Liftable peripheral transmission rake thickener operation system for coal washing

Abstract

The invention relates to the technical field of coal washing equipment, in particular to a lifting type peripheral transmission rake thickener operation system for coal washing, which comprises a working condition sensing module used for collecting slime water concentration, rake frame torque, slag layer thickness and equipment operation temperature data; the intelligent decision-making module is used for constructing a slime water sedimentation prediction model through an LSTM-GRU hybrid algorithm according to the data, and generating an equipment operation regulation and control instruction; the lifting transmission regulation and control module is used for adjusting the vertical height of a rake frame according to a regulation and control instruction to adapt to different slag layer thickness treatment requirements, and meanwhile, combining real-time feedback of the torque of the rake frame, dynamically adjusting the rotating speed of a peripheral transmission motor and optimizing coal slime and water separation conditions; and the safety early warning module is used for monitoring parameter mutation through a sliding window algorithm, and triggering corresponding strategies of different levels when the harrow torque and the slime water concentration exceed multiple groups of adaptability thresholds. Therefore, the problems that in the prior art, the height of the rake frame is fixed, regulation and control depend on manpower, and an early warning mechanism is extensive are solved.

Description

Technical Field

[0001] This invention relates to the field of coal washing equipment technology, specifically to a lifting peripheral drive rake thickener operating system for coal washing. Background Technology

[0002] As a core piece of equipment for coal slurry-water separation, the peripheral drive rake thickener for coal washing is a crucial hub in coal washing production lines, ensuring coal recovery rates and achieving water resource recycling and clean production. It plays an irreplaceable role in the field of coal processing and utilization. However, in the long-term operation process, due to factors such as the diversification of coal slurry composition, the complexity of the working environment, and the dynamic nature of processing needs, it has gradually revealed problems such as insufficient multi-dimensional state perception capabilities, lack of predictability in separation condition control, and passive and lagging safety protection. Specifically, the problems include: difficulty in integrating multi-source data such as coal slurry concentration, rake frame torque, slag layer thickness, and equipment temperature in real time for comprehensive working condition analysis; inability to dynamically optimize the rake frame height and the speed of surrounding drive motors based on the coal slurry settling trend; safety protection relying heavily on shutdown response after equipment failure rather than preventative early warning; and remote operation and maintenance only enabling basic parameter viewing, lacking real-time push notifications of abnormal information and interactive control commands. This leads to fluctuations in coal slurry separation efficiency, frequent equipment failures, coal resource loss, and energy waste. Power supply reliability, energy utilization efficiency, and intelligent management level are insufficient to meet the clean and efficient development needs of the modern coal washing industry. There is an urgent need to build a new operating system with full-dimensional perception, intelligent decision-making, dynamic adjustment, active protection, and remote collaboration capabilities.

[0003] Traditional peripheral drive rake thickener regulation systems for coal washing suffer from significant technical bottlenecks: Condition monitoring is limited to single parameter acquisition, lacking integrated perception of equipment health status and environmental factors, thus failing to support precise operational decision-making; control strategies rely on fixed thresholds, making it difficult to predict settling trends and optimize dynamic separation conditions. This results in rake height adjustments and motor speed adjustments lagging behind actual changes in slag layer thickness and coal slurry concentration, easily leading to torque overload with thick slag layers and incomplete coal slurry scraping with thin slag layers; safety protection relies primarily on passive cut-off, lacking parameter change rate monitoring and early warning, and devoid of preventative risk assessment and strategy generation mechanisms, leading to frequent false alarms and missed alarms; remote interaction functions are limited, only enabling basic operational data viewing, failing to meet the needs for real-time issuance of complex control commands and targeted push notifications of abnormal warnings. Overall, the technology faces multiple challenges: rigid regulation strategies, crude early warning mechanisms, and one-sided condition monitoring, severely restricting the continuity and economy of coal washing production lines. Summary of the Invention

[0004] This application provides a lifting peripheral drive rake thickener operating system for coal washing to solve the problems of fixed rake frame height, manual control, and crude early warning mechanism in the prior art.

[0005] The first aspect of this application provides an operating system for a liftable peripheral drive rake thickener used in coal washing, comprising: a condition sensing module, an intelligent decision-making module, a lifting transmission control module, and a safety early warning module; wherein, the condition sensing module is used to collect data on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature; the intelligent decision-making module is used to construct a coal slurry settling prediction model using an LSTM-GRU hybrid algorithm based on the coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature data, and generate equipment operation control commands; the lifting transmission control module is used to drive a hydraulic actuator to adjust the vertical height of the rake frame according to the equipment operation control commands, adapting to different slag layer thickness processing requirements, and simultaneously, combined with real-time feedback of the rake frame torque, dynamically adjust the speed of the peripheral drive motor to optimize the coal slurry and water separation conditions; the safety early warning module is used to monitor the coal slurry concentration and rake frame torque data using a sliding window algorithm, and triggers corresponding strategies at different levels when the rake frame torque and coal slurry concentration exceed multiple sets of adaptability thresholds.

[0006] The operating condition sensing module includes a coal slurry water parameter acquisition unit, an equipment status monitoring unit, and a data processing unit. The coal slurry water parameter acquisition unit is used to collect coal slurry water concentration and flow rate data. The equipment status monitoring unit is used to collect rake frame torque, slag layer thickness, system oil temperature, and motor winding temperature. The data processing unit is used to filter, reduce noise, and convert the format of the collected data to generate a standardized dataset.

[0007] The intelligent decision-making module includes a model computation unit, a control command generation unit, and a model training unit. The model computation unit, based on a standardized dataset, constructs a coal slurry settling prediction model using a hybrid LSTM-GRU algorithm. The LSTM network extracts long-term time-series features of coal slurry concentration and equipment temperature changes over time, while the GRU network optimizes the short-term correlation prediction between slag layer thickness and scraper torque. The control command generation unit outputs the coal slurry settling trend based on the prediction model and generates targeted control commands for different scenarios. The model training unit performs incremental training weekly using newly added historical operating data to control prediction errors.

[0008] The lifting transmission control module includes a hydraulic actuator, a transmission speed control unit, and a closed-loop feedback unit. The hydraulic actuator, based on the height adjustment requirements of the rake frame in the control command, drives the rake frame to perform vertical lifting movements via a hydraulic drive structure, adapting to processing needs with different slag layer thicknesses. The transmission speed control unit, based on the speed control requirements in the control command, adjusts the operating speed of the peripheral drive motors. Simultaneously, it dynamically adjusts the speed parameters based on real-time feedback information of the rake frame torque, optimizing the separation conditions of coal slime and water. The closed-loop feedback unit compares the actual rake frame height, the actual motor speed, and the target parameters in the control command in real time. When there is a deviation between the actual operating state and the command requirements, it automatically corrects the execution parameters of the hydraulic drive and speed adjustment.

[0009] Preferably, the safety early warning module includes a parameter monitoring unit, an anomaly determination unit, and an early warning push unit. The parameter monitoring unit uses a sliding window algorithm to monitor changes in coal slurry concentration and scraper torque data, quickly identifying sudden parameter mutations by calculating the rate of change of scraper torque and coal slurry concentration within each window. The anomaly determination unit uses multiple preset adaptability thresholds, combined with the parameter change trends and durations monitored by the sliding window, to logically verify and identify abnormal states of the thickener equipment. The early warning push unit triggers a corresponding level of early warning response based on the abnormal state and pushes the early warning information to management personnel via SMS or an app.

[0010] Preferably, the LSTM-GRU hybrid algorithm includes: LSTM algorithm for long-term feature extraction:

[0011]

[0012]

[0013]

[0014]

[0015]

[0016] GRU algorithm for short-term association prediction:

[0017]

[0018]

[0019]

[0020] Feature fusion using LSTM and GRU algorithms:

[0021] Output prediction results:

[0022] in, The proportion values ​​of long-term time-series information such as coal slurry concentration and equipment temperature from the previous moment are retained. For activation functions; Here is the weight matrix for the forget gate; for Long-term time-series characteristics of coal slurry water concentration and equipment temperature at any given moment; for Real-time collection of coal slurry water concentration data; for The equipment operating temperature data is collected continuously. For the bias term of the forget gate; This is the input ratio value for the current coal slurry concentration and equipment temperature information. This is the weight matrix of the input gate; This is the bias term for the input gate; These are candidate values ​​for new temporal characteristics of the current coal slurry concentration and equipment temperature. It is the hyperbolic tangent activation function; This is the weight matrix for the candidate cell states; This is a bias term for the candidate cell state; for The core characteristics of coal slurry water concentration and equipment temperature over a long period of time; For element-wise multiplication; for Long-term temporal characteristics of time; This represents the output ratio of long-term temporal features to the hidden state; This is the weight matrix of the output gate; This is the bias term for the output gate; for The final output is the long-term characteristics of the coal slurry water concentration and equipment temperature extracted at any time. This is the reset ratio for short-term information such as slag layer thickness and rake frame torque in the previous hidden state. To reset the weight matrix of the gate; for Short-term characteristic information of slag layer thickness and rake frame torque at any given time; for Data on slag layer thickness collected in real time; for Real-time collected rake frame torque data; To reset the door's bias; This represents the fusion ratio between the previous hidden state and the current candidate hidden state. To update the weight matrix of the gate; To update the gate's bias term; This is a candidate value for the new short-term correlation characteristics of the slag layer thickness and the rake frame torque at the current moment; Let be the weight matrix of the candidate hidden states; The bias term for the candidate hidden state; for The final output is the short-term correlation characteristics between the slag layer thickness extracted at any time and the rake frame torque; for The hidden state after fusing long-term temporal features and short-term correlation features; This is a vector concatenation function; for Key parameters related to the prediction of coal slurry water settling at specific times; To predict the weight matrix of the output layer; This is the bias term for predicting the output layer.

[0023] The second aspect of this application provides a method for operating a liftable peripheral drive rake thickener for coal washing, comprising: collecting data on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature; processing the data on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature, and inputting the data into a coal slurry settling prediction model constructed by an LSTM-GRU hybrid algorithm to generate equipment operation control commands; according to the equipment operation control commands, driving the hydraulic actuator to adjust the vertical height of the rake frame, adjusting the operating speed of the peripheral drive motor, and dynamically adjusting the speed parameters in conjunction with real-time feedback information of the rake frame torque to continuously optimize the separation conditions of coal slurry and water; simultaneously, monitoring changes in coal slurry concentration and rake frame torque data during the control process using a sliding window algorithm; based on the changes in coal slurry concentration and rake frame torque data, combined with multiple preset adaptability thresholds, using logical verification to judge and identify abnormal states of the thickener equipment, triggering corresponding level of early warning response, and pushing early warning information to management personnel via SMS or APP.

[0024] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement a method for operating a liftable peripheral drive rake thickener for coal washing as described in the above embodiments.

[0025] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a method for operating a liftable peripheral drive rake thickener for coal washing as described in the above embodiments.

[0026] The fifth aspect of this application provides a computer program product, including a computer program or instructions, for implementing a method for operating a liftable peripheral drive rake thickener for coal washing as described in the above embodiments.

