A method for online monitoring of asphalt mixture aggregate gradation

Abstract

The present invention relates to the field of asphalt mixing technology, and in particular to an online monitoring method for aggregate gradation of asphalt mixture. An online monitoring method for aggregate gradation of asphalt mixture, including S1, aggregate pretreatment; S2, real-time analysis; S3, production data acquisition; S4, sample acquisition; S5, input feedforward neural network model; S6, predicting mixing results. The present invention first uses labelme software and Mask RCNN instance segmentation network to grade aggregates, so that the aggregate gradation on the production line can be identified online, and at the same time, the aggregate gradation, mix ratio and the asphalt mixture produced are tracked, and a feedforward neural network model is established to obtain a mapping model between production data and the slump of asphalt mixture. The model is input to a computer terminal, and the slump performance of the asphalt mixture being produced on the production line can be predicted, thereby scientifically guiding production personnel to adjust the production formula and data.

Description

Technical Field

[0001] The invention relates to the technical field of asphalt mixing, and in particular to an online monitoring method for asphalt mixture aggregate gradation. Background Art

[0002] In the production process of asphalt mixture, since sand and stone materials come from different manufacturers and are stored in different locations, aggregate gradation, particle size and other parameters are prone to fluctuations in the actual production process, which in turn causes changes in the flow properties of the asphalt mixture leaving the factory, increasing the difficulty of construction. In addition, insufficient mixing time of the asphalt mixture will also affect the flow properties of the factory material, causing the phenomenon of white material. Too long mixing time wastes energy and affects production efficiency, so the mixing time should not be adjusted.

[0003] At present, the detection method for aggregate gradation is mostly determined by laboratory quality inspectors through sampling and screening tests. However, since random sampling can hardly represent the actual gradation of each batch of aggregates in the entire asphalt mixture production process, multiple sampling tests are required. However, there are still certain fluctuations, and the mixture proportion needs to be fine-tuned based on engineering experience. This adjustment process often runs through the entire production process and is repeated many times, which wastes manpower and the test results are not necessarily accurate. Summary of the invention

[0004] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the description and other drawings of the description.

[0005] The purpose of the present invention is to overcome the above-mentioned shortcomings and provide an online monitoring method for asphalt mixture aggregate gradation. First, labelme software and Mask RCNN instance segmentation network are used to grade aggregates, so that the aggregate gradation on the production line can be identified online. At the same time, the aggregate gradation, mix ratio and asphalt mixture produced are tracked, and a feedforward neural network model is established to obtain a mapping model between production data and the slump of asphalt mixture. The model is input into a computer terminal, and the slump performance of the asphalt mixture being produced on the production line can be predicted, thereby scientifically guiding production personnel to adjust the production formula and data to ensure the stability of asphalt mixture output.

[0006] The present invention provides an online monitoring method for asphalt mixture aggregate gradation, comprising:

[0007] S1. Aggregate preprocessing: Take photos of the aggregates to be transported to the mixer, obtain the aggregate images and transmit them to the image processing computer A, preprocess the aggregate images through labelme software, build a Mask RCNN instance segmentation network for the preprocessed aggregate images, learn image features and establish a target detection model;

[0008] S2, real-time analysis: input the target detection model to the image processing computer terminal B that can obtain the aggregate image in real time, analyze the aggregate image obtained in real time, and calculate the gradation of the batch of aggregate;

[0009] S3, production data acquisition: use the current clamp to obtain the current value of the mixer main shaft per second and store it in the PLC, the image processing computer terminal B is connected to the PLC through the first network port to obtain the current characteristics of the mixer main shaft; the image processing computer terminal B is connected to the computer terminal controlling the asphalt mixture production through the second network port to obtain the production proportion of the batch of mixture;

[0010] S4. Sample acquisition: Sampling multiple batches of produced asphalt mixtures that differ only in aggregate grading, classifying the construction performance grades, and labeling each batch of asphalt mixture according to the construction performance grade corresponding to each batch of asphalt mixture;

[0011] S5. Input feedforward neural network model: the production data of each batch of asphalt mixture, the corresponding aggregate gradation and grade label are combined into a one-dimensional array to form sample data, which are input into the feedforward neural network model after normalization. After multiple rounds of iterative training, the loss of the neural network is reduced, and a mapping relationship model between the optimal input feature and the corresponding slump label is obtained;

[0012] S6. Predicting mixing results: The trained feedforward neural network model is input into the image processing computer terminal B, and the aggregate gradation obtained can be analyzed online, so as to predict the slump of each batch of asphalt mixture before leaving the pot.

