Machine-made sand particle size and particle shape detection system

Abstract

The invention relates to the technical field of machine-made sand detection, in particular to a machine-made sand granularity and grain shape detection system. A feeding system performs self-adaptive feeding through a flow sensor in combination with a PID (Proportion Integration Differentiation) algorithm; the drying system dynamically adjusts the temperature according to the initial water content, the real-time weight and the drying time of the machine-made sand, a terahertz time-domain spectroscopy system is combined with standard sample data to construct an equation set for water content detection to calculate the water content, and a laser-induced breakdown spectroscopy and machine vision fusion algorithm is adopted for powder content detection. Obtaining the powder content through a support vector machine model; according to particle size and particle shape detection, ultrasonic vibration dispersion is utilized, laser diffraction and X-ray tomography are utilized, and particle size distribution curves, sphericity, angularity and other particle size and particle shape characteristics are output through a deep learning model and a weighted average fusion algorithm; the invention aims to solve the problems of unstable feeding, unreasonable drying and insufficient detection precision and comprehensiveness in machine-made sand detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of manufactured sand detection, and relates to a manufactured sand particle size and shape detection system. Background Art

[0002] As the main sand source in China's construction market, manufactured sand is made from raw materials such as rocks, pebbles, mine waste rocks, and tailings through processes such as soil removal, mechanical crushing, and screening. It is widely used in the construction industry, but its quality varies, and the problem of particle shape is particularly prominent. The particle size and shape of manufactured sand have a significant impact on its performance, directly related to key indicators such as the workability and strength of concrete. Therefore, accurately detecting the particle size and shape of manufactured sand is of great significance for ensuring the quality of manufactured sand and improving the quality of construction projects.

[0003] Currently, the particle shape of manufactured sand is generally poor. During the production process of manufactured sand, affected by factors such as raw material characteristics and crushing processes, most of them are irregular in shape, making it difficult to meet the demand for high-quality sand in the construction industry and reducing the workability and strength of concrete.

[0004] Existing detection methods have defects: Traditional detection methods, such as the bar screen method, are cumbersome to operate, prone to clogging of the screen holes, and the measurement results are rough, unable to comprehensively measure the particle shape of manufactured sand; The flow time method is prone to jamming when measuring coarse particle sizes, interfered by various factors, and has poor accuracy; Digital image processing technology relies on expensive equipment and professional personnel, is easily affected by factors such as light and the placement position of particles, and cannot characterize the particle shape of manufactured sand below 1.18 mm. These methods cannot accurately and comprehensively evaluate the particle shape of manufactured sand.

[0005] Low detection efficiency and inability to monitor in real time: In actual production, existing detection devices cannot achieve integrated detection in stages. Coarse sand and fine sand need to be detected separately, resulting in low efficiency. At the same time, most detection methods cannot online real-time monitor the particle size and shape of manufactured sand, making it difficult to adjust the production process in a timely manner, leading to unqualified products flowing into the market.

[0006] The current main solution is to adopt new detection methods: For example, the patent titled "Precise Characterization Method of Manufactured Sand Particle Shape", application publication number: CN115728211A proposes to divide all-sized manufactured sand into two parts, and use the unit flow time method and sphericity method to characterize the particle shapes of different particle sizes respectively, and then convert them into particle shape indices to achieve precise characterization of the particle shape of manufactured sand. Or, for example, the patent titled "Device for Detecting the Morphology and Particle Size Parameters of Manufactured Sand Using the Separation Method Based on Images", application publication number: CN116448630A, realizes follow-up detection by setting up a staged screening section, a wind guiding device, a drying guiding device, an image recording section, etc., improves the image quality, avoids ghosting, and can perform integrated detection of coarse sand and fine sand in stages. There is also a manufactured sand particle size and shape detector, which integrates systems such as feeding, drying, metering, dispersion, and image processing, can detect multiple parameters, is intelligent in operation, and has a self-cleaning function.

