by the hardness tester (the force value can be directly obtained from the control system of the hardness tester), the Vickers hardness value of the metal specimen is calculated using the Vickers hardness calculation formula. Due to the integration of a high-performance NPU, it can quickly and efficiently complete complex tasks such as image feature extraction, size measurement, and hardness calculation, greatly shortening the time of the entire detection process. Compared with the traditional manual measurement and calculation methods, the efficiency is increased several times. Finally, the calculated hardness value will be displayed on the display screen and stored in the built-in storage chip and external storage devices for subsequent quality inspection and data analysis.

[0042] The embodiments of the present invention have been described. For those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for automatically measuring hardness indentation, characterized in that: The following steps are involved: S1. Test piece testing; The test piece is subjected to pressure by a hardness tester, which is equipped with an image sensor for taking a picture of the test piece, and the image sensor is directly connected to a circuit board, which is equipped with an AI processor; S2. Upload the image; use the circuit board to control the objective lens switching and lighting system of the microscope, select the appropriate objective lens and lighting brightness according to the indentation size and measurement requirements, and trigger the image sensor to collect the image; S3. Edge recognition: the collected image will be transmitted to the circuit board, and the indentation on the image will be recognized by the AI ​​processor; S4. Hardness calculation, after identifying the edge of the indentation, calculate the area or diameter of the indentation.

2. The method for automatically measuring hardness indentation according to claim 1, characterized in that: The AI ​​processor includes an NPU.

3. The method for automatically measuring hardness indentation according to claim 1, characterized in that: The pixel resolution of the image sensor is 1.3-12 million pixels.

4. The method for automatically measuring hardness indentation according to claim 1, characterized in that: In the step S2, the image sensor converts the optical signal into an electrical signal and transmits the electrical signal to the circuit board.

5. A method for automatically measuring hardness indentation according to claim 1 or 2, characterized in that: The NPU processes the image using a pre-trained convolutional neural network model, which divides the image into a number of rows and columns in the horizontal and vertical directions and analyzes pixel changes in the rows and columns to identify the edge of the indentation; Use the convolution kernel of the neural network model to perform weighted summation operations on the local area of ​​the image to generate a feature map. Use the pooling layer of the neural network model to reduce the dimension of the feature map, reduce the number of parameters and computational complexity, and retain important feature information. Common pooling operations include maximum pooling and average pooling. The size of the feature map is reduced by aggregation operations, thereby reducing the amount of data passed to the next layer while retaining important features. Use the stacking of convolution and pooling layers of the neural network model to gradually extract higher-level features, integrate the extracted features, and output the final indentation recognition results.

6. The method for automatically measuring hardness indentation according to claim 5, characterized in that: The convolutional neural network model detects the gradient change of each pixel and determines the boundary of the indentation through the gradient change.

7. The method for automatically measuring hardness indentation according to claim 5, characterized in that: In order to further improve the accuracy of edge detection, the model will continuously repeat the above recognition operation and perform multiple detection and correction on the edge pixels of the indentation. A hardware for automatically measuring hardness indentation, comprising the method of implementing claims 1-7, characterized in that: It includes a hardness tester and an image sensor mounted on the hardness tester, and the image sensor is directly connected to a circuit board equipped with an NPU.

Images