Crusher state monitoring method based on particle material form change, medium and equipment

Abstract

The invention relates to a particle material form change-based crusher state monitoring method, medium and equipment, and the method comprises the steps: obtaining the historical point cloud data of a particle material in a crusher, marking the state of the crusher at each historical moment, and extracting the form parameters of the particle material according to the historical point cloud data; comprising morphological parameters of all particles, correlation parameters among the particles, morphological parameters of materials composed of the particles and spatial parameters between the materials and a crusher, and according to the crusher state marked by historical point cloud data and the extracted morphological parameters of the particle materials, constructing and training a model which takes the morphological parameters of the particle materials as input and takes the morphological parameters of the particle materials as output; the state of the crusher is an output monitoring model, current point cloud data is obtained, the morphological parameters of the current particle material are extracted and input into the trained monitoring model, and the state of the crusher is obtained. The problems that in the prior art, accuracy is insufficient, the data processing capacity is weak, cost is too high, environment influence is likely to happen, and adaptability is poor are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of morphological change analysis, and particularly to a crusher status monitoring method, medium and device based on the morphological change of granular materials. Background Art

[0002] In the fields of industrial production and material processing, the morphological change of materials has a crucial impact on the final quality of products, production efficiency, and cost control. The morphological change of materials refers to the change in physical form and structure of materials under the action of external forces such as stirring, compression, and shear, or due to environmental changes such as temperature and humidity.

[0003] Traditional methods for material morphological analysis mostly rely on manual experience and simple physical quantity measurements, such as weight and size measurements. However, this method has many drawbacks: firstly, the accuracy is insufficient, manual judgment is easily interfered by subjective factors, and there is a lack of quantitative analysis, making it difficult to accurately describe the morphological change of materials; secondly, real-time monitoring cannot be achieved, mostly static analysis, with obvious lag, which is not conducive to timely problem discovery; thirdly, the data processing ability is weak, manual recording and sorting are prone to errors, and there is a lack of digital support; fourthly, the cost is high, the labor cost is high, and the material processing is inefficient; fifthly, there are safety hazards, on-site inspections are easily affected by adverse environments; sixthly, the adaptability is limited, and it is difficult to effectively analyze special materials; seventhly, the R & D support is insufficient, the data provided is limited, affecting the R & D progress and efficiency.

[0004] Therefore, how to improve it is an urgent technical problem in this field. Summary of the Invention

[0005] Based on this, the purpose of this application is to provide a crusher status monitoring method, medium and device based on the morphological change of granular materials to solve at least one of the technical problems mentioned in the above background art.

[0006] In the first aspect, this application provides a crusher status monitoring method based on the morphological change of granular materials, including:

[0007] Obtain the historical point cloud data of the granular materials in the crusher and label the crusher status at each historical moment;

[0008] According to the historical point cloud data, extract the morphological parameters of the granular materials, including: the morphological parameters of each particle, the correlation parameters between particles, the morphological parameters of the materials composed of particles, and the spatial parameters between the materials and the crusher;

[0009] According to the crusher status labeled by the historical point cloud data and the extracted morphological parameters of the granular materials, construct and train a monitoring model with the morphological parameters of the granular materials as the input and the crusher status as the output;

[0010] Obtain the current point cloud data, extract the morphological parameters of the current granular material, and input them into the trained monitoring model to obtain the crusher status.

[0011] Further, the steps of extracting the morphological parameters of each particle, the correlation parameters between particles, and the morphological parameters of the material composed of particles from the historical point cloud data include:

[0012] Obtain the measurement information of the acquisition device;

[0013] Convert the historical point cloud data according to the measurement information to obtain a three-dimensional data model;

[0014] Extract features from the three-dimensional data model to obtain the morphological parameters of each particle, the correlation parameters between particles, and the morphological parameters of the material composed of particles.

[0015] Further, the morphological parameters of each particle include any one or more of volume, surface area, roundness, and sphericity; the correlation parameters between particles include any one or more of particle spacing, bulk density, flow trajectory, and layered structure; the morphological parameters of the material composed of particles include any one or more of volume, surface area, length, width, and height.