[0013] In some embodiments, the S1 step includes:

[0014] S11, a camera and a light source are arranged above the conveyor belt. When the aggregate is transported to the mixer via the conveyor belt, the angle of the light source is adjusted so that the camera can obtain an image of the aggregate and transmit it to the image processing computer terminal A;

[0015] S12. On the image processing computer terminal A, the producer uses labelme software to draw the outer contours of the aggregate particles in the aggregate image, add size labels to the aggregate particles in different diameter ranges, and define the aggregate gradation under the aggregate particles of different sizes;

[0016] S13. Input the size information of aggregate gradation into the constructed Mask RCNN instance segmentation network to learn image features and establish a target detection model.

[0017] In some embodiments, in step S2, in the aggregate image acquired by the camera in real time, the object detection model adds category labels and bounding box coordinates to the aggregate particles in the image, and masks the distinguished aggregate particles.

[0018] In some embodiments, in step S5, the feedforward neural network model includes an input layer, a hidden layer and an output layer, the input layer is provided with 32 neurons, the hidden layer is provided with 16 neurons, a batch normalization layer is added between the input layer and the hidden layer, and between the hidden layer and the output layer, and the feedforward neural network model outputs multiple results, and defines the corresponding slump label for each result, thereby realizing a mapping relationship model between the optimal input feature and the corresponding slump label.

[0019] In some embodiments, in step S5, in order to reduce the loss of the neural network and ensure the accuracy of the feedforward neural network model, a loss function is added, the loss function is a mean absolute error function, and the optimizer is an adam algorithm.

[0020] In some embodiments, in step S3, the production data also includes asphalt usage, mineral powder usage, admixture usage, and the average value of stirring current from 30 seconds to 35 seconds.

[0021] In some embodiments, in step S4, the construction performance level includes three levels: good, poor, and unqualified.

[0022] In some embodiments, in step S6, when the predicted slump performance of the asphalt mixture is poor or unqualified, the production data of the batch of asphalt mixture is called up, the aggregate grading and production mix ratio are compared, and precise adjustments are made to ensure that the slump of the asphalt mixture out of the pot meets the on-site construction requirements.

[0023] By adopting the above technical solution, the beneficial effects of the present invention are:

[0024] The present invention first uses labelme software and Mask RCNN instance segmentation network to grade aggregates, so that the aggregate gradation on the production line can be identified online, and the aggregate gradation, mix ratio and asphalt mixture produced are tracked at the same time, and a feedforward neural network model is established to obtain a mapping model between production data and the slump of asphalt mixture. The model is input into a computer terminal, and the slump performance of the asphalt mixture being produced on the production line can be predicted, thereby scientifically guiding production personnel to adjust the production formula and data to ensure the stability of asphalt mixture output.

[0025] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.

[0026] Undoubtedly, these and other objects of the present invention will become more apparent after the following detailed description of the preferred embodiment described with reference to various figures and drawings.

[0027] In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, one or several preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation of the present invention.

[0029] In the drawings, the same reference numerals are used for the same components and the drawings are schematic and not necessarily drawn to scale.

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only one or several embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on such drawings without paying creative work.

[0031] Figure 1 It is a schematic diagram of the overall operation structure of the production line in some embodiments of the present invention;

[0032] Figure 2 It is a schematic diagram of the overall flow of the online monitoring method in some embodiments of the present invention;

[0033] Figure 3 Schematic diagram of aggregate pretreatment process in some embodiments of the present invention;

[0034] Figure 4 Schematic diagram of the hierarchical structure of a feedforward neural network in some embodiments of the present invention.

[0035] Description of main reference numerals:

[0036] 1. Camera; 2. Light source; 3. Aggregate; 4. Conveyor belt; 5. Mixer; 6. Image processing computer end A; 7. Image processing computer end B. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but are not used to limit the present invention.

[0038] In addition, in the description of the present invention, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0039] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral body; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. However, if it is indicated as a direct connection, it means that the two connected bodies are not connected through a transition structure, but are connected to form a whole through a connecting structure. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0040] In the present invention, unless otherwise clearly specified and limited, the first feature "on" or "under" the second feature may be that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in an appropriate manner in any one or more embodiments or examples.

