Intelligent sorting system and method for same-size shield muck components based on multi-modal perception

Abstract

The application discloses a multi-modal perception-based intelligent sorting system and method for shield muck of the same particle size, and relates to the field of muck sorting. The system comprises a multi-modal perception module, a data processing and transmission module, an intelligent sorting decision module, an execution mechanism module and an auxiliary support module. The multi-modal perception module is used for comprehensively capturing muck and environmental characteristics through multi-dimensional and multi-type sensors, thereby providing original data support for subsequent sorting decisions. The application introduces a pebble-mudstone special perception unit in the perception link through deep integration of multi-modal perception and intelligent algorithms. An ultrasonic sensor and a hardness tester combination scheme is adopted. Not only can the internal cracks of pebbles and the weathered areas of mudstone be identified through the reflection wave characteristics, but also the quantitative differentiation of the mechanical properties of materials can be realized through the hardness value difference. Combined with high-resolution visual imaging and near-infrared spectrum analysis, a multi-dimensional feature matrix is formed, and the accuracy is significantly improved compared with traditional single visual recognition.

Description

Technical Field

[0001] This invention relates to the field of slag and soil sorting technology, specifically to an intelligent sorting system and method for slag and soil components of the same particle size from tunnel boring machines based on multimodal sensing. Background Technology

[0002] With the rapid development of urban rail transit construction, shield tunneling has been widely used, generating a large amount of shield tunneling excavated soil. Shield tunneling excavated soil of the same particle size contains various components such as sand, gravel, clay, impurities, pebbles, and mudstone. Among them, pebbles have high resource utilization value and can be used in building aggregates, landscape stone materials, and other fields.

[0003] For example, the "Construction Waste Treatment Device with Slag Screening Function and its Slag Screening Process" with publication number CN114713357B includes a base plate, a slag and steel bar separation and conveying device, a feed inlet, a slag block sorting device, a slag storage bin, a crushing device, and a steel bar storage bin. The slag and steel bar separation and conveying device is set on the base plate, the feed inlet is set on the slag and steel bar separation and conveying device, the slag block sorting device is set on the base plate and located on one side of the slag and steel bar separation and conveying device, the slag storage bin is set on one side of the slag block sorting device, the crushing device is set inside the slag storage bin, and the steel bar storage bin is set on the side of the slag storage bin away from the slag block sorting device.

[0004] However, in existing technologies, traditional slag sorting methods mainly rely on manual or simple mechanical screening. The distinction between pebbles and mudstone mainly depends on appearance and simple hardness judgment, which is inefficient and inaccurate. It is difficult to meet the needs of large-scale projects for efficient separation and high-value utilization of pebbles. Furthermore, traditional sorting devices are not effective at screening pebbles and mudstone within the same particle size range, making it difficult to separate pebbles with high purity, which affects their benefits in the process of resource utilization. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent sorting system and method for shield tunneling excavation components of the same particle size based on multimodal perception, so as to solve the problems of low efficiency and insufficient accuracy of traditional shield tunneling excavation sorting methods mentioned in the background art, especially the difficulty in effectively separating pebbles and mudstone within the same particle size range, resulting in low resource utilization rate of high-value components.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent sorting system for shield tunnel slag components of the same particle size based on multimodal perception, comprising a multimodal perception module, a data processing and transmission module, an intelligent sorting decision module, an execution mechanism module, and an auxiliary support module;

[0007] The multimodal sensing module is used to comprehensively capture the characteristics of the waste soil and environment through multi-dimensional and multi-type sensors, providing raw data support for subsequent sorting decisions. The data processing and transmission module is used to clean, enhance, transmit, and store the sensed data, providing data support for intelligent decision-making. The intelligent sorting decision-making module is used to perform feature analysis, fusion, and classification decisions on the processed data, generating execution instructions. The execution mechanism module is used to complete the physical sorting, transportation, and calibration of the waste soil according to the decision instructions. The auxiliary support module is used to provide power supply, monitoring, maintenance, and benefit evaluation support for the entire process.

[0008] Preferably, the multimodal sensing module includes a visual sensing unit, a spectral sensing unit, a mechanical sensing unit, an environmental sensing unit, and a pebble-mudstone specialized sensing unit. The visual sensing unit is used to acquire visual feature information of the color, shape, and texture of the slag; the spectral sensing unit is used to acquire chemical composition information of the slag; the mechanical sensing unit is used to acquire mechanical property information such as density, hardness, and particle distribution of the slag; the environmental sensing unit is used to monitor the temperature, humidity, and dust content of the sorting environment; and the pebble-mudstone specialized sensing unit is used to acquire specific feature information of pebbles and mudstone.