[0016] Further, the steps of extracting the spatial parameters between the material and the crusher from the historical point cloud data include:

[0017] Perform pass-through filtering on the historical point cloud data to obtain local point clouds in several directions and the positions of each local point cloud;

[0018] Select several point clouds from each local point cloud, and obtain the average distance between the selected point clouds in each local point cloud and the corresponding inner wall of the crusher;

[0019] Set weight coefficients for each local point cloud according to the local point cloud position;

[0020] Sum the average distances according to the weight coefficients to obtain the spatial parameters between the material and the crusher.

[0021] Further, the monitoring model includes: a classification unit, a first recognition unit, and a second recognition unit; where:

[0022] The classification unit is used to classify the input data into type-one data and type-two data according to the morphological parameters of each particle and the correlation parameters between particles;

[0023] The first recognition unit is connected to the classification unit and is used to make a prediction based on the type-one data to obtain the crusher status;

[0024] The second recognition unit is connected to the classification unit and is used to make a prediction based on the type-two data to obtain the crusher status.

[0025] Further, the classification unit includes an input layer, a normalization layer, a convolutional layer, a fully connected layer, an output layer, and a decision tree algorithm, where:

[0026] The input layer is used to receive the morphological parameters of each particle and the correlation parameters between particles, and input them into the normalization layer and the decision tree algorithm;

[0027] The normalization layer, the convolutional layer, and the fully connected layer are connected in sequence, and are used to extract features from the morphological parameters of each particle and the correlation parameters between particles, so as to obtain particle morphological features and particle correlation features;

[0028] The decision tree algorithm is used to extract features from the morphological parameters of each particle and the correlation parameters between particles, so as to obtain particle morphological grades and particle correlation grades;

[0029] The output layer is connected to the fully connected layer and the decision tree algorithm, and is used to classify the input data into type-one data and type-two data according to the particle morphological features, particle correlation features, particle morphological grades, and particle correlation grades, and send the type-one data to the first recognition unit and the type-two data to the second recognition unit.

[0030] Further, the structure of the first recognition unit is:

[0031] The first bottleneck layer is used to perform dimensionality reduction on the input data and retain the data feature information to obtain initial dimensionality-reduced data;

[0032] The first CBL convolutional layer is connected to the first bottleneck layer and is used to perform convolution on the initial dimensionality-reduced data to extract local features of the initial dimensionality-reduced data, and obtain 2 feature maps;

[0033] The fully connected layer is connected to the first CBL convolutional layer and is used to flatten the feature maps to obtain several one-dimensional vectors, and further combine and transform each one-dimensional vector to obtain the crusher state.

[0034] Further, the structure of the second recognition unit is:

[0035] The second bottleneck layer is used to perform dimensionality reduction on the input data and retain the data feature information to obtain initial dimensionality-reduced data;

[0036] The second CBL convolutional layer is connected to the second bottleneck layer and is used to perform convolution on the initial dimensionality-reduced data to extract local features of the initial dimensionality-reduced data, and obtain 2 initial feature maps;

[0037] The third bottleneck layer is connected to the second CBL convolutional layer and is used to perform dimensionality reduction on the initial feature maps and retain the data feature information to obtain optimized dimensionality-reduced data;

[0038] The third CBL convolutional layer, connected to the third bottleneck layer, is used to perform a convolutional operation on the optimized dimensionality-reduced data to extract local features of the initial dimensionality-reduced data, obtaining 2 optimized feature maps;

[0039] The spatial pyramid pooling layer, connected to the third CBL convolutional layer, is used to divide the optimized feature maps into multiple sub-regions and perform a pooling operation on each sub-region to obtain a number of feature vectors of a fixed size;

[0040] The fully connected layer, connected to the spatial pyramid pooling layer, is used to further combine and transform each feature vector to obtain the crusher state.

[0041] In a second aspect, the present application also provides a computer storage medium storing executable program code; the executable program code is used to execute the crusher state monitoring method based on the morphological changes of granular materials according to any one of the first aspects.

[0042] In a third aspect, the present application also provides a terminal device including a memory and a processor; the memory stores program code executable by the processor; the program code is used to execute the crusher state monitoring method based on the morphological changes of granular materials according to any one of the first aspects.

