Underground coal machine working condition monitoring system and method based on operation state

Abstract

The invention discloses an underground coal machine working condition monitoring system and method based on a running state, particularly relates to the technical field of coal machine working condition monitoring, and is used for solving the problem of poor identification of damage caused by abnormal working conditions. Sampling alignment and normalization processing are carried out on state signals of an underground coal machine, a high-precision time sequence tensor input basis is constructed, a feature vector sequence is constructed in combination with dynamic response gradient, disturbance coupling and abnormal activeness indexes, and dynamic prediction of the working condition of the coal machine is achieved through a gating cycle structure model. The method comprises the following steps: constructing a mapping relation between working conditions and unit time yield loss, energy consumption increase and scheduling delay, quantifying economic influence of future operation states, generating an economic loss trend sequence, identifying a high-loss risk window, mining a state propagation chain and a synchronous variation probability according to a scheduling dependency graph, screening intervention points and formulating a multi-path intervention strategy, the implementation cost and the response effect are evaluated, and the intelligent sensing, prediction and scheduling intervention capabilities of the coal machine in a complex underground environment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coal machine working condition monitoring, more particularly, the present application relates to an underground coal machine working condition monitoring system and method based on operating state. BACKGROUND

[0002] The underground coal machine, also known as a coal mining machine, is the core equipment for underground coal mining operations, mainly used for coal seam cutting, loading and transportation operations. With the development of intelligent and deep mining in coal mines, the underground working environment is becoming increasingly complex. Coal machines operate under complex conditions such as high humidity, high temperature, high dust and strong impact load. They often face unstable factors such as severe vibration, hydraulic fluctuation and sudden load changes. These complex conditions not only easily cause equipment wear and tear and efficiency decline, but also induce safety accidents.

[0003] The existing technology has the following deficiencies: Underground coal machine working condition monitoring mainly relies on static single-point signals for anomaly detection, which cannot reflect the evolution trend of the state in the time dimension, leading to potential faults that cannot be warned in advance. Moreover, it mainly focuses on the physical state of the equipment, lacks systematic evaluation of economic consequences such as production loss, energy consumption fluctuation and scheduling delay caused by working condition changes, and cannot provide cost-driven basis for intervention strategies. At the same time, it lacks scheduling behavior based on state propagation relationship, causing the intervention measures to be disconnected from the actual impact source, which easily leads to resource allocation imbalance and scheduling efficiency decline, affecting the efficiency of underground coal machines. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides an underground coal machine working condition monitoring system and method based on operating state to solve the problem of poor damage identification of working condition abnormalities in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions: The underground coal machine working condition monitoring method based on operating state comprises the following steps: Collecting underground coal machine operating state signals and performing sampling alignment and nonlinear normalization mapping under a unified time reference to construct a structured time series tensor; Based on the time series tensor, extracting dynamic response gradient factors, disturbance coupling factors and abnormal activity factors, constructing state feature vectors of time windows, and modeling the state evolution trend according to the gated recurrent structure to identify the working condition category of the coal machine in the future time period; According to the working condition category and the historical response experience of the equipment, the corresponding relationship between the working condition category and the unit time output loss, energy consumption increase and scheduling delay factor is established, the economic impact factor set is generated and bound to the predicted time node; Calculate the total economic loss value for each time window based on three types of economic factors, and summarize them to form an economic loss trend sequence, identify high-risk windows, and mark potential intervention areas based on the working condition category; Based on the economic loss trend sequence and high-risk windows, a scheduling dependency graph is constructed to determine the synchronous mutation probability nodes, screen out intervention points, generate intervention strategy paths, and evaluate the intervention strategy paths; Visualize intervention strategy results, status trend charts, and economic losses, and deploy and execute strategic recommendations as work orders.

[0006] In a preferred embodiment, the operating status signal of the underground coal mining machine is collected, and sampling alignment and nonlinear normalization mapping under a unified time reference are performed to construct a structured time series tensor. The specific process is as follows: The sensor group includes a three-phase current acquisition unit installed at the output end of the cutting motor, pressure transmitters and temperature sensors arranged in the front and rear chambers of the hydraulic propulsion cylinder, a three-axis vibration sensor installed on the drum structure, and a speed encoder integrated into the propulsion mechanism; The acquisition frequency of all sensor channels is uniformly set to a period synchronized with the clock signal of the coal machine control host, and time interpolation is used for completion and synchronization correction; All sensor channels are mapped and transformed using a nonlinear amplitude normalization strategy with interval compression; After completing the amplitude normalization process, a state tensor sequence is constructed using a sliding window of fixed time length. All sensor channels are arranged in chronological order to form a state matrix. The rows of the matrix represent the channel numbers, and the columns of the matrix represent the time series positions. The matrix is ​​the time series tensor corresponding to the sliding window.