[0007] The precise characterization method of the particle shape of manufactured sand used above combines two detection methods, which can give full play to their respective advantages, accurately characterize the particle shape of the whole particle size range of manufactured sand, and reduce interference factors. However, this method involves many measurement and calculation steps, is relatively complex to operate, and has high professional requirements for operators.

[0008] The device for detecting the morphology and particle size parameters of manufactured sand by the separation method based on images can achieve integrated detection, improve the detection efficiency and accuracy, can adjust the drying and blowing directions, and is applicable to the detection of manufactured sand with different humidities. However, the device structure is complex, the cost is high, and the requirements for maintenance are high. The manufactured sand particle size and shape detector has high measurement accuracy and efficiency, can detect multiple parameters simultaneously, is intelligent in operation, and has a self-cleaning function. However, the equipment price may be high, and it depends on the power supply, and has certain requirements for the use environment. Summary of the Invention

[0009] The present invention provides a manufactured sand particle size and shape detection system to solve the problems of unstable feeding, unreasonable drying, insufficient detection accuracy and comprehensiveness in the detection of manufactured sand.

[0010] In order to solve the above problems, the technical solution adopted by the invention is:

[0011] The manufactured sand particle size and shape detection system includes a feeding system, a drying system, a powder content measurement system, a moisture content measurement system, a dispersion system, a digital image processing system, a data acquisition system, and a machine vision system. The feeding system is used for feeding during the detection of manufactured sand; the drying system is used for drying the manufactured sand.

[0012] The powder content measurement system is used for detecting the powder content rate on the manufactured sand; the dispersion system is used for dispersing the manufactured sand for easy detection; the digital image processing system is used for processing image data; the data acquisition system is used for collecting data in the detection system.

[0013] The working method of the detection system includes the following steps:

[0014] S01 Adaptive feeding: A flow sensor is set on the feeding pipeline to monitor the feeding flow rate Q in real time, and according to the preset standard feeding flow rate Q0, the PID control algorithm is used to adjust the feeding speed.

[0015] S02 Pretreatment: After the manufactured sand enters the drying system, dynamic drying is carried out on the manufactured sand. According to the initial moisture content M0 and the real-time weight M1 of the manufactured sand, combined with the drying time t, through

[0016]

[0017] T0 is the initial drying temperature, m0 is the initial sample weight, k is the temperature adjustment constant, and T(t) dynamically adjusts the drying temperature;

[0018] S03 Moisture content detection: For the manufactured sand samples after pretreatment, they are evenly dispersed through a dispersion system. The terahertz time-domain spectroscopy system emits terahertz pulses. After passing through the manufactured sand samples, the terahertz pulses are received by a detector. The detector converts the received terahertz signals into electrical signals and transmits them to the data acquisition system. The data acquisition system records the time-domain waveforms of the terahertz pulses before and after passing through the manufactured sand samples to obtain the transmission time delay and amplitude attenuation of the terahertz wave;

[0019] The collected data is processed by wavelet denoising, and the propagation time Δt delay and amplitude attenuation A of the terahertz wave in the manufactured sand samples are extracted from the denoised data. Using the prepared standard samples of manufactured sand with known moisture content, with a quantity of n, experiments are carried out. Using the terahertz time-domain spectroscopy system, the propagation time delay Δt of the terahertz wave corresponding to each standard sample is measured one by one i and the amplitude attenuation A i , i = 1, 2,..., n; Substitute the measured data of the standard samples into the system of equations

[0020]

[0021] Solve through this system of equations to determine the values of the model coefficients a0, a1, a2, a3; Substitute the propagation time delay Δt and amplitude attenuation A obtained from the current manufactured sand sample detection into the relationship model, and calculate the moisture content ω = a1Δt + a2A + a3ΔtA + a0 of the manufactured sand sample through calculation.