[0027] Therefore, this application has the following beneficial effects: This application embodiment uses a working condition sensing module to collect and standardize multi-dimensional information on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature. Combined with the LSTM-GRU hybrid algorithm of the intelligent decision-making module, a coal slurry settling prediction model is constructed to accurately extract long-term time-series features and short-term correlation patterns, improving the accuracy of settling trend prediction, working condition assessment, and control command generation. The lifting transmission control module drives the hydraulic actuator to dynamically adjust the rake frame height based on control commands, and optimizes the speed of surrounding drive motors by combining real-time feedback of the rake frame torque. Through closed-loop feedback, execution parameters are corrected, effectively adapting to different slag layer thickness requirements, optimizing coal slurry-water separation conditions, and reducing rake frame overload failures and coal slurry loss. The safety early warning module uses a sliding window algorithm to monitor parameter mutations, and generates a graded early warning strategy by combining multiple sets of adaptability thresholds. Early warning information is pushed to users via SMS or an app, enabling proactive identification and preventative intervention of equipment anomalies, reducing the risk of false alarms and missed alarms. Remote early warning push and closed-loop control work together to effectively overcome the limitations of traditional systems. Therefore, this solves the problems of fixed rake frame height, reliance on manual control, and crude early warning mechanisms in existing technologies.

[0028] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0029] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the operating system of a liftable peripheral drive rake thickener for coal washing, according to an embodiment of this application. Figure 2 This is a schematic diagram of a working condition sensing module according to an embodiment of this application; Figure 3 This is a schematic diagram of an intelligent decision-making module provided according to an embodiment of this application; Figure 4 This is a schematic diagram of a settlement prediction model provided according to an embodiment of this application; Figure 5 This is a schematic diagram of lifting transmission control according to an embodiment of this application; Figure 6 This is a schematic diagram illustrating a security warning provided according to an embodiment of this application; Figure 7 This is a flowchart of an operating system for a liftable peripheral drive rake thickener for coal washing, provided according to an embodiment of this application. Figure 8 This is a flowchart illustrating the operation method of a liftable peripheral drive rake thickener for coal washing, according to an embodiment of this application. Figure 9 This is a schematic diagram of an operating method for a liftable peripheral drive rake thickener for coal washing, according to an embodiment of this application. Figure 10 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0030] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] The following description, with reference to the accompanying drawings, describes an embodiment of a liftable peripheral drive rake thickener operating system for coal washing. To address the issue of manual control mentioned in the background technology, this application provides an operating system for a liftable peripheral drive rake thickener used in coal washing. In this system, a condition sensing module collects and standardizes multi-dimensional information on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature. Combined with the LSTM-GRU hybrid algorithm of the intelligent decision-making module, a coal slurry settling prediction model is constructed to accurately extract long-term features and short-term correlation patterns, improving the accuracy of settling trend prediction, condition assessment, and control command generation. The lifting transmission control module dynamically adjusts the rake frame height based on control commands, and optimizes the peripheral drive motor speed in real-time based on rake frame torque feedback. Closed-loop feedback corrects execution parameters, effectively adapting to different slag layer thickness requirements, optimizing coal slurry-water separation conditions, and reducing rake frame overload failures and coal slurry loss. The safety early warning module uses a sliding window algorithm to monitor parameter mutations and generates a graded early warning strategy based on multiple sets of adaptability thresholds. Early warning information is pushed to users via SMS or an app, enabling proactive identification and preventative intervention of equipment anomalies and reducing the risk of false alarms and missed alarms. Remote early warning push and closed-loop control work together to effectively overcome the limitations of traditional systems. This solves the problems of fixed rake frame height, manual control, and crude early warning mechanisms in existing technologies.

[0032] Figure 1 This is a schematic diagram of the operating system of a liftable peripheral drive rake thickener for coal washing, provided as an embodiment of this application.

[0033] This application provides a lifting-type peripheral drive rake thickener operating system for coal washing, the system 10 including: Working condition perception module 100, intelligent decision-making module 200, lifting transmission control module 300, and safety early warning module 400.

[0034] The system comprises the following modules: The working condition sensing module 100 collects data on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature; the intelligent decision-making module 200 constructs a coal slurry settling prediction model using an LSTM-GRU hybrid algorithm based on these data, and generates equipment operation control commands; the lifting transmission control module 300 drives the hydraulic actuator to adjust the vertical height of the rake frame according to the equipment operation control commands, adapting to different slag layer thicknesses; simultaneously, it dynamically adjusts the speed of the surrounding drive motors based on real-time feedback of the rake frame torque, optimizing the coal slurry and water separation conditions; and the safety early warning module 400 monitors coal slurry concentration and rake frame torque data using a sliding window algorithm, triggering different levels of corresponding strategies when the rake frame torque and coal slurry concentration exceed multiple sets of adaptability thresholds.

[0035] It is understood that in this embodiment, the working condition sensing module collects and standardizes multi-dimensional information on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature. Combined with the LSTM-GRU hybrid algorithm of the intelligent decision-making module, a coal slurry settling prediction model is constructed to accurately extract long-term features and short-term correlation patterns, improving the accuracy of settling trend prediction, working condition assessment, and control command generation. The lifting transmission control module drives the hydraulic actuator to dynamically adjust the rake frame height based on control commands, and optimizes the speed of the surrounding drive motors by combining real-time feedback of the rake frame torque. Through closed-loop feedback, the execution parameters are corrected, effectively adapting to different slag layer thickness requirements, optimizing coal slurry separation conditions, and reducing rake frame overload failures and coal slurry loss. The safety early warning module uses a sliding window algorithm to monitor parameter mutations, combines multiple sets of adaptability thresholds to generate a graded early warning strategy, and pushes early warning information via SMS or APP to achieve proactive identification and preventative intervention of equipment anomalies, reducing the risk of false alarms and missed alarms. Remote early warning push and closed-loop control work together to effectively overcome the limitations of traditional systems. Thus, the problems of fixed rake frame height, reliance on manual control, and crude early warning mechanisms in existing technologies are solved.

[0036] In this embodiment of the application, the working condition sensing module 100 includes: Figure 2 As shown, there are coal slurry water parameter acquisition unit, equipment status monitoring unit, and data processing unit.

[0037] The coal slurry water parameter acquisition unit is used to collect coal slurry water concentration and flow rate data; the equipment status monitoring unit is used to collect rake frame torque, slag layer thickness, system oil temperature and motor winding temperature; and the data processing unit is used to filter, reduce noise and convert the format of the collected data to generate a standardized dataset.

[0038] It is understood that the coal slurry water parameter acquisition unit in this application embodiment captures the core characteristic parameters of coal slurry water concentration and flow rate in real time through dedicated sensors, and monitors the concentration and flow rate of coal slurry water in real time, providing data support for subsequent dynamic optimization of the rake frame height and drive motor speed, ensuring the adaptability and economy of coal slurry water separation conditions; the equipment status monitoring unit integrates torque sensors, thickness sensors and temperature sensors to continuously collect rake frame torque, slag layer thickness and system oil temperature, motor winding temperature and operating indicators, and promptly detects the risk of rake frame overload caused by excessive slag layer, and the safety hazards of equipment insulation aging and motor burnout caused by excessive oil temperature or winding temperature, thus building a safety barrier for the stable operation of the thickener; the data processing unit Focusing on the quality optimization and format standardization of raw data acquisition, this approach eliminates false signals caused by electromagnetic interference and mechanical vibration through filtering, corrects data deviations caused by instantaneous fluctuations through noise reduction, and unifies data standards from different sensors through format conversion. This generates standardized datasets, proactively avoiding biases in operational condition assessments due to data distortion and preventing subsequent model calls from being affected by format chaos. High-quality preprocessing of coal slurry water characteristics and equipment operating status data provides accurate and unified basic data, effectively connecting data acquisition and intelligent decision-making processes. This improves the accuracy of subsequent settlement prediction models and the precision of control commands, while strengthening the data's support for intelligent control and ensuring the high efficiency and stability of coal washing thickener operation.

[0039] For example, in a medium-sized coal washing plant equipped with a liftable peripheral drive rake thickener, during a continuous operation, the coal slurry water parameter acquisition unit, equipped with an ultrasonic coal slurry water concentration meter and an electromagnetic flow meter, collected data in real time at a frequency of 15 seconds per acquisition. At 10:00 AM that day, due to a brief fluctuation in the dosage of chemicals in the upstream coal washing process, the characteristics of the coal slurry water entering the thickener suddenly changed: at 10:00:00, the unit first detected that the coal slurry water concentration rose from the normal range of 200-230 g / L to 285 g / L, and when it was collected again at 10:00:15, the concentration further climbed to 310 g / L; at the same time, the electromagnetic flow meter showed that the coal slurry water flow rate increased from a stable 85 m³ / h to 110 m³ / h. The data collected for three consecutive times remained in the high concentration and high flow range, capturing in real time the change in operating conditions that suddenly increased the difficulty of coal slurry water separation, and providing core data basis for equipment status monitoring at the "coal slurry water end". The synchronously operating equipment status monitoring unit collects equipment operating data at a frequency of 30 seconds per cycle through a torque sensor installed on the rake frame main shaft, an ultrasonic slag layer thickness gauge on the thickener side wall, a platinum resistance temperature sensor in the hydraulic system oil tank, and a PT100 temperature sensor embedded in the motor windings. Affected by high-concentration coal slurry, at 10:00:30, the unit monitored a rapid increase in slag layer thickness from 1.1m (normal value) to 1.5m, and further to 1.8m at 10:01:00. Corresponding to the change in slag layer thickness, the rake frame torque gradually increased from 600N·m (60% of rated load) to 880N·m (approaching 90% of rated load). Simultaneously, the hydraulic system oil temperature rose from 42℃ to 56℃, and the motor winding temperature rose from 55℃ to 68℃, reflecting the dynamic trend of equipment load changes with the coal slurry conditions in real time, avoiding the monitoring blind spot of "only looking at the coal slurry and ignoring the equipment status." At this time, the data processing unit simultaneously receives the raw data transmitted from the two acquisition units and initiates the preset processing flow: For the instantaneous fluctuations in the coal slurry concentration data caused by electromagnetic interference (such as the ±8% fluctuation in the 310g / L data acquired at 10:00:15), a Kalman filter algorithm is used to remove the interference, reducing the concentration data fluctuation amplitude to ±1.5% and ensuring the core value of 310g / L is accurate and reliable; for the single instantaneous spurious value of 920N・m appearing in the rake frame torque data (caused by mechanical vibration), the moving average noise reduction method using 5 data points is used to remove this spurious value. The value was corrected to 875 N·m to align with the trend. Finally, all processed valid data (including timestamps, concentrations, flow rates, torques, slag thickness, oil temperatures, and motor winding temperatures) were uniformly converted into a standardized dataset in CSV format. A structured record was generated every 15 seconds, with fields including "collection time - coal slurry concentration (g / L) - coal slurry flow rate (m³ / h) - rake torque (N·m) - slag thickness (m) - system oil temperature (°C) - motor winding temperature (°C)". This provided standardized and clean basic data for subsequent data retrieval, avoiding analytical biases caused by messy raw data.

[0040] In this embodiment of the application, the intelligent decision-making module 200 includes: Figure 3 As shown, there are a model computation unit, a control instruction generation unit, and a model training unit.

[0041] The model computation unit, based on a standardized dataset, constructs a coal slurry water settling prediction model using a hybrid LSTM-GRU algorithm. The LSTM network extracts long-term time-series features of coal slurry water concentration and equipment temperature changes over time, while the GRU network optimizes the short-term correlation prediction between slag layer thickness and scraper torque. The control instruction generation unit outputs the coal slurry water settling trend based on the coal slurry water settling prediction model and generates targeted control instructions for different scenarios. The model training unit performs incremental training weekly using newly added historical operating data to control prediction errors.