[0043] A crusher state monitoring method, medium and device based on the morphological changes of granular materials provided by the present invention first obtains historical point cloud data and labels the crusher state at each historical moment; then extracts granular material morphological parameters from the historical point cloud data, including the morphological parameters of each particle, the correlation parameters between particles, the morphological parameters of the material composed of the particles, and the spatial parameters between the material and the crusher; on this basis, constructs and trains a monitoring model with the granular material morphological parameters as the input and the crusher state as the output; finally obtains the current point cloud data, extracts the current granular material morphological parameters, and inputs them into the trained monitoring model to obtain the crusher state. Among them, on the one hand, the morphological parameters of the particles and the correlation parameters between the particles can reflect the microscopic characteristics of the particles themselves and between the particles from the perspective of individual particles; on the other hand, the morphological parameters of the material composed of the particles and the spatial parameters between the material and the crusher can reflect the macroscopic characteristics of the material itself and between the material and the crusher from the perspective of the material composed of the particles; extracting the intricate morphological laws between the morphological changes of granular materials and the crusher state from both the microscopic and macroscopic levels can improve the efficiency and accuracy of state recognition. It solves the problems of insufficient accuracy, weak data processing ability, excessive cost, susceptibility to the environment and poor adaptability in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS [[ID=##]]

[0044] Figure 1 is a flowchart of the crusher state monitoring method based on the morphological changes of granular materials according to an embodiment of the present invention;

[0045] Figure 2 Schematic diagram of the monitoring model structure according to an embodiment of the present invention;

[0046] Figure 3 Schematic diagram of the classification unit structure according to an embodiment of the present invention;

[0047] Figure 4 Schematic diagram of the first recognition unit structure according to an embodiment of the present invention;

[0048] Figure 5 Schematic diagram of the second recognition unit structure according to an embodiment of the present invention. Detailed implementation manners

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments 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. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] It should be noted that if there are directional indications in the embodiments of the present invention, such as up, down, left, right, front, back..., the directional indications are only used to explain the relative position relationship and movement conditions between components in a specific posture. If the specific posture changes, the directional indications will also change accordingly. In addition, if there are descriptions such as "first, second", "S1, S2", "step one, step two" in the embodiments of the present invention, such descriptions are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features or indicating the execution order of the method, etc. Those skilled in the art can understand that all those that do not violate the inventive points under the inventive concept of the invention should be included in the protection scope of the present invention.

[0051] As Figure 1 shown, the present invention provides a crusher status monitoring method based on the morphological changes of granular materials:

[0052] S1: Obtain the historical point cloud data of the granular materials in the crusher and label the crusher status at each historical moment;

[0053] Specifically, it is optional but not limited to installing the acquisition device at the top end, side wall, etc. of the crusher, acquiring the point cloud data of the granular materials in the crusher at each moment within a set time period, and labeling the crusher status at each historical moment, such as a large amount of material accumulation, irregular material accumulation shape, abnormal increase in the material accumulation height, normal material morphology, etc., as the subsequent training set; the acquisition device includes common acquisition devices such as radar and thermal imaging cameras.

[0054] Preferably, since the acquisition device is required to have high-accuracy measurement capabilities to ensure the accurate acquisition of information such as distance, orientation, speed, and height during material flow, providing a reliable data basis for anomaly detection, and low-frequency radar has stable performance and strong anti-interference ability, and can still work stably under harsh working environments and complex electromagnetic environments, reducing false alarms and missed alarms, therefore, the acquisition device preferably uses low-frequency radar.

[0055] S2: Extract the morphological parameters of granular materials from the historical point cloud data, including: the morphological parameters of each particle, the correlation parameters between particles, the morphological parameters of the material composed of particles, and the spatial parameters between the material and the crusher;

[0056] Specifically, the morphological parameters of each particle refer to the morphological values of the particles themselves, and can optionally but not limited to include any one or more of the volume, surface area, roundness, sphericity, etc. of each particle; the correlation parameters between particles refer to the correlation values between particles, such as between adjacent particles or between a pile of particles, and can optionally but not limited to include any one or more of particle spacing, bulk density, flow trajectory, layered structure, particle size distribution, etc.; the morphological parameters of the material composed of particles refer to the overall morphological values of the material formed by the accumulation of thousands of particles, and can optionally but not limited to include any one or more of the overall volume, surface area, length, width, height, etc. of the material composed of all particles; the spatial parameters between the material and the crusher refer to the spatial values between the material formed by the accumulation of thousands of particles and the crusher as a container, and can optionally but not limited to include the average distance between the material and the crusher, the distances between the material in each direction and the side walls of the crusher, etc.