[0009] The data processing and transmission module includes a data acquisition submodule, a data cleaning submodule, a data preprocessing submodule, a data compression submodule, a data transmission submodule, a data storage submodule, and a pebble-mudstone data enhancement submodule. The data acquisition submodule is used to acquire signals from various sensors and convert them into digital signals. The data cleaning submodule is used to identify and process abnormal data. The data preprocessing submodule is used to enhance the features of the data. The data compression submodule is used to compress the data to reduce the bandwidth usage. The data transmission submodule is used to transmit the data. The data storage submodule is used to store the data and support querying and tracing. The pebble-mudstone data enhancement submodule is used to enhance the feature data of pebbles and mudstone.

[0010] The intelligent sorting decision module includes a feature extraction submodule, a feature verification submodule, a feature fusion submodule, a classification and recognition submodule, a self-learning optimization submodule, a decision instruction generation submodule, and a pebble-mudstone classification and optimization submodule. The feature extraction submodule is used to extract features from the processed data. The feature verification submodule is used to verify the consistency of the extracted features. The feature fusion submodule is used to fuse multimodal features. The classification and recognition submodule is used to classify and identify the components of the slag. The self-learning optimization submodule is used to update the model parameters through online learning. The decision instruction generation submodule is used to generate control instructions for the actuator. The pebble-mudstone classification and optimization submodule is used to optimize the classification and recognition logic for pebbles and mudstone.

[0011] The actuator module includes a pneumatic sorting unit, a mechanical sorting unit, a conveying and collecting unit, an end-of-line sorting and calibration unit, and a pebble-mudstone specialized sorting unit. The pneumatic and mechanical sorting units are used to sort the slag and soil. The conveying and collecting unit is used to convey and collect the sorted slag and soil. The end-of-line sorting and calibration unit is used to perform secondary identification on the sorted slag and soil and calibrate the actuator parameters. The pebble-mudstone specialized sorting unit is used to accurately separate pebbles and mudstone.

[0012] The auxiliary support module includes a power supply unit, a monitoring unit, a maintenance unit, a slag pretreatment unit, a quality traceability unit, an energy consumption management unit, and a pebble resource utilization benefit assessment unit. The power supply unit is used to supply power to the system, the monitoring unit is used to monitor the system's operating status, the maintenance unit is used to ensure system maintenance, the slag pretreatment unit is used to pretreat the slag, the quality traceability unit is used to record slag-related information and support traceability, the energy consumption management unit is used to manage system energy consumption, and the pebble resource utilization benefit assessment unit is used to assess the benefits of the separated pebbles in the resource utilization process.

[0013] Preferably, the pebble-mudstone specialized sensing unit includes an ultrasonic sensor and a hardness tester. The ultrasonic sensor is installed above the conveyor belt to emit ultrasonic waves and receive reflected waves. Based on the characteristics of the reflected waves, it determines the internal structure and integrity of the pebbles and mudstone. The hardness tester is in contact with the slag and measures its hardness value. The hardness of pebbles is usually higher than that of mudstone. The visual sensing unit performs specialized image acquisition and analysis on the differences in appearance between pebbles and mudstone.

[0014] Preferably, the pebble-mudstone data enhancement submodule enhances the characteristic data of pebbles and mudstone, including filtering and amplifying the ultrasonic reflection signal to highlight the differences in the internal structure of pebbles and mudstone, standardizing the hardness data to amplify the differences in hardness values ​​between pebbles and mudstone, and enhancing the display of characteristic areas of pebbles and mudstone in the visual image.

[0015] Preferably, the pebble-mudstone classification optimization submodule adjusts the parameter weights of the classification and recognition model based on the specific characteristics of pebbles and mudstone. It assigns higher weights to features such as uniform ultrasonic reflection waves, high hardness values, and smooth surfaces to identify pebbles, and assigns higher weights to features such as non-uniform ultrasonic reflection waves, low hardness values, and blocky, layered structures to identify mudstone.