[0007] In a preferred embodiment, the dynamic response gradient factor, disturbance coupling factor, and abnormal activity factor are extracted based on the time series tensor. The specific process is as follows: Scan the continuous signal data of the selected channel, calculate the numerical changes between all two adjacent data points in the entire time window, record the absolute change amplitude between each pair of adjacent data points as an instantaneous gradient value, and select the maximum value of all instantaneous gradient values ​​in the time window as the dynamic response gradient factor of the channel; Compare whether the two channels show the same change trend at each sampling moment, count the number of data point pairs with the same change trend within the entire window period, and use the proportion as the disturbance coupling factor of the channel pair; Within the time window, the numerical change between two adjacent sampling points is calculated one by one to determine whether it exceeds the disturbance threshold set for the channel. All data points that meet the condition of the change exceeding the disturbance threshold are regarded as a high-frequency disturbance event. The total number of high-frequency disturbance events in the time window is counted and compared with the total number of sampling points in the time window to serve as the abnormal activity factor. The dynamic response gradient factor is used to measure the maximum change intensity of a single channel within a time window, indicating the degree of sudden change in the operating state of the component corresponding to the channel; The perturbation coupling factor is used to measure the degree of synchronization change between different channels; The abnormal activity factor is used to measure the proportion of each channel in the high-frequency disturbance area within the time window.

[0008] In a preferred embodiment, a state feature vector of a time window is constructed, and the state evolution trend is modeled according to a gated loop structure to identify the operating condition category of the coal mining machine in the future time period. The specific process is as follows: In each time window, the dynamic response gradient factor, disturbance coupling factor and abnormal activity factor characteristics are calculated in each channel or channel pair and then uniformly spliced ​​to form the state feature vector of the time window; After constructing the feature vectors of all time windows, they are combined in chronological order to form a continuous feature sequence, which is used as the dynamic behavior trajectory input during the operation of the coal machine. The operating condition classification model is used for classification and prediction processing. The operating condition classification model uses a sequence modeling algorithm based on a gated loop structure. The model input is a sequence of state feature vectors in several continuous time windows, and the output is the predicted result of the coal machine's operating condition category in the future time period. Working condition categories include normal operation, light load operation, heavy load operation, high frequency vibration, hydraulic fluctuation and shutdown warning categories.

[0009] In a preferred embodiment, based on the operating condition category and historical equipment response experience, a correspondence is established between the operating condition category and the unit time output loss, energy consumption increase, and scheduling delay factor. A set of economic impact factors is generated and bound to the prediction time node. The specific process is as follows: Equipment historical response experience includes equipment historical data analysis and expert knowledge annotation. The three types of economic factors include unit time output loss factor, unit time energy consumption increase factor, and unit time scheduling delay factor. Under normal operating conditions, the standard output value of the coal machine per unit time is recorded as the benchmark capacity. Historical data of the target state is extracted, and all window time periods under the corresponding operating condition category are screened out. The actual average output of the coal machine in all time periods is calculated. The average output under the target state is subtracted from the benchmark capacity to obtain the output loss value. The output loss value is then converted into the ratio of the benchmark capacity to obtain the unit time output loss factor. The total energy consumption per unit time of the coal machine under normal conditions is calculated and used as a benchmark value, including the sum of the power consumption of the cutting motor, the power consumption of the traction system, and the power consumption of the hydraulic system. For the target operating state, the time window data corresponding to the historical state is extracted, and the mean values ​​of the current, pressure, and temperature channels per unit time are calculated respectively, and converted into total energy consumption. The average energy consumption per unit time under the target state is obtained. The difference between the average energy consumption under the target state and the total energy consumption under normal conditions is compared, and the difference is converted into a ratio with the benchmark value, and the ratio is used as the energy consumption increase factor; During the normal operation cycle of the coal machine, the average advance distance per unit time, the number of cutting tasks or the production rhythm are counted and used as a reference for the standard scheduling time series. For the target operating state, historical operating condition records are extracted, and the actual task pause time, delay time or task interruption time caused by the target operating state are counted. The average cumulative delay time per unit time is calculated and used as the unit time scheduling delay factor for the target operating state. The predicted working condition categories are bound together, and within each time window, the working condition category corresponding to the time window and the three types of economic factor values ​​obtained by mapping are recorded to generate a sequence of economic impact factor sets.

[0010] In a preferred embodiment, the total economic loss value of each time window is calculated based on the three types of economic factors, and the economic loss trend sequence is summarized to identify high-risk windows. The potential intervention areas are marked in combination with the working condition category. The specific process is as follows: For each time window, based on the three types of economic factors bound to it, losses are calculated from three dimensions: capacity loss, electricity cost, and dispatch cost, and the sum is taken as the total economic loss value of the time window; Taking the window number as the time axis, summarize the economic loss values ​​of all time windows in order to build a complete time series structure as the economic loss trend series; Arrange the economic loss values ​​of each time window in the order of collection to form a data sequence with time as the horizontal axis and economic loss as the vertical axis; A risk threshold is set based on the loss trend sequence, and the high quantile of the economic loss trend sequence is selected as the high-risk judgment standard. The entire economic loss trend sequence is traversed, and the window numbers of all economic loss values ​​higher than the risk threshold are extracted. They are determined as high-risk windows and marked as potential intervention areas.