[0041] Reference Figures 1 to 3 , Figure 1 It is a schematic diagram of the overall operation structure of the production line in some embodiments of the present invention; Figure 2 It is a schematic diagram of the overall flow of the online monitoring method in some embodiments of the present invention; Figure 3 Schematic diagram of aggregate pretreatment process in some embodiments of the present invention.

[0042] According to some embodiments of the present invention, the present invention provides an online monitoring method for asphalt mixture aggregate gradation, comprising:

[0043] S1, aggregate preprocessing: take a picture of the aggregate 3 to be transported to the mixer 5, obtain the aggregate image and transmit it to the image processing computer A6, preprocess the aggregate image through labelme software, build a Mask RCNN instance segmentation network for the preprocessed aggregate image, learn image features and establish a target detection model;

[0044] The S1 step includes:

[0045] S11, a camera 1 and a light source 2 are arranged above the conveyor belt 4. When the aggregate 3 is transported to the mixer 5 via the conveyor belt 4, the angle of the light source 2 is adjusted so that the camera 1 can obtain an aggregate image of the aggregate 3 and transmit it to the image processing computer terminal A6;

[0046] S12. On the image processing computer terminal A6, the producer uses labelme software to draw the outer contour of the aggregate particles in the aggregate image, add size labels to the aggregate particles in different diameter ranges, and define the aggregate gradation under aggregate particles of different sizes; specifically, the aggregate gradation is divided into 0-3mm, 3-5mm, 5-10mm, 10-16mm, 16-25mm and above 25mm according to the particle size; among them, the aggregate below 25mm enters the conveyor belt 4 in batches according to the grading to add labels, and the over-limit materials above 25mm do not meet the finished product standards, so when adding labels, the over-limit materials above 25mm are added with over-limit material labels.

[0047] S13, inputting the size information of aggregate gradation into the constructed Mask RCNN instance segmentation network, learning image features and establishing a target detection model;

[0048] Specifically, the acquired aggregate image dataset was randomly divided into training set, validation set, and test set with a ratio of 7:1:2;

[0049] The specific recognition steps of the Mask RCNN instance segmentation network are as follows: first, a target frame is drawn on the acquired aggregate image according to the aggregate particles, and then the aggregate contour in the target frame is identified. Since the size information of the aggregate gradation has been input, the corresponding gradation category can be found for each target, and the recognition of the aggregate particle size is achieved through pixel-level target segmentation. About 2,000 areas to be detected are extracted from the input aggregate image, and the convolutional neural network is used to extract features from these 2,000 areas to be detected. The RoI Pooling layer is used to extract the corresponding features of the area of ​​interest, and then the classification and target frame correction are performed through the fully connected layer;

[0050] When the Mask RCNN network is trained to update the loss function, the momentum factor is set to 0.9, the learning rate is set to 0.002, the weight decay coefficient is set to 0.0001, and the training rounds are set to 30, thus obtaining a mature object detection model;

[0051] S2, real-time analysis: input the target detection model to the image processing computer terminal B7 that can obtain the aggregate image in real time, analyze the aggregate image obtained in real time, and calculate the gradation of the batch of aggregate 3; in the aggregate image obtained in real time by the camera 1, the target detection model adds category labels and bounding box coordinates to the aggregate particles in the image, and masks the distinguished aggregate particles; specifically, in the aggregate particle size identification process, if the proportion of aggregate particles of a certain particle size is greater than 95%, the batch of aggregate is identified as the aggregate gradation of the corresponding size, such as 98% of the aggregate particles in the batch are within the range of 5-10mm, the batch of aggregate is defined as the particle size gradation of 5-10mm; however, the identification of over-limit materials is more special. If the proportion of aggregate particles above 25mm is identified to be greater than 5% during the detection process, the upstream screening process may be wrong in the actual production process, resulting in aggregate particles with too large a particle size entering the finished aggregate area, and an error alarm is sent to remind the staff to check the upstream equipment;