[0057] Preferably, the steps of extracting the morphological parameters of each particle, the correlation parameters between particles, and the morphological parameters of the material composed of particles from the historical point cloud data can optionally include:

[0058] S21: Obtain the measurement information of the acquisition device;

[0059] S22: Convert the historical point cloud data according to the measurement information to obtain a data three-dimensional model;

[0060] Specifically, it can optionally but not limited to obtain measurement information such as distance, angle, reflection intensity, etc. of the acquisition device when obtaining point cloud data, determine the positions of each point cloud in three-dimensional space according to the distance and angle, obtain the normal vectors of each point cloud according to the angle, determine the point cloud boundary according to the reflection intensity, and perform point cloud registration and noise filtering on the point cloud data using any existing technology to remove noise points, and then convert the point cloud data into a triangular mesh or polygon mesh for surface reconstruction, so as to convert the point cloud data into the corresponding data three-dimensional model.

[0061] S23: Extract features from the three-dimensional data model to obtain the morphological parameters of each particle, the correlation parameters between particles, and the morphological parameters of the material composed of the particles. Specifically, but not limited to, any existing technical solution can be selected to extract features from the three-dimensional data model, and then the morphological parameters of each particle such as volume, surface area, roundness, and sphericity can be obtained, the correlation parameters between particles such as particle spacing, packing density, flow trajectory, and layered structure can be obtained, and the morphological parameters of the material composed of the particles such as volume, surface area, length, width, and height can be obtained.

[0062] For example, when the three-dimensional data model is a triangular mesh model, the volumes of the tetrahedrons formed by each triangular face and the origin are summed to obtain the particle volume. The surface areas of all the triangular faces of each particle are summed to obtain the particle surface area. Then, the boundary points of the particle are fitted, and the deviation from the ideal circle is calculated to obtain the roundness of the particle. The sphericity of each particle is obtained according to the sphericity formula 2-1:

[0063]

[0064] where S is the sphericity of the particle, V is the volume of the model, and A is the surface area of the model. The sphericity S is an index to measure how close an object is to a spherical shape. The closer its value is to 1, the closer the object is to a spherical shape.

[0065] More preferably, the steps of extracting the spatial parameters between the material and the crusher according to the historical point cloud data may include:

[0066] S24: Perform pass-through filtering on the point cloud data to obtain local point clouds in several directions and the positions of each local point cloud.

[0067] Specifically, it is known that the contact surface between the inner wall of the crusher and the material includes five directions: the bottom, the front side, the rear side, the left side, and the right side. Therefore, but not limited to, pass-through filtering can be performed on the point cloud data to obtain the corresponding local point clouds in each direction, including the front left region point cloud, the front right region point cloud, the rear left region point cloud, and the rear right region point cloud. More specifically, the position and range of the pass-through filtering can be arbitrarily set by those skilled in the art.

[0068] S25: Select several point clouds from each local point cloud, and obtain the average distance between the selected point clouds in each local point cloud and the corresponding inner wall of the crusher.

[0069] Specifically, but not limited to, select several point clouds from each local point cloud obtained in step S23, calculate the distance between each selected point cloud and the corresponding inner wall of the crusher, and finally obtain the average distance between the selected point clouds in each local point cloud and the corresponding inner wall of the crusher.

[0070] Exemplarily, taking the point cloud in the previous left region as an example, select several points randomly from the point cloud in the front left region, and calculate the distances between each point and the front left side of the inner wall of the crusher, so as to obtain the average distance between the selected point cloud and the front left side of the corresponding inner wall of the crusher.

[0071] S26: According to the positions of the local point clouds, set weight coefficients for each local point cloud;

[0072] S27: Sum the average distances according to the weight coefficients to obtain the space parameters between the material and the crusher.