[0016] Preferably, the pebble-mudstone specialized sorting unit includes an adjustable-speed conveyor belt and a grading screen. The adjustable-speed conveyor belt transports the components suspected to be pebbles and mudstones to the specialized sorting area, and the speed is adjusted within the range of 0.3 to 1 m / s according to the identification results. The grading screen is set with different screening parameters according to the differences in particle size and hardness between pebbles and mudstones to further separate pebbles and mudstones.

[0017] Preferably, the pebble resource utilization benefit assessment unit records the number, particle size distribution, and integrity parameters of the separated pebbles, and combines this with the application scenarios and value of pebbles in the market to assess their economic and environmental benefits in the resource utilization process, and generates an assessment report.

[0018] A method for intelligent sorting of shield tunnel excavation soil of the same particle size includes the following steps:

[0019] S1: Slag pretreatment, remove oversized lumps by vibrating screen, remove dust adhering to the slag surface by high-pressure air cleaning component, and preliminarily mark areas that may contain pebbles and mudstone.

[0020] S2: Multimodal perception, which acquires the visual characteristics, chemical composition information and mechanical property information of the slag through the visual perception unit, spectral perception unit and mechanical perception unit respectively, monitors the temperature, humidity and dust content of the sorting environment through the environmental perception unit, and acquires the exclusive characteristic information of pebbles and mudstone through the pebble-mudstone specialized perception unit.

[0021] S3: Data processing and transmission. The collected multimodal data is cleaned, preprocessed, and compressed in sequence. The pebble-mudstone data enhancement submodule enhances the characteristic data of pebbles and mudstones, and then the data transmission submodule transmits the data to the data storage submodule for storage.

[0022] S4: Intelligent sorting decision-making extracts features from the processed data. After feature verification and fusion, the features are input into the classification and identification submodule for classification of slag components. The pebble-mudstone classification optimization submodule optimizes the classification and identification logic of pebbles and mudstone. The self-learning optimization submodule updates the model through manual data verification. The decision instruction generation submodule generates control instructions based on the classification results.

[0023] S5: Perform sorting. The actuator module sorts the slag and soil through the pneumatic sorting unit, mechanical sorting unit and pebble-mudstone special sorting unit according to the control command. The conveying and collection unit conveys and collects the slag and soil. The end sorting and calibration unit performs secondary identification and calibrates the parameters, focusing on ensuring the accurate separation of pebble components.

[0024] S6: Full-process control, auxiliary support modules ensure stable system operation, quality traceability unit records relevant information, energy consumption management unit manages energy consumption, and pebble resource utilization benefit assessment unit evaluates the resource utilization benefit of the separated pebbles.

[0025] Preferably, in step S2, the ultrasonic sensor of the pebble-mudstone specialized sensing unit emits ultrasonic waves at a frequency of 2-5MHz. By analyzing the propagation time and amplitude changes of the reflected waves, it is determined whether there are cracks inside the pebbles and whether there are weathered and loose areas in the mudstone. The hardness tester is a Rockwell hardness tester with a measurement range of HRC20-60, which directly contacts the surface of the slag to obtain the hardness value.

[0026] Preferably, in step S5, the mesh size of the grading screening net of the pebble-mudstone special sorting unit is set according to the common fine particle size differences between pebbles and mudstone within the same particle size range. At the same time, combined with the precise gripping of the robotic arm, the suspected pebbles screened out are subjected to a second hardness test, and after confirmation, they are sent to the pebble-specific collection container.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. In this invention, a full-chain intelligent sorting system of "perception-processing-decision-execution" is constructed through the deep integration of multimodal perception and intelligent algorithms. In the perception stage, a special perception unit for pebbles and mudstone is introduced, and a combination of ultrasonic sensors and hardness testers is adopted. This not only identifies the internal cracks of pebbles and the weathered areas of mudstone through the characteristics of reflected waves, but also realizes the quantitative differentiation of material mechanical properties through the difference in hardness values. Combined with high-resolution visual imaging and near-infrared spectral analysis, a multi-dimensional feature matrix is ​​formed, which significantly improves the accuracy compared with traditional single visual recognition. In the data processing stage, the feature data of pebbles and mudstone are processed by filtering amplification, standardization and enhancement, which greatly improves the signal-to-noise ratio of key features. Through collaborative compression technology, the transmission bandwidth requirement is significantly reduced while ensuring data integrity. With the help of a distributed storage architecture, rapid traceability and analysis of full data can be achieved.