[0052] S3, production data acquisition: use the current clamp to obtain the current value of the main shaft of the mixer 5 per second and store it in the PLC, the image processing computer terminal B7 is connected to the PLC through the first network port to obtain the current characteristics of the main shaft of the mixer 5; the image processing computer terminal B7 is connected to the computer terminal controlling the asphalt mixture production through the second network port to obtain the production proportion of the batch of mixture; wherein, the production data also includes the asphalt dosage, the mineral powder dosage, the admixture dosage, and the average value of the stirring current from the 30th second to the 35th second;

[0053] S4. Sample acquisition: Sampling is performed on multiple batches of produced asphalt mixtures that differ only in aggregate grading, and construction performance grades are divided. According to the construction performance grades corresponding to each batch of asphalt mixture, a grade label is given to each batch of asphalt mixture; the construction performance grades include good, poor, and unqualified;

[0054] For example, during the actual construction process, if the engineering staff finds that the asphalt mixture mixed with aggregate particles with a gradation of 10-16mm has a better construction effect and stronger compressive strength, then its construction performance grade is defined as good, and the production data and aggregate gradation of this batch of asphalt mixture are traced back, and its grade label is marked as level one; if the asphalt mixture mixed with aggregate particles with a gradation of 0-3mm has a poor construction effect and the concrete material cannot be formed and constructed, then its construction performance grade is defined as unqualified, and the production data and aggregate gradation of this batch of asphalt mixture are traced back, and its grade label is marked as level three; accordingly, if the production data of the next batch changes, the construction performance grade will also change, and the above tracing and marking actions are repeated again to generate multiple sets of sample data through production data, aggregate gradation and grade labels;

[0055] S5. Input feedforward neural network model: the production data of each batch of asphalt mixture, the corresponding aggregate gradation and grade label are combined into a one-dimensional array to form sample data, which is input into the feedforward neural network model after normalization. After multiple rounds of iterative training, the loss of the neural network is reduced, and the mapping relationship model between the optimal input feature and the corresponding slump label is obtained. Through the trained feedforward neural network model, it can be known that if an asphalt mixture with a good construction performance grade is to be obtained, the specific corresponding relationship between the production data and the aggregate gradation is obtained, such as when the asphalt dosage increases, the amount of aggregate with a larger particle size gradation is correspondingly reduced, and when the asphalt dosage is reduced, the amount of aggregate with a larger particle size gradation is correspondingly increased, etc.

[0056] S6. Predict mixing results: The trained feedforward neural network model is input into the image processing computer terminal B7, and the aggregate gradation obtained can be analyzed online, so as to predict the slump of each batch of asphalt mixture before it comes out of the pot; when the predicted slump performance of the asphalt mixture is poor or unqualified, the production data of the batch of asphalt mixture is called up, and the aggregate gradation and production mix ratio are compared to make precise adjustments so that the slump of the asphalt mixture out of the pot meets the requirements of on-site construction.

[0057] Reference Figure 4 , Figure 4 Schematic diagram of the hierarchical structure of a feedforward neural network in some embodiments of the present invention.

[0058] According to some embodiments of the present invention, optionally, the feedforward neural network model includes an input layer, a hidden layer and an output layer, the input layer is provided with 32 neurons, the hidden layer is provided with 16 neurons, a batch normalization layer is added between the input layer and the hidden layer, and between the hidden layer and the output layer, the feedforward neural network model outputs multiple results, and a corresponding slump label is defined for each result, thereby realizing a mapping relationship model between the optimal input feature and the corresponding slump label;

[0059] The input layer's input content is a one-dimensional array generated by asphalt dosage, mineral powder dosage, admixture dosage, the average value of stirring current from the 30th to the 35th second, the aggregate gradation of the batch, and the grade label;

[0060] In order to reduce the loss of the neural network and ensure the accuracy of the feedforward neural network model, a loss function is added. The loss function is the mean absolute error function and the optimizer is the Adam algorithm.

[0061] It should be understood that the embodiments disclosed in the present invention are not limited to the specific processing steps or materials disclosed herein, but should be extended to equivalent substitutions of such features understood by ordinary technicians in the relevant field. It should also be understood that the terms used herein are only used for the purpose of describing specific embodiments and are not meant to be limiting.

[0062] The "embodiment" mentioned in the specification means that the specific features or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, phrases or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0063] In addition, the described features or characteristics may be combined in one or more embodiments in any other suitable manner. In the above description, some specific details, such as thickness, quantity, etc., are provided to provide a comprehensive understanding of the embodiments of the present invention. However, those skilled in the relevant art will understand that the present invention can be implemented without one or more of the above specific details or can also be implemented using other methods, components, materials, etc.