[0022] S04 Powder content detection: The dried manufactured sand is dispersed by a dispersion system. Using a detection algorithm based on the fusion of laser-induced breakdown spectroscopy and machine vision, the characteristic spectral information of the elements in the manufactured sand is obtained, and the characteristic spectral line intensities I1, I2,... I n are extracted. At the same time, the machine vision system obtains particle images, and through image analysis, the area S and perimeter L of the particles are obtained. A powder content prediction model is established through a support vector machine model for powder content detection, and the decision function is

[0023]

[0024] a i is the Lagrange multiplier, y i is the sample label, K(x i , x) is the kernel function, x i is the training sample, x is the sample to be detected, and b is the bias term; By substituting the data obtained from the above analysis of the current manufactured sand sample into this decision function, the powder content is calculated;

[0025] S05 Particle Size and Shape Detection: Place the manufactured sand sample into an ultrasonic vibration dispersion container, turn on the ultrasonic vibration, and evenly adsorb and load the dispersed manufactured sand sample onto a rotatable sample stage;

[0026] Irradiate a laser beam onto the manufactured sand sample on the sample stage. By measuring the diffraction light intensity distribution after the laser passes through the particle group and using the Mie scattering theory, calculate the particle size distribution of the manufactured sand. At the same time, in the vertical direction of the laser light path, arrange an X-ray tomography device to perform tomographic scanning on the manufactured sand to obtain the cross-sectional image of the manufactured sand, and combine multiple layers of images into a three-dimensional model of the manufactured sand particles through a three-dimensional reconstruction algorithm;

[0027] Input the obtained laser diffraction data and three-dimensional tomography image data into a deep learning model trained with manufactured sand samples. Use the weighted average fusion algorithm to fuse the particle size data obtained by laser diffraction and the particle shape data obtained by tomography, and integrate them into a manufactured sand particle size and shape data set. After being analyzed and processed by the deep learning model, the characteristics of the particle size and shape of the manufactured sand are output.

[0028] The principle of this solution lies in:

[0029] The operation of this manufactured sand particle size and shape detection system is based on a complete and collaborative process. The feeding system monitors the feeding flow through a flow sensor and adaptively adjusts the feeding speed using the PID control algorithm to ensure stable feeding. The drying system dynamically adjusts the drying temperature according to the initial moisture content, real-time weight, and drying time of the manufactured sand to achieve efficient drying. In the moisture content detection link, the terahertz time-domain spectroscopy system emits terahertz pulses, passes through the manufactured sand sample evenly dispersed by the dispersion system, the detector receives the signal and converts it into an electrical signal, and the data acquisition system records the time-domain waveforms before and after the signal. By constructing an equation set from the data obtained from experiments on standard samples with known moisture content, determine the model coefficients, and then calculate the moisture content of the sample to be measured. For the powder content detection, use the laser-induced breakdown spectroscopy and machine vision fusion algorithm to obtain the elemental characteristic spectral information and particle image data of the manufactured sand, and establish a powder content prediction model using the support vector machine model. During particle size and shape detection, calculate the diffraction light intensity distribution after the laser passes through the particle group through the Mie scattering theory to obtain the particle size distribution, the X-ray tomography device obtains the cross-sectional image and performs three-dimensional reconstruction, and then input the laser diffraction and tomography data into the deep learning model, integrate the data through the weighted average fusion algorithm, and output the particle size and shape characteristics of the manufactured sand

[0030] Compared with the prior art, in the feeding process, adaptive feeding is achieved through a flow sensor and a PID control algorithm, which can accurately control the feeding speed and avoid interference caused by excessive or insufficient feeding to subsequent detections. For example, the traditional feeding method often leads to large fluctuations in detection data due to instability, while this system can ensure the stable progress of the detection process and improve the accuracy of detection data. The drying system dynamically adjusts the drying temperature, which fully considers factors such as the initial moisture content and real-time weight of the manufactured sand. Compared with traditional fixed-temperature drying, it can better adapt to different manufactured sand samples, prevent over-drying or under-drying, improve the drying effect, and provide a more reliable sample state for subsequent moisture content and powder content detections. In the moisture content detection, a method of using a terahertz time-domain spectroscopy system combined with a standard sample with a known moisture content to construct an equation set to determine the model coefficients has higher accuracy compared with traditional detection methods, can measure the moisture content of the manufactured sand more accurately, and reduce errors. For the powder content detection, a fusion algorithm of laser-induced breakdown spectroscopy and machine vision, as well as a support vector machine model, can predict the powder content by integrating various data compared with a single detection method, improving the accuracy and reliability of the detection. For the particle size and shape detection, Mie scattering theory, X-ray tomography imaging, and a deep learning model are combined, and the data is integrated through a weighted average fusion algorithm. Compared with traditional methods for single particle size or shape detection, it can comprehensively and accurately obtain the particle size and shape characteristics of the manufactured sand.