[0029] 2. In this invention, the actuator adopts a three-level linkage scheme of "pneumatic + mechanical + specialized sorting". The pneumatic nozzle and the robotic arm work together, combined with the fine distinction of the graded screening net, to achieve precise capture of pebbles within the same particle size range. The end sorting calibration unit identifies the pebbles twice and automatically fine-tunes the parameters, which ultimately greatly improves the purity of the pebbles sorting. This completely solves the industry pain point of the difficulty in separating pebbles and mudstone in slag of the same particle size. The coordinated operation of each link of the system not only improves the sorting efficiency, but also transforms the traditional sorting mode that relies on experience into a quantifiable and optimizable intelligent process. Attached Figure Description

[0030] Figure 1This is a flowchart of the multimodal sensing module of the intelligent sorting system for shield tunnel slag and soil components of the same particle size based on multimodal sensing, according to the present invention.

[0031] Figure 2 This is a flowchart of the data processing and transmission module of the intelligent sorting system for shield tunnel slag based on multimodal perception of the same particle size according to the present invention.

[0032] Figure 3 This is a flowchart of the intelligent sorting decision module of the intelligent sorting system for shield tunnel slag and soil components based on multimodal perception of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example 1: Refer to Figure 1 , Figure 2 and Figure 3 As shown: A multimodal sensing-based intelligent sorting system for shield tunnel slag of the same particle size includes a multimodal sensing module, a data processing and transmission module, an intelligent sorting decision module, an execution mechanism module, and an auxiliary support module. The multimodal sensing module is used to comprehensively acquire relevant information about the slag and the environment, including a visual sensing unit, a spectral sensing unit, a mechanical sensing unit, an environmental sensing unit, and a pebble-mudstone specialized sensing unit. The visual sensing unit uses a high-resolution industrial camera, installed above the slag conveyor belt, equipped with professional lighting equipment, to take multi-angle pictures of the slag on the conveyor belt, especially for the appearance differences between pebbles and mudstone (such as pebbles having a smoother surface, and mudstone being mostly blocky and layered) to collect and analyze their visual feature information. The spectral sensing unit uses a near-infrared spectral sensor to emit near-infrared light and receive the spectral signals reflected back from the slag. By analyzing the spectral data, it identifies the chemical composition information of the slag and helps to distinguish between pebbles and mudstone (such as mudstone may contain more clay minerals).

[0035] The mechanical sensing unit installs pressure and vibration sensors below the slag conveyor belt. The pressure sensors measure the pressure of the slag on the conveyor belt to reflect density information, while the vibration sensors detect vibration signals to obtain mechanical property information such as hardness and particle distribution. The environmental sensing unit includes a temperature and humidity sensor (measurement range: temperature -10~50℃, humidity 20%~90%RH) and a dust concentration sensor (detection accuracy 0.1mg / m³). 3Installed within 1 meter of the conveyor belt, it monitors the temperature, humidity, and dust content of the sorting environment in real time. If the temperature or humidity exceeds the sensor's normal operating range or the dust concentration exceeds the standard (>10mg / m³), the sensor will detect the problem. 3 When this occurs, an early warning is triggered and the dust removal equipment is activated.

[0036] The pebble-mudstone specialized sensing unit includes an ultrasonic sensor and a hardness tester. The ultrasonic sensor is installed above the conveyor belt to emit 2-5MHz ultrasonic waves and receive reflected waves. Based on the characteristics of the reflected waves (propagation time, amplitude variation), it judges the internal structure and integrity of pebbles and mudstone. For example, pebbles have a denser internal structure and more uniform reflected waves, while mudstone may have pores or bedding that cause chaotic reflected waves. The hardness tester uses a Rockwell hardness tester with a measurement range of HRC20-60. It is in contact with the slag and soil to measure its hardness value. The hardness of pebbles is usually higher than that of mudstone.

[0037] Example 2: According to Figure 2 As shown, the data processing and transmission module is used to process, transmit, and store the data acquired by the multimodal sensing module. It includes a data acquisition submodule, a data cleaning submodule, a data preprocessing submodule, a data compression submodule, a data transmission submodule, a data storage submodule, and a pebble-mudstone data enhancement submodule. The data acquisition submodule is responsible for acquiring the signals output by each sensor and converting them into digital signals. The data cleaning submodule is independent of the preprocessing stage and is specifically designed to identify and process abnormal data: it removes the instantaneous pulse interference values ​​of the pressure sensor using the 3σ criterion, fills in the occasional data gaps of the spectrometer using interpolation, marks motion-blurred frames in the visual image and triggers a retake command, and separately marks and processes abnormal data (such as abnormal changes in ultrasonic reflection waves and hardness values ​​exceeding the normal range) from the pebble-mudstone specialized sensing unit.