[0042] It is understood that the computing unit in this application embodiment is based on a standardized dataset and constructs a coal slurry water settling prediction model using an LSTM-GRU hybrid algorithm. LSTM extracts long-term time-series features of coal slurry water concentration and equipment temperature, while GRU optimizes the short-term correlation prediction between slag layer thickness and scraper torque, avoiding the defects of a single algorithm and providing support for the prediction of settling trends. The control instruction generation unit generates targeted instructions according to the settling trends output by the model for different scenarios, avoiding the one-size-fits-all approach of traditional fixed instructions, preventing scraper overload and reducing coal slurry waste. The model training unit uses newly added historical data for incremental training every week, updates parameters to control prediction errors, avoids the problem of model prediction accuracy decay due to long-term operating condition fluctuations, and maintains the continuous effectiveness and scientific nature of the intelligent decision-making module.

[0043] It should be noted that the LSTM-GRU hybrid algorithm is used to construct a coal slurry settling prediction model. First, in the data input stage, a standardized dataset generated by the data processing unit of the operating condition perception module is received. This dataset has undergone filtering, noise reduction, and format conversion, and includes key parameters such as coal slurry concentration, equipment operating temperature (system oil temperature, motor winding temperature), slag layer thickness, and scraper torque. The dataset is recorded as time-series data at fixed time intervals. The LSTM-GRU hybrid algorithm is then used to process the standardized dataset. The LSTM network extracts long-term time-series features of the coal slurry concentration and equipment operating temperature, capturing the gradual change patterns of these parameters over long periods. The GRU network focuses on the slag layer thickness and scraper torque, optimizing their short-term correlation prediction and identifying the instantaneous linkage between parameters within a short timeframe. This compensates for the incomplete capture of long and short-term features by a single algorithm. Finally, the long-term time-series features extracted by LSTM and the short-term correlation features extracted by GRU are fused using the Concat vector stitching function. A comprehensive feature vector is used to construct a coal slurry sedimentation prediction model. The model ultimately outputs key prediction results related to coal slurry sedimentation. The specific parameters of the LSTM-GRU hybrid algorithm are set as follows: the LSTM layer has 64 neurons and 1 hidden layer, with a 2-dimensional input feature dimension corresponding to coal slurry concentration and equipment temperature, a time step of 30 steps, and activation functions using gated sigmoid and candidate cell state tanh; the GRU layer has 32 neurons and 1 hidden layer, with a 2-dimensional input feature dimension, and activation functions using gated sigmoid and candidate cell state tanh; the feature fusion layer forms a 96-dimensional comprehensive feature vector after concatenation, with a 64-dimensional LSTM output and a 32-dimensional GRU output. The subsequent fully connected layer has two layers: the first layer has 128 nodes to fully integrate long-term and short-term feature information, and the second output layer has 2 nodes, with a linear activation function; a weekly incremental training mechanism is also used, with a batch size of 32, a learning rate of 0.001, and a prediction error controlled to ≤5%.

[0044] For example, such as Figure 4As shown, during the daily operation of a lifting-type peripheral drive rake thickener in a coal washing plant, the operating condition sensing module continuously outputs a standardized dataset (updated every 15 seconds) containing coal slurry concentration, equipment temperature, slag layer thickness, and rake frame torque. The model computation unit calls this dataset to construct a coal slurry settling prediction model: the LSTM network extracts long-term temporal features over 24 hours—such as capturing the gradual change in coal slurry concentration from 190 g / L to 280 g / L and hydraulic oil temperature from 41℃ to 53℃ from 0:00 to 24:00 on the same day; the GRU network optimizes short-term correlation prediction—identifying the instantaneous linkage relationship where the rake frame torque increases from 630 N·m to 890 N·m when the slag layer thickness increases from 1.3 m to 1.7 m between 14:00 and 15:00. Based on this, the model outputs the coal slurry settling trend for the next 10 minutes: the slag layer thickness is expected to exceed 1.9 m (exceeding the safety threshold of 1.8 m). Based on this trend, the control command generation unit matches the "slag layer thickening overload" scenario and generates targeted control commands such as "increasing the height of the rake frame by 0.6m and reducing the speed of the peripheral drive motor to 0.8r / min". Subsequently, each week, the model training unit integrates 12,000 newly added historical operating data points (including data from special operating conditions such as high concentration and low flow) for incremental training, updating the long-term feature weights of the LSTM and the short-term correlation coefficients of the GRU, gradually reducing the prediction error of the coal slurry settling velocity from the initial 7.2% to 3.5%, ensuring that the model closely matches actual operating conditions in the long term.

[0045] In this embodiment of the application, the LSTM-GRU hybrid algorithm includes: LSTM algorithm for long-term feature extraction:

[0046]

[0047]

[0048]

[0049]

[0050]

[0051] GRU algorithm for short-term association prediction:

[0052]

[0053]

[0054]

[0055] Feature fusion using LSTM and GRU algorithms:

[0056] Output prediction results:

[0057] in, The proportion values ​​of long-term time-series information such as coal slurry concentration and equipment temperature from the previous moment are retained. For activation functions; Here is the weight matrix for the forget gate; for Long-term time-series characteristics of coal slurry water concentration and equipment temperature at any given moment; for Real-time collection of coal slurry water concentration data; for The equipment operating temperature data is collected continuously. For the bias term of the forget gate; This is the input ratio value for the current coal slurry concentration and equipment temperature information. This is the weight matrix of the input gate; This is the bias term for the input gate; These are candidate values ​​for new temporal characteristics of the current coal slurry concentration and equipment temperature. It is the hyperbolic tangent activation function; This is the weight matrix for the candidate cell states; This is a bias term for the candidate cell state; for The core characteristics of coal slurry water concentration and equipment temperature over a long period of time; For element-wise multiplication; for Long-term temporal characteristics of time; This represents the output ratio of long-term temporal features to the hidden state; This is the weight matrix of the output gate; This is the bias term for the output gate; for The final output is the long-term characteristics of the coal slurry water concentration and equipment temperature extracted at any time. This is the reset ratio for short-term information such as slag layer thickness and rake frame torque in the previous hidden state. To reset the weight matrix of the gate; for Short-term characteristic information of slag layer thickness and rake frame torque at any given time; for Data on slag layer thickness collected in real time; for Real-time collected rake frame torque data; To reset the door's bias; This represents the fusion ratio between the previous hidden state and the current candidate hidden state. To update the weight matrix of the gate; To update the gate's bias term; This is a candidate value for the new short-term correlation characteristics of the slag layer thickness and the rake frame torque at the current moment; Let be the weight matrix of the candidate hidden states; The bias term for the candidate hidden state; for The final output is the short-term correlation characteristics between the slag layer thickness extracted at any time and the rake frame torque; for The hidden state after fusing long-term temporal features and short-term correlation features; This is a vector concatenation function; for Key parameters related to the prediction of coal slurry water settling at specific times; To predict the weight matrix of the output layer; This is the bias term for predicting the output layer.

[0058] It is understood that the embodiments of this application extract long-term time-series features of coal slurry concentration and equipment operating temperature over time using an LSTM network. This accurately captures the gradual change patterns of these two parameters over a long period, avoiding interference from short-term parameter fluctuations on long-term trends. Simultaneously, by focusing on optimizing the short-term correlation prediction between slag layer thickness and scraper torque using a GRU network, the instantaneous linkage between these two parameters in a short period can be keenly identified. This compensates for the shortcomings of single time-series algorithms in capturing both long-term and short-term features incompletely. Furthermore, the fusion logic of long-term and short-term features is clearly defined, effectively combining the long-term time-series features extracted by LSTM with the short-term correlation features extracted by GRU to form comprehensive features covering both long-term trends and short-term changes. This provides a more comprehensive input basis for the coal slurry settling prediction model and scientific data support for the subsequent control command generation unit.

[0059] In this embodiment, the lifting transmission control module 300 includes: Figure 5 As shown, it includes a hydraulic actuator, a transmission speed control unit, and a closed-loop feedback unit.

[0060] The hydraulic actuator is used to drive the rake frame to move vertically up and down according to the height adjustment requirements of the rake frame in the control command, adapting to the processing needs of different slag layer thicknesses. The transmission speed control unit is used to adjust the operating speed of the peripheral drive motor according to the speed control requirements of the control command. At the same time, it dynamically adjusts the speed parameters based on the real-time feedback information of the rake frame torque to optimize the separation conditions of coal slime and water. The closed-loop feedback unit is used to compare the actual height of the rake frame, the actual speed of the motor and the target parameters in the control command in real time. When there is a deviation between the actual operating state and the command requirements, it automatically corrects the execution parameters of the hydraulic drive and speed adjustment.

[0061] It is understood that, in this embodiment, the hydraulic actuator drives the rake frame to rise and fall vertically according to the control command, adapting to different slag layer thickness requirements and solving the problems of traditional fixed rake frames being difficult to dynamically adapt, prone to failure, or having coal slime lost; the transmission speed control unit adjusts the motor speed according to the command, and dynamically adjusts it in combination with the rake frame torque feedback to optimize the separation conditions and avoid the defects of traditional fixed speed control being lagging, inefficient, or energy-intensive; the closed-loop feedback unit compares the actual height of the rake frame, the actual speed of the motor, and the target parameters in real time, and corrects the execution parameters when there is a deviation, ensuring accurate control.

[0062] It should be noted that in the hydraulic actuator, the hydraulic drive structure serves as the core power source for the vertical lifting of the rake frame. A hydraulic pump supplies pressurized oil, and a hydraulic cylinder converts hydraulic energy into mechanical energy, driving the rake frame to smoothly complete the lifting motion along a preset guide rail. When the slag layer thickness increases during coal slurry-water separation, the hydraulic drive structure can raise the rake frame height according to control commands, preventing a sudden increase in load and excessive torque due to excessive contact with the thick slag layer, thus preventing rake frame deformation or motor overload failure. When the slag layer thickness decreases, the hydraulic drive structure can lower the rake frame height, ensuring that the rake teeth on the rake frame fully contact the coal slurry layer, achieving thorough scraping and conveying of the coal slurry, and preventing coal slurry loss with the supernatant due to an excessively high rake frame. Ultimately, through this dynamically adjustable lifting action, it precisely adapts to changes in slag layer thickness under different working conditions, ensuring stable and efficient operation of the thickener under various processing requirements.

[0063] In the transmission speed control unit, adjusting the operating speed of the peripheral drive motors is to adapt to the coal slurry-water separation rhythm. When the coal slurry-water concentration is high and the scraping speed needs to be slowed down to ensure sufficient settling, the speed can be reduced. When the concentration is low and the coal slurry conveying needs to be accelerated to improve efficiency, the speed can be increased. At the same time, the real-time feedback information of the rake frame torque can directly reflect the equipment load status. For example, if the slag layer is too thick, it will cause the rake frame resistance to increase and the torque to rise. If the slag layer is too thin, the torque will decrease. The system will analyze this feedback in relation to the current motor speed: if the torque is detected to exceed the safety threshold, it means that the rake frame is overloaded at the current speed, and the speed needs to be reduced appropriately to reduce the running resistance of the rake frame and avoid motor overload or rake frame damage. If the torque is consistently low, it indicates that the current speed may be lower than the actual demand. The speed can be appropriately increased to accelerate the scraping and separation of coal slurry and avoid coal slurry retention or loss with the supernatant. This ensures that the motor speed always matches the coal slurry-water conditions and equipment load, improving the balanced separation efficiency and equipment safety. The torque safety threshold is set at 95% of the equipment's rated torque.