[0073] Specifically, it is optional but not limited to set different weight coefficients for each local point cloud according to the positions of the local point clouds, and sum the average distances according to the weight coefficients and the average distances obtained in step S24, so as to obtain the space parameters between the material and the crusher.

[0074] Exemplarily, since the material will be gradually crushed during the process of moving along the set direction of the inner wall of the crusher, the farther the material travels, the higher the degree of crushing of the material. Optionally, set the first weight coefficient for the point cloud in the front left region and the point cloud in the front right region, and set the second weight coefficient for the point cloud in the rear left region and the point cloud in the rear right region, and then sum the average distances according to the weights, so as to obtain the space parameters between the material and the crusher; specifically, the weight coefficients can be arbitrarily set according to the properties of the material in the crusher, the working principle and mechanical structure of the current crusher, such as the positions of the feeding port and the discharging port, the stirring type, the spiral type, etc.

[0075] S3: According to the crusher status marked by the historical point cloud data and the extracted particle material morphology parameters, construct and train a monitoring model with the particle material morphology parameters as the input and the crusher status as the output;

[0076] Specifically, the monitoring model can adopt any neural network model in the prior art; according to the crusher status at each historical moment marked in step S1 and the particle material morphology parameters extracted in step S2, construct a training data set; then, according to the training data set, input the initial neural network model to train the parameters of the initial neural network model to obtain the trained neural network model as the monitoring model. It should be noted that this step is a preparatory work, which can be constructed and trained in advance, as long as it is constructed and trained before the neural network model is needed for prediction.

[0077] Preferably, as Figure 2 shown, the monitoring model includes: a classification unit, a first recognition unit, and a second recognition unit; where:

[0078] The classification unit is used to classify the input data into type-one data and type-two data according to the morphology parameters of each particle and the correlation parameters between particles;

[0079] The first recognition unit, connected to the classification unit, is used to make a prediction based on a certain type of data to obtain the crusher status;

[0080] The second recognition unit, connected to the classification unit, is used to make a prediction based on another type of data to obtain the crusher status.

[0081] Preferably, as Figure 3 shown, the classification unit includes an input layer, a normalization layer, a convolutional layer, a fully connected layer, an output layer, and a decision tree algorithm, where:

[0082] The input layer is used to receive the morphological parameters of each particle and the correlation parameters between particles, and input them into the normalization layer and the decision tree algorithm;

[0083] The normalization layer, the convolutional layer, and the fully connected layer are connected in sequence and are used to extract features from the morphological parameters of each particle and the correlation parameters between particles to obtain particle morphological features and particle correlation features; Preferably, the normalization layer is connected to the input layer and is used to extract features from the morphological parameters of each particle and the correlation parameters between particles to obtain particle morphological features and particle correlation features; the convolutional layer is connected to the normalization layer and is used to perform a convolution operation on the particle morphological features and particle correlation features to obtain 2 feature maps; the fully connected layer is connected to the convolutional layer and is used to flatten the feature maps to obtain several one-dimensional vectors, and further combine and transform each one-dimensional vector to obtain the updated particle morphological features and particle correlation features;

[0084] The decision tree algorithm is used to extract features from the morphological parameters of each particle and the correlation parameters between particles to obtain particle morphological grades and particle correlation grades; Preferably, the decision tree algorithm can optionally construct a tree structure, where each internal node represents a test on an attribute, each branch represents a test output, and each leaf node represents a category, so as to obtain particle morphological grades such as large, medium, and small, and particle correlation grades such as far, flat, and near;

[0085] The output layer is connected to the fully connected layer and the decision tree algorithm and is used to classify the input data into a certain type of data and another type of data according to the particle morphological features, particle correlation features, particle morphological grades, and particle correlation grades, and send the certain type of data to the first recognition unit and send the another type of data to the second recognition unit.