Claims

1. An online monitoring method for asphalt mixture aggregate gradation, characterized in that: include S1. Aggregate preprocessing: Take photos of the aggregates to be transported to the mixer, obtain the aggregate images and transmit them to the image processing computer A, preprocess the aggregate images through labelme software, build a Mask RCNN instance segmentation network for the preprocessed aggregate images, learn image features and establish a target detection model; S2, real-time analysis: input the target detection model to the image processing computer terminal B that can obtain the aggregate image in real time, analyze the aggregate image obtained in real time, and calculate the gradation of the batch of aggregate; S3, production data acquisition: use the current clamp to obtain the current value of the mixer main shaft per second and store it in the PLC, the image processing computer terminal B is connected to the PLC through the first network port to obtain the current characteristics of the mixer main shaft; the image processing computer terminal B is connected to the computer terminal controlling the asphalt mixture production through the second network port to obtain the production proportion of the batch of mixture; S4. Sample acquisition: Sampling multiple batches of produced asphalt mixtures that differ only in aggregate grading, classifying the construction performance grades, and labeling each batch of asphalt mixture according to the construction performance grade corresponding to each batch of asphalt mixture; S5. Input feedforward neural network model: the production data of each batch of asphalt mixture, the corresponding aggregate gradation and grade label are combined into a one-dimensional array to form sample data, which are input into the feedforward neural network model after normalization. After multiple rounds of iterative training, the loss of the neural network is reduced, and a mapping relationship model between the optimal input feature and the corresponding slump label is obtained; S6. Predicting the mixing result: inputting the trained feedforward neural network model to the image processing computer terminal B, the aggregate gradation obtained can be analyzed online, so as to predict the slump of each batch of asphalt mixture before leaving the pot; The S1 step includes: S11, a camera and a light source are arranged above the conveyor belt. When the aggregate is transported to the mixer via the conveyor belt, the angle of the light source is adjusted so that the camera can obtain an image of the aggregate and transmit it to the image processing computer terminal A; S12. On the image processing computer terminal A, the producer uses labelme software to draw the outer contours of the aggregate particles in the aggregate image, add size labels to the aggregate particles in different diameter ranges, and define the aggregate gradation under the aggregate particles of different sizes; S13, inputting the size information of aggregate gradation into the constructed Mask RCNN instance segmentation network, learning image features and establishing a target detection model; In step S2, in the aggregate image acquired by the camera in real time, the object detection model adds category labels and bounding box coordinates to the aggregate particles in the image, and masks the distinguished aggregate particles.

2. The method for online monitoring of asphalt mixture aggregate gradation according to claim 1, characterized in that: In step S5, the feedforward neural network model includes an input layer, a hidden layer and an output layer. The input layer is provided with 32 neurons, the hidden layer is provided with 16 neurons, and a batch normalization layer is added between the input layer and the hidden layer, and between the hidden layer and the output layer. The feedforward neural network model outputs multiple results, and a corresponding slump label is defined for each result, thereby realizing a mapping relationship model between the optimal input feature and the corresponding slump label.

3. The method for online monitoring of asphalt mixture aggregate gradation according to claim 1, characterized in that: In step S5, in order to reduce the loss of the neural network and ensure the accuracy of the feedforward neural network model, a loss function is added. The loss function is the mean absolute error function, and the optimizer is the Adam algorithm.

4. The method for online monitoring of asphalt mixture aggregate gradation according to claim 1, characterized in that: In step S3, the production data also includes the asphalt dosage, mineral powder dosage, admixture dosage, and the average value of the stirring current from the 30th second to the 35th second.

5. The method for online monitoring of asphalt mixture aggregate gradation according to claim 1, characterized in that: In step S4, the construction performance levels include good, poor, and unqualified.

6. The method for online monitoring of asphalt mixture aggregate gradation according to claim 1, characterized in that: In step S6, when the predicted slump performance of the asphalt mixture is poor or unqualified, the production data of the batch of asphalt mixture is retrieved, the aggregate gradation and production mix ratio are compared, and precise adjustments are made to ensure that the slump of the asphalt mixture meets the requirements of on-site construction.

Images