[0064] For example, in the daily operation of a peripheral drive rake thickener in a large coal washing plant, the machine's rated torque is 1000 N·m, and the safety threshold is set at 950 N·m (95% of the rated value). Under normal operating conditions, the coal slurry concentration is stable at 200-300 g / L, and the rake torque is maintained in the range of 600-850 N·m. At 8:30 AM one morning, the system monitoring showed a coal slurry concentration of 260 g / L, a slag layer thickness of 1.3 m, a rake height of 1.8 m, and a torque of 720 N·m. The peripheral drive motor was running at 1.2 r / min, and all parameters were within the normal range. At 9:05 AM, fluctuations in the upstream coal washing chemical dosing system caused the coal slurry concentration to surge to 380 g / L. Within 10 minutes, the slag layer thickness rapidly increased to 1.9 m. At 9:15 AM, the torque sensor detected that the rake torque reached 980 N·m, exceeding the safety threshold of 950 N·m, triggering the intelligent decision-making module to generate control commands. The hydraulic actuator immediately responded to the command to "raise the rake frame height to 2.2m," adjusting the working pressure from 12MPa to 16MPa and the flow rate from 30L / min to 45L / min via the variable hydraulic pump. This drove the hydraulic cylinder to raise the rake frame vertically along the guide rail. At 9:16, the rake frame height increased from 1.8m to 2.1m, initially reducing the contact area with the thick slag layer. Simultaneously, the transmission speed control unit received the basic command to "reduce the motor speed to 0.8r / min," and, combined with real-time torque feedback data (980N・m), further reduced the speed to 0.7r / min. By reducing the scraping speed, the equipment load was decreased. At 9:17, the system detected a 0.1m deviation between the actual rake frame height of 2.1m and the target value of 2.2m, and immediately and automatically corrected the hydraulic drive parameters: the variable pump output pressure increased by 0.5MPa, and the flow rate was fine-tuned to 47L / min, allowing the rake frame to reach the set height of 2.2m at 9:18. To address the deviation between the actual motor speed of 0.75 r / min and the commanded value of 0.7 r / min, the transmission speed control unit adjusted the output frequency of the motor driver, reducing it from 35 Hz to 33 Hz, ultimately stabilizing the speed at 0.7 r / min. At 9:20, the system entered a stable operating state: after adjustment by the hydraulic actuator, the rake frame remained at a height of 2.2 m; the transmission speed control unit stabilized the motor speed at 0.7 r / min; the torque sensor showed that the current torque had decreased to 820 N·m, returning to the normal range. Continuous monitoring by the closed-loop feedback unit revealed that the rake frame height fluctuation was controlled within ±0.02 m, and the motor speed deviation was less than 0.03 r / min, fully meeting the accuracy requirements. At this time, the thickener throughput remained at 65 t / h (80% of the design throughput), and the solids content of the overflow water was controlled below 300 ppm, meeting the operating standards of a high-efficiency thickener.

[0065] In this embodiment of the application, the security warning module 400 includes, as follows: Figure 6 As shown, there are parameter monitoring unit, anomaly detection unit, and early warning push unit.

[0066] The parameter monitoring unit uses a sliding window algorithm to monitor changes in coal slurry concentration and scraper torque data. By calculating the rate of change of scraper torque and coal slurry concentration within each window, it can quickly identify sudden parameter mutations. The anomaly judgment unit uses multiple preset adaptability thresholds, combined with the parameter change trend and duration monitored by the sliding window, to logically verify and identify abnormal states of the thickener equipment. The early warning push unit triggers corresponding level of early warning response based on the abnormal state and pushes the early warning information to management personnel via SMS or APP.

[0067] It is understood that the parameter monitoring unit in this application embodiment calculates the window change rate of the scraper torque and coal slurry concentration using a sliding window algorithm, which can quickly identify sudden parameter mutations and avoid equipment risks caused by slow accumulation of parameters and monitoring lag. The anomaly determination unit identifies equipment anomalies based on a preset adaptability threshold, combined with the parameter change trend and duration, through logical verification, thereby improving the accuracy and reliability of anomaly determination. The early warning push unit triggers graded early warnings according to the abnormal state and pushes information to management personnel via SMS or APP, thereby improving the response efficiency and safety of equipment operation and maintenance.

[0068] It should be noted that the sliding window algorithm refers to an algorithm that sets a fixed time window and calculates the rate of change of key parameters such as rake frame torque and coal slurry concentration in real time within each window to quickly identify sudden parameter changes. The formula is as follows:

[0069]

[0070] in, For the first The average value of coal slurry water concentration or equipment temperature within each sliding window; The size of the sliding window; For the first Real-time values ​​of coal slurry water concentration or equipment temperature at each sampling time; For the first The rate of change of parameters of each window relative to the previous window; For the first The average parameter values ​​of each window.

[0071] For example, a coal washing plant in a mountainous area, located in a rainy region, often faces the problem of fluctuating coal slurry properties due to a sudden increase in the moisture content of raw coal during the rainy season. Its Φ18m peripheral drive rake thickener has a rated torque of 850 N·m, with a torque safety threshold set at 90% of the rated value, i.e., 765 N·m. Under normal operating conditions, the coal slurry concentration is controlled at 180-250 g / L, and the rake torque is stable at 500-700 N·m. After three consecutive days of rainfall, the moisture content of the raw coal increased from the normal 10% to 18%. At 8:40 AM on July 15th, the coal slurry concentration fed into the coal washing system suddenly began to climb from 220 g / L. Parameter monitoring... The unit immediately activates a sliding window algorithm with a 15-second window interval to calculate the parameter change rate in real time within each window. In the window from 8:40:00 to 8:40:15, the coal slurry concentration increased from 220 g / L to 260 g / L, a change rate of 18.2%. In the window from 8:40:15 to 8:40:30, the concentration further increased to 300 g / L, a change rate of 15.4%. At the same time, the rake torque increased from 680 N·m to 730 N·m, a change rate of 7.4%. The algorithm quickly identifies the abnormal trend of "sudden increase in coal slurry concentration and synchronous increase in torque" and transmits the data to the anomaly judgment unit in real time. The anomaly detection unit calls multiple preset adaptation thresholds (for the working condition of raw coal moisture content of 15%-20%, the normal range of coal slurry concentration is widened to 180-300g / L, the corresponding normal range of scraper torque is 500-750N・m, and anomalies are only detected if "parameter exceeds the range or the rate of change exceeds 10% + duration ≥ 5 seconds"). Combined with the trend monitored by the sliding window: after 8:40:30, the coal slurry concentration stabilized at 300g / L (reaching the upper limit of the widened range), and the scraper torque was maintained at 740-750N・m (close to the upper limit of the range), and this state has lasted for 8 seconds. After logical verification, it is determined to be an equipment anomaly state of "moderate coal slurry concentration abnormality accompanied by high load", and the anomaly signal is sent to the early warning push unit.The early warning push unit immediately triggers a response mechanism based on the anomaly level (the torque safety threshold of 765 N·m has not been reached, which is judged as a level 2 early warning): At 8:40:40, an early warning information (including the time of the anomaly, the current concentration of 300 g / L, the torque of 750 N·m, the type of anomaly, and a suggestion to check the moisture content of the raw coal and the dosing system) is pushed to the on-duty maintenance personnel via the coal washing plant equipment management APP, and at the same time, a reminder is sent via SMS (to avoid the APP not being checked in time); at 8:40:55, after seeing the early warning information, the maintenance personnel immediately go to the dosing room to adjust the flocculant dosage (from 2). The concentration of coal slurry was increased from 5g / t to 3.0g / t. The motor speed was reduced from 1.1r / min to 1.0r / min through the remote command transmission speed regulation unit of the control system, reducing the load on the scraper frame. At 8:42:30, the parameter monitoring unit showed that the coal slurry concentration gradually dropped to 270g / L and the torque dropped to 690N・m. The anomaly judgment unit confirmed that the parameters returned to the normal range and lasted for 10 seconds, and judged that the anomaly was resolved. The early warning push unit automatically sent a "anomaly resolved" notification, which avoided the problem of excessive scraper frame torque and excessive solid content in overflow water caused by a further increase in coal slurry concentration.

[0072] The following will describe a specific embodiment of an operating system for a liftable peripheral drive rake thickener used in coal washing, such as... Figure 7 As shown, it includes: To improve coal slurry thickening efficiency and equipment operational safety, a large-scale thermal coal washing plant implemented a liftable peripheral drive rake thickener system. This system is designed to process 600 cubic meters of coal slurry per hour, with a rated underflow concentration control target of 500±20 g / L. The safe operating range for the rake frame torque is set at 1200-1800 N·m, the normal slag layer thickness is 0.8-1.5 m, the system oil temperature must be ≤65℃, and the motor winding temperature must be ≤85℃. During the production period from 8:00 AM to 8:30 AM on a certain workday, the operating condition sensing module was the first to activate. The coal slurry water parameter acquisition unit collects key data on feed concentration and flow rate every 5 minutes: 8:00 concentration 260g / L, flow rate 580m³ / h; 8:05 concentration 258g / L, flow rate 590m³ / h; 8:10 concentration 285g / L (fluctuating by ±7% due to electromagnetic interference), flow rate 600m³ / h; 8:15 concentration 272g / L, flow rate 595m³ / h; 8:20 concentration 269g / L, flow rate 588m³ / h; 8:25 concentration 274g / L, flow rate 592m³ / h; 8:30 concentration 268g / L, flow rate 596m³ / h. The equipment status monitoring unit collects core parameters every 8 minutes: 8:00: rake frame torque 1300 N·m, slag layer thickness 1.0 m, system oil temperature 55℃, motor winding temperature 75℃; 8:08: torque 1380 N·m, oil temperature 56℃, winding temperature 76℃; 8:16: torque 930 N·m (instantaneous false value due to mechanical vibration), slag layer thickness 1.1 m, oil temperature 56.5℃, winding temperature 76.5℃; 8:24: torque 1420 N·m, oil temperature 57℃, winding temperature 77℃; 8:30: torque 1450 N·m, slag layer thickness 1.2 m, oil temperature 58℃, winding temperature 78℃. After the collected raw data is transmitted to the data processing unit, the ±7% electromagnetic interference fluctuation accompanying the coal slurry water concentration of 285 g / L at 8:10 is eliminated by using a Kalman filter algorithm. After processing, the concentration data fluctuation amplitude is reduced to ±1.2%, and the core value is corrected to 282 g / L, ensuring that the data truly reflects the actual concentration. For the instantaneous spurious value of 930 N·m for the scraper torque at 8:16, this data point and five adjacent data points (8:10 1390 N·m, 8:12 1400 N·m, 8:16 930 N·m, 8:20 1410 N·m, 8:24 1420 N·m) are selected, and the average value is calculated to be 1396 N·m using the moving average noise reduction method. The spurious value is corrected to make it consistent with the continuous upward trend of the torque. Finally, the data processing unit converts all processed valid data (including timestamps, concentrations, flow rates, torques, slag thickness, oil temperatures, and motor winding temperatures) into a standardized dataset in CSV format. The data fields corresponding to each timestamp are standardized as floating-point numbers, providing reliable input for the intelligent decision-making module.