[0038] The data preprocessing submodule performs preprocessing operations such as denoising, filtering, and normalization on the cleaned data, focusing on feature enhancement to improve data quality. The data compression submodule uses H.265 encoding to compress visual image data (compression ratio up to 10:1) and performs sparse representation of spectral data through wavelet transform (compressing data volume by 60% while retaining 95% of feature values), reducing transmission bandwidth usage. The data transmission submodule uses high-speed Ethernet and TCP / IP protocol for data transmission. Data packetization and verification mechanisms ensure the integrity and accuracy of data transmission, and data transmission priority is set, with data related to pebbles and mudstone being transmitted first.

[0039] The data storage submodule adopts a distributed storage architecture, including a local SSD cache (capacity 512GB, storing nearly 1 hour of real-time data) and a cloud database (using a hybrid MySQL+MongoDB storage, with structured data stored in MySQL and raw image / spectral files stored in MongoDB). The data retention period is set to 1 year, supporting fast querying and tracing by time, source of slag, and whether it is pebbles or mudstone. The pebble-mudstone data enhancement submodule enhances the characteristic data of pebbles and mudstone, including filtering and amplifying ultrasonic reflection signals to highlight the differences in the internal structure of pebbles and mudstone, standardizing hardness data to amplify the differences in hardness values ​​between pebbles and mudstone, and enhancing the display of characteristic areas of pebbles and mudstone in visual images (such as the smooth surface of pebbles and the stratification of mudstone) to facilitate subsequent feature extraction.

[0040] Example 3: According to Figure 3 As shown, the intelligent sorting decision module is used to analyze the processed data and generate decision instructions. It includes a feature extraction submodule, a feature verification submodule, a feature fusion submodule, a classification and recognition submodule, a self-learning optimization submodule, a decision instruction generation submodule, and a pebble-mudstone classification optimization submodule. The feature extraction submodule uses a convolutional neural network (CNN) to extract visual features, focusing on extracting the appearance features of pebbles and mudstone. It uses an autoencoder to extract spectral features and extracts mechanical features through time-domain and frequency-domain analysis. It extracts features such as frequency and amplitude of reflected waves from ultrasonic sensor data and hardness value features from hardness tester data. The feature verification submodule performs consistency verification on the extracted multimodal features. For example, when the visual feature identifies it as "pebble" but the hardness value is low and the ultrasonic reflected wave is messy, it is marked as a suspicious sample and a manual review process is initiated to ensure the reliability before feature fusion.

[0041] The feature fusion submodule uses a weighted fusion algorithm to fuse visual, spectral, mechanical, and pebble-mudstone-specific feature vectors. The weights of each feature vector are determined through training sample data. For the classification of pebbles and mudstone, the weights of ultrasonic and hardness features are appropriately increased to obtain fused feature vectors. The classification and recognition submodule constructs a classification model based on a deep belief network (DBN), which contains three hidden layers. The input layer is the fused feature vector, and the output layer contains pebbles, mudstone, and other original components. The model is trained using a large amount of labeled slag sample data (including different types of pebble and mudstone samples) and outputs the category information of slag components. The self-learning optimization submodule has a built-in online learning engine that automatically extracts 500 manually reviewed misclassified samples (especially misclassified samples of pebbles and mudstone) every week. The DBN model parameters are updated through incremental training, which improves the monthly recognition accuracy of the model for pebbles and mudstone by an average of 0.5% to 1%.

[0042] The decision command generation submodule calculates the time and location parameters for sorting based on the classification and identification results, combined with the conveyor belt speed and the location information of the slag particles, and generates corresponding control commands. For the sorting commands of pebbles and mudstone, the accuracy parameter requirements are added and sent to the execution mechanism module. The pebble-mudstone classification optimization submodule adjusts the parameter weights of the classification and identification model based on the specific characteristics of pebbles and mudstone. It assigns higher weights to features such as uniform ultrasonic reflection waves, high hardness values, and smooth surfaces to identify pebbles, and assigns higher weights to features such as non-uniform ultrasonic reflection waves, low hardness values, and blocky, layered structures to identify mudstone, continuously optimizing the classification and identification logic.