[0031] Further, the flow sensor of the feeding system in S01 is an electromagnetic flow sensor.

[0032] Further, in S02, the detector equipped with the terahertz time-domain spectroscopy system is a mercury cadmium telluride detector, and the response time is less than 1 picosecond.

[0033] Further, in S04, the frequency is set at 30 - 50 kHz.

[0034] Further, in the particle size and shape detection of S05, the ultrasonic vibration dispersion container uses a device with an automatic frequency adjustment function, which can automatically search for the dispersion frequency within the range of 30 - 50 kHz when the characteristics of the manufactured sand sample change.

[0035] Further, the feature extraction in S05 includes particle size distribution characteristics, the major axis, minor axis, surface area, and volume of the particles.

[0036] Further, in S05, the laser beam is a polarized laser beam.

[0037] Further, in S05, the features of the particle size and shape of the manufactured sand output include the particle size distribution curve, sphericity, and angularity. Description of the Drawings

[0038] Figure 1 It is the structural flowchart of the present invention. Detailed Embodiments

[0039] Example 1, as Figure 1 shown, the detection system for the particle size and shape of manufactured sand includes a feeding system, a drying system, a powder content measurement system, a moisture content measurement system, a dispersion system, a digital image processing system, a data acquisition system, and a machine vision system. The feeding system is used to feed the manufactured sand for detection; the drying system is used to dry the manufactured sand;

[0040] The powder content measurement system is used to detect the powder content rate on the manufactured sand; the dispersion system is used to disperse the manufactured sand for easy detection; the digital image processing system is used to process the image data; the data acquisition system is used to collect the data in the detection system;

[0041] The working method of the detection system includes the following steps:

[0042] S01 Adaptive feeding: A flow sensor is set on the feeding pipeline to real-time monitor the feeding flow rate Q, and according to the preset standard feeding flow rate Q0, the PID control algorithm is used to adjust the feeding speed;

[0043] S02 Pretreatment: After the manufactured sand enters the drying system, dynamic drying is carried out on the manufactured sand. According to the initial moisture content M0 and the real-time weight M1 of the manufactured sand, combined with the drying time t, through

[0044]

[0045] where T0 is the initial drying temperature, m0 is the initial sample weight, k is the temperature adjustment constant, and T(t) dynamically adjusts the drying temperature;

[0046] S03 Moisture content detection: After the manufactured sand sample undergoes pretreatment, it is evenly dispersed through the dispersion system. The terahertz time-domain spectroscopy system emits terahertz pulses. After passing through the manufactured sand sample, the terahertz pulses are received by the detector. The detector converts the received terahertz signal into an electrical signal and transmits it to the data acquisition system. The data acquisition system records the time-domain waveforms of the terahertz pulses before and after passing through the manufactured sand sample, and obtains the transmission time delay and amplitude attenuation of the terahertz wave;

[0047] The collected data is processed by wavelet denoising, and the propagation time Δt delay and amplitude attenuation A of the terahertz wave in the manufactured sand sample are extracted from the denoised data. Using the prepared standard samples of manufactured sand with known moisture content, the number of which is n, experiments are carried out. Using the terahertz time-domain spectroscopy system, the propagation time delay Δt i and amplitude attenuation A i of each standard sample are measured one by one, i = 1, 2,..., n; the measured standard sample data is substituted into the system of equations

[0048]

[0049] Solve through this system of equations to determine the values of the model coefficients a0, a1, a2, and a3; substitute the propagation time delay Δt and amplitude attenuation A obtained from the current test of the manufactured sand sample into the relational model, and calculate the moisture content ω of the manufactured sand sample through calculation: ω = a1Δt + a2A + a3ΔtA + a0.