[0073] The intelligent decision-making module starts its computation based on the standardized dataset generated by the working condition perception module. The model computation unit uses an LSTM layer with 64 neurons and 1 hidden layer, with a 2D input feature dimension (corresponding to coal slurry concentration and equipment temperature) and a time step of 30 steps; and a GRU layer with 32 neurons and 1 hidden layer, with a 2D input feature dimension (corresponding to slag layer thickness and scraper torque). The activation functions are sigmoid with gating and tanh with candidate states. After feature fusion, the results are integrated through a fully connected layer with 128 nodes, and finally, the prediction results are output by a 2-node output layer using an LSTM-GRU hybrid algorithm. The LSTM network is responsible for extracting long-term temporal features of coal slurry concentration and equipment temperature changes over time. Historical operating data of the coal washing plant shows that when the coal slurry concentration is in the range of 260-280 g / L for 3 hours, the system oil temperature rises from 55℃ to 60℃, and the motor winding temperature rises from 75℃ to 80℃, the slag layer thickness increases by an average of 0.3 m and the scraper torque increases by an average of 150 N·m in the following 4 hours. The LSTM network learns these long-term historical correlations and outputs a feature vector of the influence of parameters such as concentration and temperature on the long-term settling behavior of coal slurry. The GRU network, on the other hand, focuses on the slag layer thickness and scraper torque data from the most recent two hours, as GRU is superior in predicting short-term dynamic correlations—historical data shows that for every 0.1m increase in slag layer thickness within the most recent two hours, the scraper torque increases by an average of 20 N·m. Based on this, the GRU network optimizes the short-term correlation prediction weights between slag layer thickness and scraper torque. The hybrid model combines the outputs of LSTM and GRU to construct a coal slurry settling prediction model. Starting with standardized data at 8:30, the model predicts the coal slurry settling trend for the next 2 hours: the feed concentration is expected to remain at 270±5 g / L, and the flow rate is expected to remain at 595±5 m³ / h; based on the long-term characteristics of LSTM, it is predicted that the slag layer thickness may reach 1.5 m after 1 hour (9:30), and the rake torque may rise to 1550 N·m; the GRU network further corrects this prediction, taking into account the current rate of increase of torque from 8:00 to 8:30 (an increase of 150 N·m in 30 minutes, an average of 5 N·m per minute), and finally predicts that the torque at 9:30 will be 1530 N·m, and the slag layer thickness will be 1.45 m. Based on this predicted trend, the control command generation unit generates targeted control commands: Since the predicted slag layer thickness is close to the upper limit of 1.5m and the torque is also close to the warning value of 1600N·m (the upper limit of the safe range is 1800N·m, with a reserve of 200N·m buffer space), the hydraulic execution unit is instructed to raise the height of the rake frame from the current 0.5m (relative to the bottom of the pool) to 0.8m to accommodate the thicker slag layer; at the same time, the transmission speed control unit is instructed to adjust the speed of the peripheral drive motor from the current 0.8r / min to 0.7r / min, reducing the speed to reduce the load on the rake frame, while ensuring the efficiency of coal slime conveying.In addition, the model training unit uses historical operating data added in the past week for incremental training at a frequency of once a week, a batch size of 32, and a learning rate of 0.001. The new data is first subjected to anomaly removal by the 3σ principle, that is, the historical mean and standard deviation of parameters such as coal slurry concentration, equipment temperature, slag layer thickness, and rake torque are calculated, and data that deviate from the mean by more than 3 times the standard deviation are judged as anomalies and screened out. Assuming that the actual torque at 9:30 is 1520 N·m and the slag layer thickness is 1.4 m after this prediction, the model will calculate a torque prediction error of 10 N·m (error rate of approximately 0.65%) and a thickness prediction error of 0.05 m (error rate of approximately 3.33%), and adjust the network parameters according to the errors: if the error rate is >3%, the learning rate of the corresponding network (GRU for torque error, LSTM or GRU for thickness error, determined by the dominant factor) will be temporarily increased to 0.002, and the weights of that network layer will be updated first; if the error rate is ≤3%, the original learning rate of 0.001 will be maintained, and the parameters of the entire network will be updated according to normal gradient descent to ensure that the subsequent prediction error is controlled within 5% and continuously optimize the model accuracy.

[0074] Upon receiving the control command from the intelligent decision-making module, the lifting transmission control module immediately executes it. In response to the command to "raise the rake frame height from 0.5m to 0.8m," the hydraulic actuator activates the hydraulic drive structure, setting the oil pump pressure to 12MPa. The extension and retraction of the hydraulic cylinders causes the rake frame to rise vertically. The closed-loop feedback unit monitors the actual height of the rake frame in real time every 2 seconds: the initial height is 0.5m, rising to 0.55m after 10 seconds, 0.6m after 20 seconds, and so on. After approximately one minute, the rake frame height stabilizes at 0.8m. The closed-loop feedback unit compares the actual height with the command target of 0.8m, controlling the error within ±0.01m. The hydraulic actuator then stops operating, maintaining the rake frame at a height of 0.8m. Meanwhile, in response to the instruction to "adjust the motor speed from 0.8 r / min to 0.7 r / min," the transmission speed control unit adjusted the inverter output frequency of the peripheral drive motors. The initial motor speed of 0.8 r / min corresponds to a frequency of 30 Hz, therefore the target frequency was calculated to be 26.25 Hz (speed and frequency are proportional, 0.7 / 0.8 = 26.25 / 30). During the speed adjustment process, the transmission speed control unit combined real-time feedback of the rake frame torque (torque was collected every 10 seconds): when the speed started to decrease from 0.8 r / min, the torque dropped from 1450 N·m to 1430 N·m in the first 10 seconds, verifying the effectiveness of the speed adjustment in reducing the load on the rake frame. Therefore, the speed was stabilized at 0.7 r / min as instructed. The closed-loop feedback unit collected the actual motor speed every 5 seconds via an encoder: after 5 seconds, the speed was 0.72 r / min, and after 10 seconds, it reached 0.70 r / min, with the error controlled within ±0.01 r / min, confirming that the speed adjustment was in place. In subsequent operation, for example at 9:00, the torque monitoring data was 1480 N·m, compared with 1450 N·m at 8:30 before the adjustment. The rate of increase was significantly slower, which indicates that the control commands for the lifting and speed adjustment of the rake frame effectively optimized the separation conditions of coal slime and water, and the load increase of the rake frame was reasonably controlled.

[0075] The safety early warning module continuously monitors for parameter anomalies throughout the process. The parameter monitoring unit uses a sliding window algorithm, with a window size of 5 minutes and a sliding window every 3 minutes, focusing on monitoring the rake frame torque and coal slurry concentration: In the first window (8:00-8:05), the average torque was 180 N·m and the average concentration was 340 g / L; in the previous window (7:57-8:02), the average torque was 175 N·m and the average concentration was 335 g / L, with calculated torque change rates of approximately 2.86% and concentration change rates of approximately 1.49%; in the second window (8:05-8:10), the average torque was 190 N·m and the average concentration was 345 g / L (data after Kalman filtering), and compared to the previous window, the torque change rate was approximately 5.56% and the concentration change rate was approximately 2.96% (not exceeding the preset mutation thresholds: torque mutation threshold ≥8%, concentration mutation threshold ≥5%). Although the rate of change of a single parameter did not reach the abrupt change threshold, the anomaly detection unit verified the following based on the coordinated trend of the parameters: From 8:00 to 8:30, the rake torque increased from 170 N·m (normal threshold 150-250 N·m) to 220 N·m (slightly approaching the upper limit of the normal threshold), the slag layer thickness increased from 1.1m (normal threshold 0.8-1.2m) to 1.2m (close to the upper limit of the normal threshold), the system oil temperature increased from 42℃ (within the normal range) to 44℃ (with an upward trend), and the motor winding temperature increased from 43℃ (within the normal range) to 45℃ (with an upward trend). The above changes fully conform to the potential anomaly logic of "slag layer thickening → load increase → temperature increase," and the rake torque and slag layer thickness both slightly deviated from or approached the "normal" threshold range. Therefore, it was determined to be a "potential anomaly state requiring attention," triggering a level one warning. The early warning push unit responded immediately: it sent an SMS to the management personnel stating, "Potential anomaly in the thickener: current torque 220 N·m (normal threshold 150-250 N·m), slag layer 1.2m (normal threshold 0.8-1.2m), please pay attention." Simultaneously, it pushed the warning and parameter trend chart to the equipment management APP. Subsequent monitoring showed that the lifting transmission control module slowed the torque increase rate, and the management personnel determined that the control was effective, requiring no additional intervention. This fully demonstrated the reliability of the module's coordination and improved the safety of coal slurry thickening.

[0076] In summary, this application's embodiments collect coal slurry parameters and equipment status data through a working condition perception module, providing a reliable input basis for subsequent decision-making. The intelligent decision-making module utilizes an LSTM-GRU hybrid algorithm to extract long-term features and optimize short-term correlation predictions, constructing a coal slurry sedimentation prediction model. Incremental training continuously optimizes accuracy, generating targeted control commands for different scenarios, avoiding the blindness of traditional control methods and ensuring measures align with actual operating trends. The lifting transmission control module's hydraulic execution unit adapts to slag layer thickness, and the transmission speed control unit adjusts motor speed. A closed-loop feedback unit compares actual parameters with target values ​​in real time and corrects deviations, ensuring accurate matching of equipment operating parameters to command requirements and improving operational stability and coal slurry separation efficiency. The safety early warning module uses a sliding window algorithm to monitor key parameters, combining parameter co-change trends and preset thresholds to determine abnormal states, triggering tiered early warnings and pushing information to promptly capture potential risks and prevent escalation of faults. This addresses the problems of unreliable data, lagging control, and difficulty in risk prediction in traditional thickeners, enabling intelligent, safe, and efficient operation of a liftable peripheral drive rake thickener for coal washing.

[0077] Next, referring to the accompanying drawings, a method for operating a liftable peripheral drive rake thickener for coal washing, according to an embodiment of this application, is described.

[0078] like Figure 8 As shown, the operating method of a liftable peripheral drive rake thickener for coal washing includes the following steps: In step S101, data on coal slurry water concentration, rake frame torque, slag layer thickness, and equipment operating temperature are collected.

[0079] It is understood that the embodiments of this application collect data on coal slurry concentration, rake torque, slag layer thickness, and equipment operating temperature to accurately grasp the core state of coal slurry concentration and the actual operation of the thickener in real time. This provides data support for coal slurry settling prediction, equipment operation control, and abnormal state identification. It also captures parameter fluctuation signals in a timely manner, providing early warning of potential separation efficiency decline or equipment overload problems. This provides a basis for adjusting the rake height and dynamically optimizing the motor speed, thereby improving coal slurry concentration efficiency, ensuring the safety of thickener operation, and reducing equipment failure rate.

[0080] In step S102, the data on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature are processed and then input into the coal slurry settling prediction model constructed by the LSTM-GRU hybrid algorithm to generate equipment operation control commands.

[0081] In processing the raw data of coal slurry concentration, rake torque, slag layer thickness, and equipment operating temperature, the following steps were taken: First, to eliminate electromagnetic interference fluctuations in the coal slurry concentration data, a Kalman filter algorithm was used to remove interference signals, significantly reducing the fluctuation range of the concentration data and ensuring that the core values ​​accurately reflect the actual characteristics of the coal slurry. Then, the rake torque, slag layer thickness, and equipment operating temperature data were denoised using a moving average noise reduction method to filter out high-frequency noise during equipment operation. Finally, all valid data processed by the algorithm were uniformly converted into a standardized dataset containing timestamps and standardized field formats.

[0082] It should be noted that the Kalman filter algorithm formula is:

[0083]

[0084]

[0085]

[0086]

[0087] in, For the first The prior prediction of the concentration of coal slurry water at any given time; This is the state transition matrix; For the first The posterior optimal estimate of the coal slurry water concentration at time t; To control the input matrix; For the first Time-based control input; For the first Error covariance of prior predictions at time points; For the first The error covariance of the posterior optimal estimate at time step; For process noise covariance; For the first Kalman gain at time step; The observation matrix; Observation matrix The transpose of the matrix; To observe the noise covariance; For the first The posterior optimal estimate at time step; For the first The raw concentration measurement value collected by the time sensor; For the first The error covariance of the posterior optimal estimate at time step; It is an identity matrix.

[0088] The formula for moving average noise reduction is:

[0089] in, For the first The parameter values ​​after moving average processing at any given time; The width is half the width of the window; This is the offset within the window; These are the original data points within the window.