[0086] Preferably, as Figure 4 shown, the first recognition unit includes:

[0087] The first bottleneck layer is used to perform a dimensionality reduction operation on the input data and retain the data feature information to obtain the initial dimensionality-reduced data;

[0088] The first CBL convolutional layer, connected to the first bottleneck layer, is used to perform a convolutional operation on the initial dimensionality-reduced data to extract the local features of the initial dimensionality-reduced data, obtaining 2 feature maps;

[0089] The fully connected layer, connected to the first CBL convolutional layer, is used to flatten the feature maps to obtain several one-dimensional vectors, and further combine and transform each one-dimensional vector to obtain the crusher state.

[0090] Specifically, the first recognition unit includes the first bottleneck layer, the first CBL convolutional layer, and the fully connected layer, and the structure is relatively simple. Since the features of large particles are relatively easy to identify, for the recognition of large particles, there is no need for an overly complex feature extraction and processing process. The particle features can be extracted through a group of data bottleneck layers and the first CBL convolutional layer, reducing the data processing volume and improving the efficiency to efficiently complete the task through a simple model structure.

[0091] Preferably, as Figure 5 shown, the second recognition unit includes:

[0092] The second bottleneck layer is used to perform a dimensionality reduction operation on the input data and retain the data feature information, obtaining the initial dimensionality-reduced data;

[0093] The second CBL convolutional layer, connected to the second bottleneck layer, is used to perform a convolutional operation on the initial dimensionality-reduced data to extract the local features of the initial dimensionality-reduced data, obtaining 2 initial feature maps;

[0094] The third bottleneck layer, connected to the second CBL convolutional layer, is used to perform a dimensionality reduction operation on the initial feature maps and retain the data feature information, obtaining the optimized dimensionality-reduced data;

[0095] The third CBL convolutional layer, connected to the third bottleneck layer, is used to perform a convolutional operation on the optimized dimensionality-reduced data to extract the local features of the initial dimensionality-reduced data, obtaining 2 optimized feature maps;

[0096] The spatial pyramid pooling layer, connected to the third CBL convolutional layer, is used to divide the optimized feature maps into multiple sub-regions and perform a pooling operation on each sub-region to obtain several feature vectors of a fixed size;

[0097] The fully connected layer, connected to the spatial pyramid pooling layer, is used to further combine and transform each feature vector to obtain the crusher state.

[0098] Specifically, the second recognition unit includes two sets of bottleneck layers and CBL convolutional layers. With a more complex structure, it can extract richer features. Since the features of medium and small particles are more difficult to identify, a more complex model structure is required to capture these details. The two sets of bottleneck layers and CBL convolutional layers are used to extract features of the particles, obtaining deeper features, improving the recognition accuracy and precision. And through the spatial pyramid pooling layer, feature maps of any size are converted into fixed-size feature vectors, so as to extract feature information of different scales through multi-scale pooling operations, enhancing the robustness of the model to scale changes of the input image, thereby providing a standardized input for the subsequent fully connected layer and reducing the computational burden of the fully connected layer.

[0099] In this embodiment, a preferred embodiment of the monitoring model is given, including: a classification unit, a first recognition unit, and a second recognition unit; using the morphological parameters of the granular material, including: the morphological parameters of each particle, the correlation parameters between particles, the morphological parameters of the material composed of the particles, and the spatial parameters between the material and the crusher as the overall input data. First, through the morphological parameters of the particles and the correlation parameters between particles, via the classification unit, it is divided into type-one data and type-two data. Then, the first recognition unit with a simple structure or the second recognition unit with a complex structure is selected, which can selectively use different recognition models for different particles, improving the recognition accuracy and efficiency. Preferably, various preferred embodiments of the classification unit, the first recognition unit, and the second recognition unit are given. It should be noted that the specific structures of the monitoring model, the classification unit, the first recognition unit, and the second recognition unit are only for illustrative purposes and are not limited thereto. For example, the monitoring module includes the classification unit, the first recognition unit, and the second recognition unit only as a preferred embodiment. When no data classification is performed, the classification unit may not be included, and preferably any one of the first recognition unit or the second recognition unit is used; or according to the results of the first recognition unit or the second recognition unit, feature fusion is performed to obtain an updated result.

[0100] S4: Obtain the current point cloud data, extract the current morphological parameters of the granular material, and input them into the trained monitoring model to obtain the crusher status.