[0043] Example 4: The actuator module is used to sort the slag according to decision commands. It includes a pneumatic sorting unit, a mechanical sorting unit, a conveying and collecting unit, an end-of-line sorting and calibration unit, and a pebble-mudstone specialized sorting unit. The pneumatic sorting unit consists of 10 sets of pneumatic nozzles, evenly arranged along the width of the conveyor belt. The nozzle diameter is 5mm, and the working air pressure is 0.6-0.8MPa. Each nozzle corresponds to an electromagnetic control valve, which controls the opening and closing of the valve by receiving decision commands. High-pressure airflow blows lightweight, small-particle slag to the corresponding collection tank. Smaller pebbles and mudstone can be assisted in sorting. The sorting unit is equipped with two 6-DOF industrial robotic arms with a reach of 1.5m and a repeatability of ±0.1mm. The arms are equipped with both vacuum suction cups and mechanical grippers at the end effector, which can automatically switch according to the size and weight of the soil particles. Upon receiving instructions, the arms grab larger pebbles, mudstone, and impurities and place them into the corresponding collection bins. A more stable gripping method is used for pebbles to prevent slippage. The conveyor belt of the conveyor and collection unit is made of wear-resistant rubber, is 1.2m wide and 5m long, and is driven by a variable frequency motor with a speed adjustable from 0.5 to 2m / s. Multiple collection containers are installed at the end of the conveyor belt.

[0044] How to use and how to work this device:

[0045] The pretreatment of slag and soil involves removing oversized clumps using a vibrating screen and removing dust adhering to the surface of the slag and soil using a high-pressure air-blowing cleaning component. At the same time, areas that may contain pebbles and mudstone are preliminarily marked.

[0046] Multimodal sensing acquires visual characteristics, chemical composition, and mechanical properties of the waste soil through visual, spectral, and mechanical sensing units, respectively. An environmental sensing unit monitors the temperature, humidity, and dust content of the sorting environment. A dedicated pebble-mudstone sensing unit acquires specific characteristic information of pebbles and mudstone. The ultrasonic sensor in the pebble-mudstone sensing unit emits ultrasonic waves at a frequency of 2-5 MHz. By analyzing the propagation time and amplitude changes of the reflected waves, it determines whether there are internal cracks in the pebbles and whether there are weathered and loose areas in the mudstone. A Rockwell hardness tester with a measurement range of HRC20-60 is used, directly contacting the surface of the waste soil to obtain hardness values.

[0047] Data processing and transmission involves cleaning, preprocessing, and compressing the collected multimodal data in sequence. The pebble-mudstone data enhancement submodule enhances the characteristic data of pebbles and mudstones, and then the data transmission submodule transmits the data to the data storage submodule for storage.

[0048] The intelligent sorting decision extracts features from the processed data. After feature verification and fusion, the features are input into the classification and identification submodule for the classification of slag components. The pebble-mudstone classification optimization submodule optimizes the classification and identification logic of pebbles and mudstone. The self-learning optimization submodule updates the model by manually reviewing the data. The decision instruction generation submodule generates control instructions based on the classification results.

[0049] The sorting process is performed by the actuator module, which sorts the slag and soil through a pneumatic sorting unit, a mechanical sorting unit, and a pebble-mudstone specialized sorting unit according to control commands. The conveying and collection unit conveys and collects the slag and soil, and the end sorting and calibration unit performs secondary identification and parameter calibration to ensure the accurate separation of pebble components. The mesh size of the graded screening screen in the pebble-mudstone specialized sorting unit is set according to the common fine particle size differences between pebbles and mudstone within the same particle size range. At the same time, combined with the precise gripping of the robotic arm, the suspected pebbles are subjected to a second hardness test. After confirmation, they are sent to a special pebble collection container.

[0050] The system features full-process control and support modules to ensure stable operation, a quality traceability unit to record relevant information, an energy consumption management unit to manage energy consumption, and a pebble resource utilization benefit assessment unit to evaluate the resource utilization benefits of the separated pebbles.