[0090] It is understood that the embodiments of this application use the Kalman filter algorithm to remove electromagnetic interference fluctuations in the coal slurry concentration data, significantly reducing the fluctuation amplitude of the concentration data to ensure that the core values ​​truly reflect the actual characteristics. Then, the moving average denoising method is used to denoise the data of scraper torque, slag layer thickness, and equipment operating temperature to filter out high-frequency noise. The effective data is uniformly converted into a standardized dataset containing timestamps and standardized field formats, providing high-quality data for the subsequent input of the LSTM-GRU hybrid algorithm to construct the coal slurry sedimentation prediction model. This enables the model to accurately explore the long-term and short-term evolution patterns of the coal slurry system and improve the prediction accuracy of the sedimentation process. At the same time, it avoids problems such as scraper torque overload, excessive slag layer thickness, and abnormal equipment temperature, ensuring the stable and safe operation of the equipment.

[0091] For example, during the 8:00-8:10 AM operating period at a coal washing plant, the raw data collected by the liftable peripheral drive rake thickener used for coal washing is processed as follows: The raw data of coal slurry water concentration during 8:00-8:10 AM are 275 g / L, 278 g / L, 280 g / L, 292 g / L (fluctuating by ±8% due to electromagnetic interference), 281 g / L, and 279 g / L, respectively. First, the Kalman filter algorithm is used for processing—assuming the state transition matrix A=1, the observation matrix... With array H=1 and observation noise covariance R=12 (adapting to electromagnetic interference), after filtering and calculation, the concentration at 8:06 was corrected to 285 g / L, and the overall concentration data fluctuation range decreased from ±8% to ±1.2%. The core values ​​accurately reflect the actual characteristics of coal slurry water. The original data for the scraper torque during the same period were 1350 N·m, 1380 N·m, 1290 N·m (high-frequency noise caused by mechanical vibration), 1340 N·m, and 1360 N·m, and the original slag layer thickness was... The data were 1.08m, 1.12m, 1.05m (sensor fluctuations), 1.10m, and 1.11m. The original equipment operating temperature data were 55℃, 56℃, 54.5℃ (heat dissipation fluctuations), 55.5℃, and 56℃. Using a moving average noise reduction method with a window size of 5, the torque correction value was calculated to be 1344 N·m, the slag layer thickness correction value to be 1.09m, and the temperature correction value to be 55.4℃. High-frequency noise was filtered out. Finally, all processed data were... The data was uniformly converted into a standardized dataset in CSV format. Each data entry included standardized fields such as timestamp (e.g., 8:00, 8:02), coal slurry concentration (275.0 g / L, 278.0 g / L, etc.), rake frame torque (1350.0 N·m, 1344.0 N·m, etc.), slag layer thickness (1.08 m, 1.09 m, etc.), and equipment operating temperature (55.0℃, 55.4℃, etc.), providing reliable data support for subsequent settlement prediction model input.

[0092] In step S103, according to the equipment operation control command, the hydraulic actuator is driven to adjust the vertical height of the rake frame, adjust the operating speed of the peripheral drive motor, and dynamically adjust the speed parameters in combination with the real-time feedback information of the rake frame torque to continuously optimize the separation conditions of coal slime and water. At the same time, the changes in coal slime water concentration and rake frame torque data during the control process are monitored through the sliding window algorithm.

[0093] Among them, the continuous optimization of the separation conditions of coal slime and water involves the separation of coal slime and water by using external forces such as gravity and centrifugal force, or by adding coagulants to reduce the dispersibility of coal slime particles. This is combined with adjusting the particle size of coal slime particles and controlling the viscosity and flow rate of water, so that the coal slime particles can overcome the buoyancy and viscous resistance of water and achieve solid-liquid separation by aggregation or sedimentation. Continuous optimization refers to adjusting the vertical height of the rake frame by driving the hydraulic actuator to adapt to changes in slag layer thickness. When the slag layer thickens, the rake frame is raised to avoid excessive compression that would cause coal slime particles to redisperse into the water, thus maintaining the stability of the slag layer structure to facilitate water penetration. The agitation intensity of the rake frame is controlled by adjusting the speed of the peripheral drive motor. Lowering the speed reduces disturbance to the settled coal slime, while increasing the speed enhances the compaction effect on loose coal slime, balancing coal slime aggregation and water discharge efficiency. Simultaneously, the speed is dynamically adjusted by combining real-time feedback of the rake frame torque. When the torque increases, the slag layer resistance increases, so the speed is reduced to reduce the load and prevent the coal slime from being over-agitated and hindering separation. When the torque is too low, the slag layer is loose, so the speed is appropriately increased to enhance compaction and promote water separation.

[0094] It is understood that the embodiments of this application optimize the separation conditions of coal slime and water by adjusting the height of the rake frame to match the thickness of the slag layer and adjusting the motor speed in combination with torque feedback. This can avoid abnormal slag layer or excessive agitation, improve the efficiency of coal slime water treatment, and ensure the processing capacity of the equipment. At the same time, it reduces the moisture content of coal slime, improves the utilization rate of circulating water, and optimizes the quality of the separated products. It can also reduce equipment overload and wear, extend the life of vulnerable parts, reduce energy consumption and reagent consumption, and help achieve environmental protection standards.

[0095] For example, when a lifting-type peripheral drive rake thickener in a coal washing plant is running, the initial slag layer thickness is 0.9m. The system, according to the control instructions, sets the rake frame vertical height to 0.7m and the peripheral drive motor speed to 3.2 r / min. At this time, the rake frame torque is stable at 13.0 N·m. Three minutes later, the sensor detects that the slag layer thickness has increased to 1.2m. The system immediately drives the hydraulic actuator to raise the rake frame height to 1.0m to prevent excessive compression of the slag layer by the rake teeth, which would cause the coal slime particles to redisperse. Five minutes later, the rake frame torque suddenly increases to 17.2 N·m (due to increased slag layer resistance). The controller, based on real-time torque feedback, dynamically reduces the motor speed to 2.3 r / min. After the parameters stabilize, the torque drops back to 13.5 N·m. During the monitoring process, a sliding window algorithm with a window size of 5 minutes was used: the coal slime water concentration fluctuated from an initial 285 g / L ± 6%, and after optimization, it stabilized at 283-287 g / L (fluctuation ± 0.7%); the overflow turbidity decreased from 102 NTU to 45 NTU, and the underflow concentration increased to 70.2%. Finally, through coordinated adjustment of height and rotation speed, the coal slime moisture content decreased from 33% to 27%, and the circulating water reuse rate increased to 93%, effectively optimizing the separation conditions of coal slime and water.

[0096] In step S104, based on the changes in coal slurry concentration and rake frame torque data, combined with multiple preset adaptability thresholds, the abnormal state of the thickener equipment is judged and identified using logic verification, and the corresponding level of early warning response is triggered. The early warning information is pushed to the management personnel via SMS or APP.

[0097] Among them, the multiple sets of adaptability thresholds refer to key parameters such as slag layer thickness, rake frame torque, motor speed, coal slurry water feed concentration, feed flow rate, underflow concentration, overflow turbidity, liquid level height in the pool, dosage of coagulation or flocculation agents, hydraulic actuator pressure, and current of peripheral transmission mechanisms during thickener operation. The preset multiple sets of critical values ​​can match the corresponding thresholds according to the actual parameter changes, thereby triggering rake frame height adjustment, speed regulation, agent dosage optimization, and liquid level control adaptation and regulation actions. The specific thresholds are as follows: slag layer thickness: low threshold 0.5m, normal threshold 0.8-1.2m, high threshold 1.5m; rake frame torque: low threshold 100N·m, normal threshold 150-250N·m, high threshold 300N·m; motor speed: low threshold 2.0r / min, normal threshold 2.5-3.5r / min, high threshold 4.0r / min; coal slurry water concentration: low threshold 200g / L, normal threshold 250-350g / L, high threshold 400g / L; overflow turbidity: low threshold 30NTU, medium threshold 50NTU, high threshold 80NTU. The following parameters are specified for the liquid level in the pool: low threshold 1.5m, normal threshold 1.8-2.5m, high threshold 3.0m; feed flow rate: low threshold 80m³ / h, normal threshold 100-300m³ / h, high threshold 320m³ / h; coagulation or flocculation agent dosage: low threshold 5mg / L, normal threshold 8-15mg / L, high threshold 20mg / L; hydraulic actuator pressure: low threshold 1.5MPa, normal threshold 2.0-3.5MPa, high threshold 4.0MPa; peripheral transmission mechanism current: low threshold 15A, normal threshold 20-30A, high threshold 35A.

[0098] It should be noted that logical verification refers to a method of identifying abnormal device states by comparing a single parameter of device operation with a preset threshold and verifying the logical relationship between multiple parameters.

[0099] Triggering corresponding levels of early warning responses includes: Level 1 warning: triggered when a single non-core parameter approaches the "low" or "high" threshold boundary, or when a core parameter slightly deviates from the "normal" threshold range. At this time, the impact on equipment operation and separation effect is minimal and there is no safety risk. Only a text warning is sent via the APP. Management personnel must view and record it within 2 hours; no on-site intervention is required, and subsequent tracking of parameter change trends is sufficient. Level 2 warning: triggered when a core parameter exceeds the "normal" threshold and enters the "low" or "high" threshold range, or when two or more non-core parameters simultaneously exceed limits. This may lead to a decrease in separation effect and slight fluctuations in equipment load, requiring timely intervention to prevent the risk from escalating. This triggers APP push notifications and SMS reminders. Management personnel must respond within 1 hour. Parameters can be adjusted remotely; if remote adjustment is ineffective, on-site investigation is required. Level 3 warning: triggered when a core parameter significantly exceeds the "high" threshold (or triggers a critical risk on the "low" threshold side), potentially leading to equipment damage and process interruption. Immediate action is required. This triggers APP push notifications and SMS reminders, along with a telephone reminder to management personnel. Management personnel must arrive on-site within 15 minutes. If necessary, an "emergency load reduction" command can be remotely triggered first to prevent the accident from escalating.

[0100] It is understood that the embodiments of this application, by analyzing the changes in various parameters during the operation of the thickener and combining multiple preset adaptability thresholds, use logical verification to identify abnormal equipment states and trigger corresponding level warning responses. Furthermore, warning information is pushed to management personnel via SMS or APP, breaking the limitations of traditional manual inspections that rely on a single parameter and are prone to missing hidden anomalies. This avoids equipment overload damage or deterioration of coal-slurry-water separation effect due to untimely detection of anomalies. By matching different levels of urgency with the response requirements through graded warnings, it prevents minor anomalies from excessively consuming management resources and ensures that serious anomalies can quickly reach management personnel. Combined with mobile push notifications, it solves the pain point of management personnel not being able to be on-site in real time, ensuring the long-term stable operation of the thickener, reducing maintenance costs, and maintaining the continuous compliance of the coal-slurry-water separation process.

[0101] For example, a coal washing plant's 15m diameter liftable peripheral drive rake thickener daily processes 150m³ / h of coal slurry. Initially, the coal slurry concentration was 300g / L, the coagulant dosage was 12mg / L, the overflow turbidity was 40NTU, and the tank level was 2.2m—all parameters were within normal ranges, and the equipment issued no warnings. One hour later, the coal slurry concentration sensor reading rose to 420g / L, but the coagulant dosage remained at 12mg / L without adjustment. Simultaneously, the overflow turbidity rose to 62NTU. Logical verification identified a correlation between "exceeding coal slurry concentration + failure to adjust coagulant dosage + abnormal overflow turbidity," triggering a level-two warning. The system then sent a notification to equipment management personnel via the app: "Equipment No. XM-08, Level-two warning: Coal slurry concentration 420g / L, overflow turbidity 62..." "NTU, please adjust the dosage of the reagent in time" and send a text message with the same information. After the manager checks the warning within 5 minutes, the reagent dosage is remotely increased to 18 mg / L. After 30 minutes, the coal slurry concentration drops to 380 g / L, the overflow turbidity drops to 48 NTU, and the system automatically cancels the warning. If this is not handled in time, according to the subsequent parameter change trend, the coal slurry concentration will continue to rise to 450 g / L after 1 hour, the overflow turbidity will reach 85 NTU, and the liquid level in the pool will rise to 2.9m. The logic verification will further determine to trigger a level 3 warning. In addition to the APP and text message, the system will also automatically call the manager every 5 minutes to remind him. After the manager arrives, he can trigger an emergency command to "suspend feeding for 10 minutes + increase the speed of the underflow pump to 1.2 times the original speed" to ultimately avoid the risk of slag overflow and motor overload.