[0101] Specifically, it is optional but not limited to obtaining the current point cloud data, and according to step S2, also extracting the current morphological parameters of the granular material and inputting them into the monitoring model obtained in step S3, then the current crusher status can be obtained.

[0102] In this embodiment, a method for monitoring the state of a crusher based on the morphological changes of granular materials of the present invention is presented. First, historical point cloud data is obtained, and the states of the crusher at each historical moment are labeled. Then, morphological parameters of the granular materials are extracted from the historical point cloud data, including the morphological parameters of each particle, the correlation parameters between particles, the morphological parameters of the materials composed of the particles, and the spatial parameters between the materials and the crusher. On this basis, a monitoring model with the morphological parameters of the granular materials as the input and the crusher state as the output is constructed and trained. Finally, the current point cloud data is obtained, the current morphological parameters of the granular materials are extracted, and input into the trained monitoring model to obtain the crusher state. Among them, on the one hand, the morphological parameters of the particles and the correlation parameters between particles can reflect the microscopic characteristics of the particles themselves and between the particles from the perspective of individual particles. On the other hand, the morphological parameters of the materials composed of the particles and the spatial parameters between the materials and the crusher can reflect the macroscopic characteristics of the materials themselves and between the materials and the crusher from the perspective of the materials composed of the particles. Extracting the intricate morphological laws between the morphological changes of the granular materials and the crusher state from both the microscopic and macroscopic levels can improve the efficiency and accuracy of state recognition. It solves the problems of insufficient accuracy, weak data processing ability, too high cost, susceptibility to environmental influence and poor adaptability in the prior art.

[0103] On the other hand, the present invention also provides a computer storage medium storing executable program code; the executable program code is used to execute any of the above-mentioned methods for monitoring the state of a crusher based on the morphological changes of granular materials.

[0104] On the other hand, the present invention also provides a terminal device, including a memory and a processor; the memory stores program code executable by the processor; the program code is used to execute any of the above-mentioned methods for monitoring the state of a crusher based on the morphological changes of granular materials.

[0105] Exemplarily, the program code can be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the program code in the terminal device.

[0106] The terminal device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the terminal device may further include input / output devices, network access devices, a bus, etc.

[0107] The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0108] The memory may be an internal storage unit of the terminal device, such as a hard disk or memory. The memory may also be an external storage device of the terminal device, such as a plug-in hard disk equipped on the terminal device, a SmartMedia Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory may also include both the internal storage unit and the external storage device of the terminal device. The memory is used to store the program code and other programs and data required by the terminal device. The memory may also be used to temporarily store the data that has been output or will be output.

[0109] The above computer storage medium and terminal device are created based on the above crusher status monitoring method based on the morphological change of granular materials. Their technical functions and beneficial effects will not be elaborated here. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as these combinations of technical features do not conflict, they should all be considered as the scope described in this specification.

[0110] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the invention patent should be subject to the appended claims.

Claims

1. A crusher status monitoring method based on the morphological changes of granular materials, characterized in that, Including: Obtain the historical point cloud data of the granular material in the crusher, and label the crusher status at each historical moment; According to the historical point cloud data, extract the morphological parameters of the granular material, including: the morphological parameters of each particle, the correlation parameters between particles, the morphological parameters of the material composed of particles, and the spatial parameters between the material and the crusher; According to the crusher status labeled by the historical point cloud data and the extracted morphological parameters of the granular material, construct and train a monitoring model with the morphological parameters of the granular material as the input and the crusher status as the output; Obtain the current point cloud data, extract the current morphological parameters of the granular material, and input them into the trained monitoring model to obtain the crusher status.

2. The method according to claim 1, characterized in that The steps of extracting the morphological parameters of each particle, the correlation parameters between particles, and the morphological parameters of the material composed of particles according to the historical point cloud data include: Obtain the measurement information of the acquisition device; Convert the historical point cloud data according to the measurement information to obtain a data three-dimensional model; Extract features from the data three-dimensional model to obtain the morphological parameters of each particle, the correlation parameters between particles, and the morphological parameters of the material composed of particles.

3. The method according to claim 2, wherein The morphological parameters of each particle include any one or more of volume, surface area, roundness, and sphericity; the correlation parameters between particles include any one or more of particle spacing, bulk density, flow trajectory, and layered structure; the morphological parameters of the material composed of particles include any one or more of volume, surface area, length, width, and height.