[0051] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart sorting system for shield tunneling excavation soil components of the same particle size based on multimodal perception, characterized in that: It includes a multimodal perception module, a data processing and transmission module, an intelligent sorting decision module, an execution mechanism module, and an auxiliary support module; The multimodal sensing module is used to comprehensively capture the characteristics of the slag and the environment through multi-dimensional and multi-type sensors, providing raw data support for subsequent sorting decisions. The data processing and transmission module is used to clean, enhance, transmit, and store the sensed data, providing data support for intelligent decision-making. The intelligent sorting decision-making module is used to perform feature analysis, fusion, and classification decisions on the processed data, generating execution instructions. The execution mechanism module is used to complete the physical sorting, transportation, and calibration of the slag according to the decision instructions. The auxiliary support module is used to provide power supply, monitoring, maintenance, and benefit evaluation support for the entire process. The multimodal sensing module includes a visual sensing unit, a spectral sensing unit, a mechanical sensing unit, an environmental sensing unit, and a pebble-mudstone specialized sensing unit. The visual sensing unit is used to acquire visual feature information of the color, shape, and texture of the slag. The spectral sensing unit is used to acquire chemical composition information of the slag. The mechanical sensing unit is used to acquire mechanical property information of the density, hardness, and particle distribution of the slag. The environmental sensing unit is used to monitor the temperature, humidity, and dust content of the sorting environment. The pebble-mudstone specialized sensing unit is used to acquire specific feature information of pebbles and mudstone. The data processing and transmission module includes a data acquisition submodule, a data cleaning submodule, a data preprocessing submodule, a data compression submodule, a data transmission submodule, a data storage submodule, and a pebble-mudstone data enhancement submodule. The data acquisition submodule is used to acquire signals from various sensors and convert them into digital signals. The data cleaning submodule is used to identify and process abnormal data. The data preprocessing submodule is used to enhance the features of the data. The data compression submodule is used to compress the data to reduce the bandwidth usage. The data transmission submodule is used to transmit the data. The data storage submodule is used to store the data and support querying and tracing. The pebble-mudstone data enhancement submodule is used to enhance the feature data of pebbles and mudstone. The intelligent sorting decision module includes a feature extraction submodule, a feature verification submodule, a feature fusion submodule, a classification and recognition submodule, a self-learning optimization submodule, a decision instruction generation submodule, and a pebble-mudstone classification and optimization submodule. The feature extraction submodule is used to extract features from the processed data. The feature verification submodule is used to verify the consistency of the extracted features. The feature fusion submodule is used to fuse multimodal features. The classification and recognition submodule is used to classify and identify the components of the slag. The self-learning optimization submodule is used to update the model parameters through online learning. The decision instruction generation submodule is used to generate control instructions for the actuator. The pebble-mudstone classification and optimization submodule is used to optimize the classification and recognition logic for pebbles and mudstone. The actuator module includes a pneumatic sorting unit, a mechanical sorting unit, a conveying and collecting unit, an end-of-line sorting and calibration unit, and a pebble-mudstone specialized sorting unit. The pneumatic and mechanical sorting units are used to sort the slag and soil. The conveying and collecting unit is used to convey and collect the sorted slag and soil. The end-of-line sorting and calibration unit is used to perform secondary identification on the sorted slag and soil and calibrate the actuator parameters. The pebble-mudstone specialized sorting unit is used to accurately separate pebbles and mudstone. The auxiliary support module includes a power supply unit, a monitoring unit, a maintenance unit, a slag pretreatment unit, a quality traceability unit, an energy consumption management unit, and a pebble resource utilization benefit assessment unit. The power supply unit is used to supply power to the system, the monitoring unit is used to monitor the system's operating status, the maintenance unit is used to ensure system maintenance, the slag pretreatment unit is used to pretreat the slag, the quality traceability unit is used to record slag-related information and support traceability, the energy consumption management unit is used to manage system energy consumption, and the pebble resource utilization benefit assessment unit is used to assess the benefits of the separated pebbles in the resource utilization process. The pebble-mudstone specialized sensing unit includes an ultrasonic sensor and a hardness tester. The ultrasonic sensor is installed above the conveyor belt to emit ultrasonic waves and receive reflected waves. Based on the characteristics of the reflected waves, it judges the internal structure and integrity of pebbles and mudstone. The hardness tester is in contact with the slag and measures its hardness value. The hardness of pebbles is usually higher than that of mudstone. The visual sensing unit performs specialized image acquisition and analysis on the appearance differences between pebbles and mudstone. The pebble-mudstone data enhancement submodule enhances the feature data of pebbles and mudstones, including filtering and amplifying ultrasonic reflection signals to highlight the differences in the internal structure of pebbles and mudstones, standardizing hardness data to amplify the differences in hardness values ​​between pebbles and mudstones, and enhancing the display of feature areas of pebbles and mudstones in visual images. The pebble-mudstone classification optimization submodule adjusts the parameter weights of the classification and recognition model based on the specific characteristics of pebbles and mudstone. It assigns higher weights to the characteristics of uniform ultrasonic reflection waves, high hardness values ​​and smooth surfaces to identify pebbles, and assigns higher weights to the characteristics of non-uniform ultrasonic reflection waves, low hardness values ​​and blocky and layered features to identify mudstone. The pebble-mudstone specialized sorting unit includes an adjustable-speed conveyor belt and a grading screen. The adjustable-speed conveyor belt transports the components suspected to be pebbles and mudstones to the specialized sorting area, and the speed is adjusted within the range of 0.3 to 1 m / s according to the identification results. The grading screen is set with different screening parameters according to the differences in particle size and hardness between pebbles and mudstones to further separate pebbles and mudstones. The pebble resource utilization benefit assessment unit records the number, particle size distribution, and integrity parameters of the separated pebbles, and combines this with the application scenarios and value of pebbles in the market to assess their economic and environmental benefits in the resource utilization process, and generates an assessment report.