[0102] According to the embodiments of this application, a method for operating a liftable peripheral drive rake thickener for coal washing is proposed. This method involves a working condition sensing module that collects and standardizes multi-dimensional information on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature. Combined with an intelligent decision-making module using an LSTM-GRU hybrid algorithm, a coal slurry settling prediction model is constructed to accurately extract long-term features and short-term correlation patterns, improving the accuracy of settling trend prediction, working condition assessment, and control command generation. The lifting transmission control module dynamically adjusts the rake frame height based on control commands, and optimizes the peripheral drive motor speed in real-time based on rake frame torque feedback. Closed-loop feedback corrects execution parameters, effectively adapting to different slag layer thickness requirements, optimizing coal slurry-water separation conditions, and reducing rake frame overload failures and coal slurry loss. The safety early warning module uses a sliding window algorithm to monitor parameter mutations and generates a graded early warning strategy based on multiple sets of adaptability thresholds. Early warning information is pushed to users via SMS or an app, enabling proactive identification and preventative intervention of equipment anomalies and reducing the risk of false alarms and missed alarms. The remote early warning push and closed-loop control work together to effectively overcome the limitations of traditional systems. This solves the problems of fixed rake frame height, manual control, and crude early warning mechanisms in existing technologies.

[0103] The following will illustrate the operation method of a liftable peripheral drive rake thickener for coal washing through a specific embodiment, such as... Figure 9 As shown, it includes: A large coal washing plant is equipped with a liftable, peripherally driven rake thickener with a diameter of 15m, an effective water depth of 4.5m, and a designed processing capacity of 150m³ / h. Its core function is to separate and recover coal slime and water. The coal slime-water concentration, rake torque, slag layer thickness, and equipment operating temperature are key parameters affecting separation efficiency and equipment safety. To accurately collect these data, the plant installed a CY-2000 ultrasonic concentration meter 1m above the tank via the thickener's feed pipe. High-frequency ultrasonic waves penetrate the coal slime-water, and the concentration value is calculated based on the sound wave attenuation rate. The sampling frequency is set to 1 time / minute to avoid instantaneous data fluctuations caused by feed impact. A JM-300 torque sensor is installed at the connection between the rake drive shaft and the reducer. Strain gauges sense torque changes and output values ​​in real time, with a sampling frequency of 1 time / 20 seconds to ensure the capture of subtle fluctuations in the rake load. Three CS-10 [unclear - possibly referring to a specific type of thickener] are evenly distributed around the tank. The Type 0 ultrasonic level gauge is installed 0.3m above the bottom of the pool. It calculates the distance between the top of the slag layer and the sensor by measuring the ultrasonic reflection time, and then calculates the slag layer thickness. The average value is taken every 30 seconds to reduce the impact of water flow disturbance in the pool. At the same time, PT100 thermocouple temperature sensors are installed at three locations: the stator winding of the peripheral drive motor, the hydraulic oil tank, and the rake frame bearing housing. These sensors monitor the motor operating temperature, hydraulic oil temperature, and the temperature of key transmission components. The sampling frequency is 1 time / minute, the temperature measurement range is -50℃ to 200℃, and the accuracy is ±0.5℃. At 8:00 AM on a regular production day, the thickener starts up. In the initial stage, the ultrasonic concentration meter at the feed pipe shows a coal slurry concentration of 302 g / L, the torque sensor monitors a rake frame torque of 152 N·m, the average value of the three ultrasonic level gauges calculates a slag layer thickness of 0.82 m, the motor temperature is 34.8℃, the hydraulic oil temperature is 32.5℃, and the bearing temperature is 33.2℃. This real-time collected data is transmitted to the data analysis terminal in the central control room via the 485 communication protocol to prepare for subsequent processing and model input.

[0104] After receiving the raw data transmitted from various sensors, the data analysis terminal first performs preprocessing. Due to electromagnetic interference generated by the operation of on-site equipment, the coal slurry concentration data is prone to fluctuations. For example, at 8:03, the concentration momentarily jumped to 325 g / L, significantly deviating from the actual operating conditions. In this case, the terminal uses a Kalman filter algorithm on the coal slurry concentration data, combined with real-time estimation of the measurement noise covariance and process noise covariance, to eliminate electromagnetic interference signals and correct the outlier to 305 g / L. This reduces the concentration data fluctuation range from ±15 g / L to ±3 g / L, ensuring that the core values ​​accurately reflect the characteristics of the coal slurry. Subsequently, the data on the rake frame torque, slag layer thickness, and equipment operating temperature are processed using a moving average noise reduction method (with a sliding window size set to 5 sampling points). For the rake frame torque (sampling once every 20 seconds), the average of the samples over 100 seconds is used to filter out high-frequency noise generated by transmission vibration. For example, the original torque value fluctuated between 150-158 N·m between 8:04 and 8:05, and after processing, it became stable. The value is 154 N·m. For the slag layer thickness (one average value every 30 seconds) and equipment temperature (sampling once per minute), the data fluctuations are also smoothed by moving average to reduce the errors caused by water flow disturbance and changes in ambient temperature. Finally, all effective data processed by the algorithm are uniformly converted into a standardized dataset containing the standard fields of "collection timestamp (accurate to the second), parameter name, value, and unit". According to the formula "(actual value - minimum value) / (maximum value - minimum value)", the coal slurry water concentration (200g / L~400g / L), the rake torque (100N·m~300N·m), the slag layer thickness (0.5m~1.5m), and the equipment operating temperature (30℃~60℃) are mapped to the interval [0,1]. At 8:05, the normalized value of the processed coal slurry water concentration is 0.51, the rake torque is 0.26, the slag layer thickness is 0.32, and the motor temperature is 0.16. The preprocessed dataset was input into a coal slurry sedimentation prediction model constructed using an LSTM-GRU hybrid algorithm. This model had previously been trained using historical operating data from the plant over the past three months (collected 24 hours a day, once per minute, totaling 129,600 data sets). The model structure includes one LSTM layer with 64 neurons (used to capture long-term dependencies in the sedimentation process), one GRU layer with 32 neurons (used to quickly handle short-term parameter fluctuations), and two fully connected layers (which ultimately output commands for the vertical height of the rake frame and the motor speed). Testing showed that the model achieved a 92.3% accuracy rate in predicting sedimentation effects. Combining the currently processed data, the model analyzed historical similar operating conditions (such as the optimal control scheme at a concentration of approximately 300 g / L and a torque of approximately 150 N·m) and generated equipment operation control commands at 8:06: the vertical height of the rake frame was adjusted to 0.7 m, and the speed of the peripheral drive motor was set to 3.2 r / min.The logic for generating this instruction is based on the principle that "when the slag layer thickness is 0.82m, the height of the rake frame should be 0.12m lower than the slag layer to avoid idling and ensure compaction effect; the motor speed of 3.2r / min is within the normal threshold range of 2.5-3.5r / min, which can balance settling efficiency and energy consumption." The instruction is then transmitted to the thickener's actuator control unit in the form of a digital signal.

[0105] Upon receiving the control command, the actuator control unit immediately drives the corresponding hydraulic actuator and peripheral drive motor. The hydraulic actuator uses a YHG1 type metallurgical hydraulic cylinder with a stroke of 1.2m. The control unit first sends a signal to the hydraulic pump, smoothly increasing the system pressure from the initial 2.0MPa to 2.5MPa, pushing the cylinder piston rod upwards and causing the rake frame to slowly rise. The entire adjustment process takes 22 seconds. During this time, the torque sensor provides real-time feedback on the load changes of the rake frame, ensuring a smooth ascent. The peripheral drive motor is a 15kW Y2-160M2-2 type three-phase asynchronous motor, paired with an ACS510 frequency converter. The control unit adjusts the frequency converter's output frequency, gradually increasing the motor speed from the initial 3.0r / min to 3.2r / min, with a frequency adjustment increment of 0.1Hz / second, to avoid torque shocks caused by sudden speed changes. At 8:10, the height of the rake frame stabilized at 0.7m and the motor speed stabilized at 3.2r / min. At this time, the torque sensor showed a torque of 165N·m, which was within the normal threshold range of 150-250N·m. However, at 8:15, the feed concentration increased to 330g / L due to the adjustment of the coal washing process, and the slag layer thickness increased to 0.95m. The torque of the rake frame increased to 205N·m, which was close to the upper limit of the normal threshold. The control unit, combined with the real-time torque feedback, automatically reduced the motor speed to 3.0r / min. After 10 minutes, the torque dropped back to 180N·m and remained in the stable range. Meanwhile, the central control room used a sliding window algorithm with a window size of 5 minutes to monitor the changes in various parameters during the control process: In the first window from 8:00 to 8:05, the coal slurry water concentration fluctuated between 300-305 g / L, the WGZ-2000 turbidity meter installed at the overflow trough showed a turbidity of 45 NTU, and the gamma-ray concentration meter at the bottom flow pump outlet showed a bottom flow concentration of 65.2%; In the window from 8:10 to 8:15, the concentration fluctuation narrowed to 298-302 g / L, the turbidity dropped to 42 NTU, and the bottom flow concentration rose to 66.8%; In the window from 8:20 to 8:25, the concentration stabilized at 299-301 g / L, the turbidity further dropped to 40 NTU, and the bottom flow concentration reached 67.5%. The changing trends of various parameters indicate that the separation conditions of coal slurry and water are continuously optimized. The sliding window algorithm continuously records the change curves of these parameters, providing real-time data support for subsequent anomaly judgment.