[0050] S04 Powder Content Detection: The dried manufactured sand is dispersed by a dispersion system. An inspection algorithm based on the fusion of laser-induced breakdown spectroscopy and machine vision is used to obtain the characteristic spectral information of the elements in the manufactured sand, and the characteristic spectral line intensities I1, I2,... I n are obtained. At the same time, the machine vision system obtains particle images, and through image analysis, the area S and perimeter L of the particles are obtained. A powder content prediction model is established through a support vector machine model for powder content detection, and the decision function is

[0051]

[0052] a i is the Lagrange multiplier, y i is the sample label, K(x i , x) is the kernel function, x i is the training sample, x is the sample to be detected, and b is the bias term; substitute the data obtained from the above analysis of the current manufactured sand sample into this decision function to calculate the powder content;

[0053] S05 Particle Size and Shape Detection: Put the manufactured sand sample into an ultrasonic vibration dispersion container, turn on the ultrasonic vibration, and evenly adsorb and load the dispersed manufactured sand sample onto a rotatable sample stage;

[0054] Irradiate a laser beam onto the manufactured sand sample on the sample stage. By measuring the diffraction light intensity distribution after the laser passes through the particle group, use Mie scattering theory to calculate the particle size distribution of the manufactured sand. At the same time, in the vertical direction of the laser light path, arrange an X-ray tomography device to perform tomographic scanning on the manufactured sand to obtain the cross-sectional image of the manufactured sand, and combine multiple layers of images into a three-dimensional model of the manufactured sand particles through a three-dimensional reconstruction algorithm;

[0055] Input the obtained laser diffraction data and three-dimensional tomography image data into a deep learning model trained with manufactured sand samples. Use the weighted average fusion algorithm to fuse the particle size data obtained by laser diffraction and the particle shape data obtained by tomography, and integrate them into a manufactured sand particle size and shape data set. After analysis and processing by the deep learning model, the characteristics of the manufactured sand particle size and shape are output.

[0056] The feeding system monitors the feeding flow through a flow sensor and adaptively adjusts the feeding speed using the PID control algorithm to ensure stable feeding. The drying system dynamically adjusts the drying temperature based on the initial moisture content, real-time weight, and drying time of the manufactured sand to achieve efficient drying. In the moisture content detection section, the terahertz time-domain spectroscopy system emits terahertz pulses, which pass through the manufactured sand sample evenly dispersed by the dispersion system. The detector receives the signal and converts it into an electrical signal. The data acquisition system records the time-domain waveforms before and after the signal. By constructing an equation system from the data obtained from experiments on standard samples with known moisture content, the model coefficients are determined, and then the moisture content of the sample to be measured is calculated. For the powder content detection, an algorithm that fuses laser-induced breakdown spectroscopy and machine vision is used to obtain the elemental characteristic spectral information and particle image data of the manufactured sand. A powder content prediction model is established using a support vector machine model. When detecting the particle size and shape, the diffraction light intensity distribution of the laser passing through the particle group is calculated through the Mie scattering theory to obtain the particle size distribution. The X-ray tomography device acquires cross-sectional images and performs three-dimensional reconstruction. Then, the laser diffraction and tomography data are input into a deep learning model, and the data are integrated through a weighted average fusion algorithm to output the particle size and shape characteristics of the manufactured sand.