2. A method for intelligent sorting of components in shield tunnel excavation soil of the same particle size, characterized in that, The intelligent sorting system for shield tunneling excavation soil components based on multimodal perception as described in claim 1 includes the following steps: S1: Slag pretreatment, remove oversized lumps by vibrating screen, remove dust adhering to the slag surface by high-pressure air cleaning component, and preliminarily mark areas that may contain pebbles and mudstone. S2: Multimodal perception, which acquires the visual characteristics, chemical composition information and mechanical property information of the slag through the visual perception unit, spectral perception unit and mechanical perception unit respectively, monitors the temperature, humidity and dust content of the sorting environment through the environmental perception unit, and acquires the exclusive characteristic information of pebbles and mudstone through the pebble-mudstone specialized perception unit. S3: Data processing and transmission. The collected multimodal data is cleaned, preprocessed, and compressed in sequence. The pebble-mudstone data enhancement submodule enhances the characteristic data of pebbles and mudstones, and then the data transmission submodule transmits the data to the data storage submodule for storage. S4: Intelligent sorting decision-making extracts features from the processed data. After feature verification and fusion, the features are input into the classification and identification submodule for classification of slag components. The pebble-mudstone classification optimization submodule optimizes the classification and identification logic of pebbles and mudstone. The self-learning optimization submodule updates the model through manual data verification. The decision instruction generation submodule generates control instructions based on the classification results. S5: Perform sorting. The actuator module sorts the slag and soil through the pneumatic sorting unit, mechanical sorting unit and pebble-mudstone special sorting unit according to the control command. The conveying and collection unit conveys and collects the slag and soil. The end sorting and calibration unit performs secondary identification and calibrates the parameters, focusing on ensuring the accurate separation of pebble components. S6: Full-process control, auxiliary support modules ensure stable system operation, quality traceability unit records relevant information, energy consumption management unit manages energy consumption, and pebble resource utilization benefit assessment unit evaluates the resource utilization benefit of the separated pebbles.

3. The intelligent sorting method for shield tunneling slag components of the same particle size according to claim 2, characterized in that: In step S2, the ultrasonic sensor of the pebble-mudstone specialized sensing unit emits ultrasonic waves with a frequency of 2-5MHz. By analyzing the propagation time and amplitude changes of the reflected waves, it is determined whether there are cracks inside the pebbles and whether there are weathered and loose areas in the mudstone. The hardness tester uses a Rockwell hardness tester with a measurement range of HRC20-60, which directly contacts the surface of the slag to obtain the hardness value.

4. The intelligent sorting method for shield tunneling slag components of the same particle size according to claim 3, characterized in that: In step S5, the mesh size of the graded screening net of the pebble-mudstone special sorting unit is set according to the common fine particle size differences between pebbles and mudstone within the same particle size range. At the same time, combined with the precise gripping of the robotic arm, the suspected pebbles screened out are subjected to a second hardness test, and after confirmation, they are sent to the special pebble collection container.

Images