[0106] The anomaly detection module in the central control room retrieves parameter data monitored by the sliding window in real time and performs logical verification in conjunction with multiple preset adaptability thresholds. Among the preset thresholds, the low threshold for coal slurry concentration is 200g / L, the normal threshold is 250-350g / L, and the high threshold is 400g / L; the low threshold for rake frame torque is 100N·m, the normal threshold is 150-250N·m, and the high threshold is 300N·m; the low threshold for slag layer thickness is 0.5m, the normal threshold is 0.8-1.2m, and the high threshold is 1.5m; among the equipment operating temperatures, the normal threshold for motor temperature is 30-45℃, and the high threshold is 50℃; the normal threshold for hydraulic oil temperature is 30-40℃, and the high threshold is 45℃; and the normal threshold for bearing temperature is 30-45℃, and the high threshold is 50℃. At the same time, the judgment criteria for Level 1 warning (single parameter approaching the threshold), Level 2 warning (single parameter exceeding the limit or two parameters approaching the threshold), and Level 3 warning (two or more parameters exceeding the limit) are set. At 10:00 AM that day, a malfunction in the coal washing machine caused a surge of high-concentration coal slurry into the feed end. The ultrasonic concentration meter detected a rapid increase in concentration from 320 g / L to 435 g / L within 10 minutes, exceeding the high threshold of 400 g / L. Simultaneously, the motor temperature rose from 38°C to 48°C due to increased load, exceeding the normal upper limit of 45°C; the rake frame torque rose to 285 N·m, approaching the high threshold of 300 N·m. The anomaly detection module, through logical verification, identified the associated anomalies of "exceeding coal slurry concentration + exceeding motor temperature + approaching exceeding torque," triggering a level-two warning. Subsequently, the communication module in the central control room immediately pushed an early warning message to the management personnel's dedicated APP. The message stated: "A coal washing plant, equipment number XM-08, Level 2 warning: At 10:10, coal slurry concentration 435g / L (high threshold 400g / L), motor temperature 48℃ (normal ≤45℃), rake torque 285N・m (high threshold 300N・m). It is recommended to adjust the dosage of coagulant and the speed of the peripheral drive motors." Simultaneously, a text message was sent to the management personnel's mobile phone, with the same content as the APP message, ensuring timely delivery of information. After checking the APP warning at 10:12, the management personnel remotely operated the system through the central control room's human-machine interface: increasing the dosage of the coagulant (polyaluminum chloride) from 12mg / L to 18mg / L, increasing the agent pump speed from 1500r / min to 2200r / min; decreasing the peripheral drive motor speed from 3.0r / min to 2.8r / min, and reducing the output frequency from 50Hz to 46.7Hz via the frequency converter. At 10:25, the sliding window monitoring showed that the coal slurry concentration dropped to 380g / L, the motor temperature dropped to 43℃, and the rake frame torque dropped to 220N・m. All parameters returned to the normal range, and the anomaly judgment module automatically lifted the warning.If this warning is not handled in time, based on the parameter change trend, the coal slurry concentration will rise to 460g / L at 11:00, the rake frame torque will exceed the high threshold of 300N·m, and the motor temperature will rise to 52℃. At this time, the anomaly judgment module will determine that a level three warning has been triggered. In addition to APP and SMS push notifications, it will also automatically call the management personnel and backup management personnel every 5 minutes to remind them. If there is no response within 15 minutes, the central control room will automatically trigger the "emergency load reduction" command, suspend the feeding for 10 minutes, and at the same time increase the speed of the underflow pump to 1.2 times the original speed to accelerate the discharge of coal slurry, avoid overflow or motor overload damage, and ensure equipment and process safety.

[0107] In summary, this application embodiment accurately collects core parameters, generates control instructions using an LSTM-GRU model after preprocessing, drives the actuator to dynamically adjust, monitors and optimizes separation conditions using a sliding window, and then combines threshold and logic verification to trigger graded early warnings to ensure the continuous compliance of the coal slurry-water separation process. Furthermore, it avoids the risks of motor overload and slag overflow through real-time monitoring and anomaly early warning, and reduces labor costs and improves response efficiency by relying on remote operation and maintenance, thus ensuring the stable and efficient operation of the thickener.

[0108] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 1001, the processor 1002, and the computer program stored on the memory 1001 and capable of running on the processor 1002.

[0109] When the processor 1002 executes the program, it implements the operating method of the liftable peripheral drive rake thickener for coal washing provided in the above embodiments.

[0110] Furthermore, electronic devices also include: Communication interface 1003 is used for communication between memory 1001 and processor 1002.

[0111] The memory 1001 is used to store computer programs that can run on the processor 1002.

[0112] The memory 1001 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0113] If the memory 1001, processor 1002, and communication interface 1003 are implemented independently, then the communication interface 1003, memory 1001, and processor 1002 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0114] Optionally, in a specific implementation, if the memory 1001, processor 1002, and communication interface 1003 are integrated on a single chip, then the memory 1001, processor 1002, and communication interface 1003 can communicate with each other through an internal interface.

[0115] The processor 1002 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0116] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for operating a liftable peripheral drive rake thickener for coal washing.

[0117] In addition, this application also provides a computer program product, including a computer program or instructions, which, when executed, implement the above-described method for operating a liftable peripheral drive rake thickener for coal washing.

[0118] In the description of this specification, the references to "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0119] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0120] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0121] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0122] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0123] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A lifting peripheral drive rake thickener operating system for coal washing, characterized in that, include: The module includes a working condition sensing module, an intelligent decision-making module, a lifting transmission control module, and a safety early warning module; among them, The operating condition sensing module is used to collect data on coal slurry concentration, rake frame torque, slag layer thickness and equipment operating temperature. The intelligent decision-making module is used to construct a coal slurry settling prediction model based on the coal slurry concentration, rake torque, slag layer thickness and equipment operating temperature data, and generate equipment operation control commands. The lifting transmission control module is used to drive the hydraulic actuator to adjust the vertical height of the rake frame according to the equipment operation control command, so as to adapt to the processing requirements of different slag layer thicknesses. At the same time, combined with the real-time feedback of the rake frame torque, the speed of the peripheral transmission motor is dynamically adjusted to optimize the separation conditions of coal slime and water. The safety early warning module is used to monitor the coal slurry concentration and scraper torque data through a sliding window algorithm. When the scraper torque and coal slurry concentration exceed multiple sets of adaptability thresholds, corresponding strategies at different levels are triggered.

2. The operating system of a liftable peripheral drive rake thickener for coal washing according to claim 1, characterized in that, The operating condition sensing module includes a coal slurry water parameter acquisition unit, an equipment status monitoring unit, and a data processing unit. The coal slurry water parameter acquisition unit is used to collect coal slurry water concentration and flow rate data. The equipment status monitoring unit is used to collect rake frame torque, slag layer thickness, system oil temperature, and motor winding temperature. The data processing unit is used to filter, reduce noise, and convert the format of the collected data to generate a standardized dataset.

3. The operating system of a liftable peripheral drive rake thickener for coal washing according to claim 1, characterized in that, The intelligent decision-making module includes a model computation unit, a control command generation unit, and a model training unit. The model computation unit, based on a standardized dataset, constructs a coal slurry settling prediction model using a hybrid LSTM-GRU algorithm. The LSTM network extracts long-term time-series features of coal slurry concentration and equipment temperature changes over time, while the GRU network optimizes the short-term correlation prediction between slag layer thickness and scraper torque. The control command generation unit outputs the coal slurry settling trend based on the prediction model and generates targeted control commands for different scenarios. The model training unit performs incremental training weekly using newly added historical operating data to control prediction errors.

4. The operating system of a liftable peripheral drive rake thickener for coal washing according to claim 1, characterized in that, The lifting transmission control module includes a hydraulic actuator, a transmission speed control unit, and a closed-loop feedback unit. The hydraulic actuator, based on the height adjustment requirements of the rake frame in the control command, drives the rake frame to perform vertical lifting movements via a hydraulic drive structure, adapting to processing needs with different slag layer thicknesses. The transmission speed control unit, based on the speed control requirements in the control command, adjusts the operating speed of the peripheral drive motors. Simultaneously, it dynamically adjusts the speed parameters based on real-time feedback information of the rake frame torque, optimizing the separation conditions of coal slime and water. The closed-loop feedback unit compares the actual rake frame height, the actual motor speed, and the target parameters in the control command in real time. When there is a deviation between the actual operating state and the command requirements, it automatically corrects the execution parameters of the hydraulic drive and speed adjustment.

5. The operating system of a liftable peripheral drive rake thickener for coal washing according to claim 1, characterized in that, The safety early warning module includes a parameter monitoring unit, an anomaly determination unit, and an early warning push unit. The parameter monitoring unit uses a sliding window algorithm to monitor changes in coal slurry concentration and scraper torque data. By calculating the rate of change of scraper torque and coal slurry concentration within each window, it quickly identifies sudden parameter mutations. The anomaly determination unit uses multiple preset adaptability thresholds, combined with the parameter change trends and durations monitored by the sliding window, to logically verify and identify abnormal states of the thickener equipment. The early warning push unit triggers a corresponding level of early warning response based on the abnormal state and pushes the early warning information to management personnel via SMS or an app.

6. The operating system of a liftable peripheral drive rake thickener for coal washing according to claim 1, characterized in that, The LSTM-GRU hybrid algorithm includes: LSTM algorithm for long-term feature extraction: ； ； ； ； ； ； GRU algorithm for short-term association prediction: ； ； ； ； Feature fusion using LSTM and GRU algorithms: ； Output prediction results: ； in, The proportion values ​​of long-term time-series information such as coal slurry concentration and equipment temperature from the previous moment are retained. For activation functions; Here is the weight matrix for the forget gate; for Long-term time-series characteristics of coal slurry water concentration and equipment temperature at any given moment; for Real-time collection of coal slurry water concentration data; for The equipment operating temperature data is collected continuously. For the bias term of the forget gate; This is the input ratio value for the current coal slurry concentration and equipment temperature information. This is the weight matrix of the input gate; This is the bias term for the input gate; These are candidate values ​​for new temporal characteristics of the current coal slurry concentration and equipment temperature. It is the hyperbolic tangent activation function; This is the weight matrix for the candidate cell states; This is a bias term for the candidate cell state; for The core characteristics of coal slurry water concentration and equipment temperature over a long period of time; For element-wise multiplication; for Long-term temporal characteristics of time; This represents the output ratio of long-term temporal features to the hidden state; This is the weight matrix of the output gate; This is the bias term for the output gate; for The final output is the long-term characteristics of the coal slurry water concentration and equipment temperature extracted at any time. This is the reset ratio for short-term information such as slag layer thickness and rake frame torque in the previous hidden state. To reset the weight matrix of the gate; for Short-term characteristic information of slag layer thickness and rake frame torque at any given time; for Data on slag layer thickness collected in real time; for Real-time collected rake frame torque data; To reset the door's bias; This represents the fusion ratio between the previous hidden state and the current candidate hidden state. To update the weight matrix of the gate; To update the gate's bias term; This is a candidate value for the new short-term correlation characteristics of the slag layer thickness and the rake frame torque at the current moment; Let be the weight matrix of the candidate hidden states; The bias term for the candidate hidden state; for The final output is the short-term correlation characteristics between the slag layer thickness extracted at any time and the rake frame torque; for The hidden state after fusing long-term temporal features and short-term correlation features; This is a vector concatenation function; for Key parameters related to the prediction of coal slurry water settling at specific times; To predict the weight matrix of the output layer; This is the bias term for predicting the output layer.

7. A method for operating a liftable peripheral drive rake thickener for coal washing, which can be applied to the operating system of a liftable peripheral drive rake thickener for coal washing as described in any one of claims 1-6, characterized in that, include: Collect data on coal slurry concentration, rake frame torque, slag layer thickness, and equipment operating temperature; After processing the data on coal slurry concentration, rake frame torque, slag layer thickness and equipment operating temperature, the data are input into the coal slurry settling prediction model constructed by the LSTM-GRU hybrid algorithm to generate equipment operation control commands. According to the equipment operation control command, the hydraulic actuator is driven to adjust the vertical height of the rake frame, adjust the operating speed of the peripheral drive motor, and dynamically adjust the speed parameters in combination with the real-time feedback information of the rake frame torque to continuously optimize the separation conditions of coal slime and water. At the same time, the changes in coal slime water concentration and rake frame torque data during the control process are monitored through the sliding window algorithm. Based on the changes in coal slurry concentration and rake frame torque data, combined with multiple preset adaptability thresholds, the system uses logic verification to judge and identify abnormal states of the thickener equipment, triggers corresponding level of early warning response, and pushes the early warning information to management personnel via SMS or APP.

8. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement the operating method of the liftable peripheral drive rake thickener for coal washing as described in claim 7.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When a computer program or instruction is executed, it implements the operating method of a liftable peripheral drive rake thickener for coal washing as described in claim 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When a computer program or instruction is executed, it implements the operating method of a liftable peripheral drive rake thickener for coal washing as described in claim 7.

Images