[0057] Compared with the prior art, in the feeding process, adaptive feeding is achieved through a flow sensor and the PID control algorithm, which can accurately control the feeding speed and avoid interference with subsequent detections caused by excessive or insufficient feeding. For example, the traditional feeding method often leads to large fluctuations in detection data due to instability, while this system can ensure the stable progress of the detection process and improve the accuracy of detection data. The drying system dynamically adjusts the drying temperature, fully considering factors such as the initial moisture content and real-time weight of the manufactured sand. Compared with traditional fixed-temperature drying, it can better adapt to different manufactured sand samples, prevent over-drying or under-drying, improve the drying effect, and provide a more reliable sample state for subsequent moisture content and powder content detections. In the moisture content detection, the method of using the terahertz time-domain spectroscopy system combined with standard samples with known moisture content to construct an equation system to determine the model coefficients has higher accuracy compared with traditional detection methods, can measure the moisture content of the manufactured sand more accurately, and reduce errors. For the powder content detection, an algorithm that fuses laser-induced breakdown spectroscopy and machine vision, as well as a support vector machine model, can predict the powder content by integrating various data compared with single detection methods, improving the accuracy and reliability of detection. For the particle size and shape detection, the combination of the Mie scattering theory, X-ray tomography, and deep learning model is used, and the data are integrated through a weighted average fusion algorithm. Compared with traditional methods of detecting only the particle size or shape individually, it can comprehensively and accurately obtain the particle size and shape characteristics of the manufactured sand.

[0058] The flow sensor in S01 for the feeding system is an electromagnetic flow sensor. Mechanism sand is a solid granular material. When measuring its flow rate, the electromagnetic flow sensor is not affected by changes in physical properties such as the temperature, pressure, density, and viscosity of the mechanism sand. Different from other types of flow sensors, the electromagnetic flow sensor can stably measure the flow rate of mechanism sand. As long as the mechanism sand can pass through the sensor in a suspended or flowing state in the feeding pipeline, it can be accurately measured, which makes it have good adaptability under the complex working conditions of mechanism sand feeding.

[0059] In S02, the detector equipped with the terahertz time-domain spectroscopy system is a mercury cadmium telluride detector with a response time less than 1 picosecond. When the terahertz time-domain spectroscopy system is used for detecting the moisture content of mechanism sand, after the terahertz pulse passes through the mechanism sand sample, the signal changes very fast. This detector with high time resolution can accurately capture the details of the temporal waveform changes of the terahertz pulse before and after passing through the mechanism sand sample. For example, when the terahertz wave propagates in the mechanism sand, the presence of moisture will cause a transmission time delay and amplitude attenuation, and the detector can quickly record these instantaneous changes, providing a basis for accurately extracting the propagation time delay and amplitude attenuation A subsequently, so that the moisture content of the mechanism sand can be calculated more precisely.

[0060] In S04, the frequency is set at 30 - 50 kHz. When the laser frequency is 30 - 50 kHz, it can provide an appropriate number of high-energy pulses per unit time, making the elements in the mechanism sand easier to be excited, thus generating stronger characteristic spectral information. For example, for trace elements with low content in the mechanism sand, an appropriate laser frequency helps to enhance the intensity of their characteristic spectral lines, facilitating more accurate acquisition of the spectral information of these elements. At the same time, the tumbling and vibration of the particles can expose the surface features at different angles, enabling the images obtained by the machine vision system to present better contrast and more details after processing. These details are very helpful for distinguishing whether the particles are sand or powder and accurately calculating the powder content. For example, there may be differences in surface roughness, shape regularity, etc. between powder particles and sand particles, and they can be better distinguished through clear images.

[0061] In the S05 particle size and shape detection, the ultrasonic vibration dispersion container uses a device with an automatic frequency adjustment function. When the characteristics of the manufactured sand sample change, it can automatically search for the dispersion frequency within the range of 30 - 50 kHz. The characteristics of the manufactured sand sample, such as particle size, density, and particle shape, may vary due to factors such as raw material source and production process. When these characteristics change, the optimal dispersion frequency may also change. For example, coarser manufactured sand particles may require a higher frequency to achieve good dispersion, while finer particles or particles with irregular shapes may be able to achieve uniform dispersion at a lower frequency. The ultrasonic vibration dispersion container with an automatic frequency adjustment function can automatically search for the dispersion frequency suitable for the current characteristics of the manufactured sand sample within the range of 30 - 50 kHz, ensuring that the manufactured sand can be uniformly dispersed. In this way, regardless of how the characteristics of the manufactured sand change, it can provide a uniformly dispersed sample basis for subsequent laser diffraction particle size detection and X-ray tomography particle shape detection, improving the accuracy of the detection results.