4. The method according to claim 1, characterized in that, The steps of extracting the spatial parameters between the material and the crusher according to the historical point cloud data include: Perform direct filtering on the historical point cloud data to obtain local point clouds in several directions and the positions of each local point cloud; Select several point clouds from each local point cloud, and obtain the average distance between the selected point cloud in each local point cloud and the corresponding inner wall of the crusher; According to the positions of the local point clouds, set weight coefficients for each local point cloud; Sum the average distances according to the weight coefficients to obtain the spatial parameters between the material and the crusher.

5. The method according to claim 1, wherein The monitoring model includes: a classification unit, a first recognition unit, and a second recognition unit; where: The classification unit is used to classify the input data into type-one data and type-two data according to the morphological parameters of each particle and the correlation parameters between particles; The first recognition unit is connected to the classification unit and is used to make a prediction based on the type-one data to obtain the crusher status; The second recognition unit is connected to the classification unit and is used to make a prediction based on the type-two data to obtain the crusher status.

6. The method according to claim 5, wherein The classification unit includes an input layer, a normalization layer, a convolutional layer, a fully connected layer, an output layer, and a decision tree algorithm, where: The input layer is used to receive the morphological parameters of each particle and the correlation parameters between particles, and input them into the normalization layer and the decision tree algorithm; The normalization layer, the convolutional layer, and the fully connected layer are connected in sequence and are used to extract features from the morphological parameters of each particle and the correlation parameters between particles to obtain particle morphological features and particle correlation features; The decision tree algorithm is used to extract features from the morphological parameters of each particle and the correlation parameters between particles to obtain particle morphological grades and particle correlation grades; The output layer, connected to the fully connected layer and the decision tree algorithm, is used to classify the input data into type-one data and type-two data according to the particle morphological features, particle association features, particle morphological grades, and particle association grades, and send the type-one data to the first recognition unit and the type-two data to the second recognition unit.

7. The method according to claim 5, wherein The structure of the first recognition unit is as follows: The first bottleneck layer is used to perform dimensionality reduction on the input data and retain the data feature information to obtain the initial dimensionality-reduced data; The first CBL convolutional layer, connected to the first bottleneck layer, is used to perform convolutional operations on the initial dimensionality-reduced data to extract the local features of the initial dimensionality-reduced data and obtain 2 feature maps; The fully connected layer, connected to the first CBL convolutional layer, is used to flatten the feature maps to obtain several one-dimensional vectors, and further combine and transform each one-dimensional vector to obtain the crusher state.

8. The method according to claim 5, characterized in that, The structure of the second recognition unit is as follows: The second bottleneck layer is used to perform dimensionality reduction on the input data and retain the data feature information to obtain the initial dimensionality-reduced data; The second CBL convolutional layer, connected to the second bottleneck layer, is used to perform convolutional operations on the initial dimensionality-reduced data to extract the local features of the initial dimensionality-reduced data and obtain 2 initial feature maps; The third bottleneck layer, connected to the second CBL convolutional layer, is used to perform dimensionality reduction on the initial feature maps and retain the data feature information to obtain the optimized dimensionality-reduced data; The third CBL convolutional layer, connected to the third bottleneck layer, is used to perform convolutional operations on the optimized dimensionality-reduced data to extract the local features of the initial dimensionality-reduced data and obtain 2 optimized feature maps; The spatial pyramid pooling layer, connected to the third CBL convolutional layer, is used to divide the optimized feature maps into multiple sub-regions and perform pooling operations on each sub-region to obtain several feature vectors of a fixed size; The fully connected layer, connected to the spatial pyramid pooling layer, is used to further combine and transform each feature vector to obtain the crusher state.

9. A computer storage medium, characterized in that, Stores executable program code; the executable program code is used to execute the crusher state monitoring method based on the morphological changes of granular materials according to any one of claims 1-8.

10. A terminal device, characterized in that, Includes a memory and a processor; the memory stores program code executable by the processor; the program code is used to execute the crusher state monitoring method based on the morphological changes of granular materials according to any one of claims 1-8.

Images