[0062] In the S05, the laser beam is a polarized laser beam. When the laser beam is a polarized laser beam, the vibration direction of its electric field vector is specific. This polarization characteristic makes the change in the diffraction light intensity distribution more regular after the laser beam passes through the manufactured sand particle group. For example, according to Mie scattering theory, the intensity difference of the scattered light generated by the interaction between the polarized laser beam and the particles is more obvious in different directions. For particle size detection, this obvious intensity difference helps to more accurately measure the diffraction light intensity distribution after the laser passes through the particle group, thereby improving the accuracy of calculating the particle size distribution of the manufactured sand.

[0063] Compared with a non-polarized laser beam, a polarized laser beam can reduce the interference of background noise and stray light. Due to the consistency of the polarization direction, the stray light components in the scattered light that are inconsistent with the polarization direction can be filtered out or their influence can be reduced. This enables the detection system to obtain a purer signal, enhancing the signal contrast, and thus being able to more clearly distinguish the light intensity change information related to the particle size of the manufactured sand, which is beneficial to improving the accuracy of particle size detection.

[0064] In the S05, the characteristics of the output particle size and shape of the manufactured sand include the particle size distribution curve, sphericity, and angularity. The particle size distribution curve can intuitively show the proportion of particles with different particle sizes in the manufactured sand. For example, in concrete production, a suitable particle size distribution can ensure the workability and strength of the concrete. By comparing the particle size distribution curve with industry standards or product design requirements, producers can quickly determine whether the manufactured sand is qualified. If the curve shows too many coarse or fine particles, the production process can be adjusted in a timely manner to ensure stable product quality and compliance with standards.

[0065] The above are only embodiments of the present invention. Common general knowledge such as specific structures and characteristics known in the art is not described in detail herein. Those of ordinary skill in the art know all the common general knowledge in the technical field to which the invention pertains before the filing date or the priority date, are able to obtain all the prior art in this field, and have the ability to apply conventional experimental means before this date. Those of ordinary skill in the art can, under the inspiration given in this application, combine their own abilities to complete and implement this solution. Some typical well-known structures or well-known methods should not become obstacles for those of ordinary skill in the art to implement this application. It should be noted that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can still be made, and these should also be regarded as the protection scope of the present invention, and these will not affect the implementation effect of the present invention and the practicality of the patent. The protection scope required by this application should be based on the content of its claims, and the specific implementation manners and the like recorded in the specification can be used to interpret the content of the claims.

Claims

1. The machine-made sand particle size and shape detection system is characterized by: It includes feeding system, drying system, powder content metering system, moisture content metering system, dispersing system, digital image processing system, data acquisition system and machine vision system. Feeding system: used for feeding for machine-made sand detection; drying system: used for drying machine-made sand; powder content metering system: used for detecting the powder content on machine-made sand; dispersing system: used for dispersing machine-made sand for easy detection; digital image processing system: used for processing image data; data acquisition system: used for collecting data in the detection system; machine vision acquisition system: used for obtaining particle images; The working method of the detection system comprises the following steps: S01 Adaptive feeding: A flow sensor is installed on the feed pipeline to monitor the feed flow Q in real time, and the feed speed is adjusted using the PID control algorithm according to the preset standard feed flow Q0; S02 pretreatment: After the machine-made sand enters the drying system, it is dynamically dried according to the initial moisture content M0 and real-time weight M1 of the machine-made sand, combined with the drying time t. T0 is the initial drying temperature, m0 is the initial sample weight, k is the temperature adjustment constant, and T(t) dynamically adjusts the drying temperature; S03 moisture content detection: the pre-treated machine-made sand sample is evenly dispersed by the dispersion system, and the terahertz time-domain spectroscopy system emits a terahertz pulse, which passes through the machine-made sand sample and is received by the detector. The detector converts the received terahertz signal into an electrical signal and transmits it to the data acquisition system. The data acquisition system records the time domain waveform of the terahertz pulse before and after passing through the machine-made sand sample, and obtains the transmission time delay and amplitude attenuation of the terahertz wave; The collected data were processed by wavelet denoising, and the propagation time delay Δt and amplitude attenuation A of the terahertz wave in the machine-made sand sample were extracted from the denoised data. The experiment was carried out using n prepared standard samples of machine-made sand with known moisture content. The terahertz time-domain spectroscopy system was used to measure the terahertz wave propagation time delay Δt corresponding to each standard sample one by one. i With amplitude attenuation A i , i = 1, 2, ..., n; Substitute the measured standard sample data into the equation group By solving the equation group, the values ​​of the model coefficients a0, a1, a2, and a3 are determined; the propagation time delay Δt and amplitude attenuation A obtained from the detection of the artificial sand sample are substituted into the relationship model, and the moisture content of the artificial sand sample is calculated to be ω=a1Δt+a2A+a3ΔtA+a0. S04 powder content detection: The dried machine-made sand is dispersed by a dispersion system, and a detection algorithm based on the fusion of laser-induced breakdown spectroscopy and machine vision is used to obtain the characteristic spectrum information of the elements in the machine-made sand and extract the characteristic spectrum line intensities I1, I2, ...I n At the same time, the machine vision system obtains the particle image, and obtains the area S and perimeter L of the particle through image analysis. The powder content prediction model is established through the support vector machine model to detect the powder content. The decision function is a i is the Lagrange multiplier, y i is the sample label, K(x i ,x) is the kernel function, x i is the training sample, x is the sample to be tested, and b is the bias term; the powder content is calculated by substituting the data of the current machine-made sand sample obtained by the above analysis into the decision function; S05 Grain size and shape detection: Place the machine-made sand sample into an ultrasonic vibration dispersion container, turn on the ultrasonic vibration, and evenly adsorb the dispersed machine-made sand sample and load it onto a rotatable sample table; A laser beam is irradiated on the machine-made sand sample on the sample stage. The diffraction light intensity distribution after the laser passes through the particle group is measured, and the particle size distribution of the machine-made sand is calculated using the Mie scattering theory. At the same time, an X-ray tomography device is arranged in the vertical direction of the laser light path to perform tomography on the machine-made sand to obtain the cross-sectional image of the machine-made sand. The multi-layer images are combined into a three-dimensional model of the machine-made sand particles through a three-dimensional reconstruction algorithm; The acquired laser diffraction data and three-dimensional tomography image data are input into a deep learning model trained with machine-made sand samples. The particle size data obtained by laser diffraction and the particle shape data obtained by tomography are fused using a weighted average fusion algorithm to form a machine-made sand particle size and shape dataset. After analysis and processing by the deep learning model, the dataset outputs the characteristics of the machine-made sand particle size and particle shape.

2. The machine-made sand grain size and shape detection system according to claim 1 is characterized in that: The flow sensor in the feeding system in S01 is an electromagnetic flow sensor.

3. The machine-made sand grain size and shape detection system according to claim 1 is characterized in that: In S02, the detector equipped in the terahertz time-domain spectroscopy system is a mercury cadmium telluride detector, and the response time is less than 1 picosecond.

4. The machine-made sand grain size and shape detection system according to claim 1 is characterized in that: In the S04, the frequency is set at 30-50kHz.

5. The machine-made sand grain size and shape detection system according to claim 1 is characterized in that: In the particle size and shape detection, the ultrasonic vibration dispersion container adopts a device with an automatic frequency adjustment function, which can automatically search for the dispersion frequency within the frequency range of 30-50kHz when the characteristics of the machine-made sand sample change.

6. The machine-made sand grain size and shape detection system according to claim 1 is characterized in that: In the above S05, the laser beam is a polarized laser beam.

7. The machine-made sand grain size and shape detection system according to claim 1 is characterized in that: In the S05, the characteristics of the particle size and shape of the machine-made sand are output, including the particle size distribution curve, sphericity, and angularity.

Images