[0011] In a preferred embodiment, for each time window, losses are calculated from three dimensions: capacity loss, power cost, and dispatch cost, based on the three types of economic factors bound to it. The sum is taken as the total economic loss value of the time window. The specific process is as follows: Calculate the theoretical output value loss caused by the decline in production capacity during the time window based on the unit time output loss factor and the normal coal mining quota; Calculate the additional energy costs incurred within the time window based on the energy consumption increase factor per unit time and the equipment's electricity price per kilowatt-hour; Calculate the labor and equipment idle costs caused by scheduling delays within the time window based on the unit time scheduling delay factor and the scheduling labor cost benchmark; Add the three results together to form the total economic loss value within the time window.

[0012] In a preferred embodiment, a scheduling dependency graph is constructed based on the economic loss trend sequence and the high-risk window, the synchronous mutation probability nodes are determined, the intervention points are screened, and the intervention strategy path is generated. The specific process is as follows: Define a set of nodes in the scheduling dependency graph. Each node corresponds to an independent device unit. The edge set in the scheduling dependency graph is used to indicate the existence of a scheduling dependency or operation linkage relationship between two nodes. The direction of the edge indicates the impact flow. The weight of the edge is set to the degree of state correlation, which is calculated based on the probability of synchronous mutation of two channels in the high-risk window in the historical data; For each window marked as high-risk, we analyze its corresponding state feature vector and operating condition category sequence to determine the main operating condition type that causes economic losses. Using the rising economic loss factor as the signal source, we expand outward in the graph and track the associated scheduling dependency graph nodes. Identify the node groups with continuous abnormal responses and determine the intervention points based on the status propagation priority score and status impact index; After identifying the intervention points, an intervention strategy path is generated based on the operation control rules, equipment adjustment parameters and historical intervention effect records.

[0013] In a preferred embodiment, the intervention strategy path is evaluated, and the specific process is as follows: The implementation cost is the composite cost of labor costs, material costs and equipment downtime losses required to complete all actions in the intervention strategy path; Response latency is the average time interval from the issuance of a strategy to the actual effect feedback; The historical success rate is the ratio of the intervention strategy path that successfully reduces economic losses to below the set threshold under historical operating conditions.

[0014] The underground coal mining machine condition monitoring system based on the operating status is used to implement the above-mentioned underground coal mining machine condition monitoring method based on the operating status, including: The coal mining machine data acquisition module collects the operating status signals of underground coal mining machines and performs sampling alignment and nonlinear normalization mapping under a unified time base to construct a structured time series tensor. The operating condition classification and acquisition module extracts the dynamic response gradient factor, disturbance coupling factor, and abnormal activity factor based on the time series tensor, constructs the state feature vector of the time window, and models the state evolution trend based on the gated loop structure to identify the operating condition category of the coal mining machine in the future time period; The economic impact module establishes a correspondence between the operating condition category and the unit time output loss, energy consumption increase, and scheduling delay factor based on the operating condition category and historical equipment response experience, generates a set of economic impact factors, and binds them to the forecast time node; The risk identification module calculates the total economic loss value of each time window based on three types of economic factors, summarizes them into an economic loss trend sequence, identifies high-risk windows, and marks potential intervention areas based on the working condition category; The path intervention module builds a scheduling dependency graph based on the economic loss trend sequence and high-risk windows, determines the synchronous mutation probability nodes, screens out intervention points, generates intervention strategy paths, and evaluates the intervention strategy paths; The coal machinery adjustment module visualizes intervention strategy results, status trend charts, and economic losses, and deploys and executes strategy recommendations in the form of work orders.

[0015] The technical effects and advantages of the present invention are as follows: The present invention constructs a high-precision time series tensor input foundation by performing sampling alignment and normalization processing on key state signals such as cutting motor current, voltage, vibration, hydraulic oil pressure and drum temperature under a unified time base. It constructs a feature vector sequence by combining indicators such as dynamic response gradient, disturbance coupling and abnormal activity, and uses a gated loop structure model to realize dynamic prediction of coal machine operating conditions, improve the forward-looking recognition capability of abnormal operating conditions, and quantify the economic impact of future operating conditions by constructing a mapping relationship between operating conditions and unit time production loss, energy consumption increase and scheduling delay. It generates an economic loss trend sequence and effectively identifies high-loss risk windows. It further mines the state propagation chain and synchronous mutation probability based on the scheduling dependency graph, accurately screens intervention points and formulates multi-path intervention strategies, evaluates their implementation cost and response effect, ensures the feasibility and timeliness of the scheduling plan, and finally combines a visualization platform to realize intuitive display and linkage deployment of monitoring results and strategies, significantly improving the intelligent perception, prediction and scheduling intervention capabilities of the coal machine system in complex underground environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1The figure is a flow chart of the underground coal mining machine operating condition monitoring method based on the operating status of the present invention.

[0017] Figure 2 It is a structural schematic diagram of the underground coal machine working condition monitoring system based on the operating status of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 As shown, the underground coal machine operating condition monitoring method based on the operating status includes the following steps: Collect the operating status signals of underground coal mining machines, perform sampling alignment and nonlinear normalization mapping under a unified time base, and construct a structured time series tensor; Based on the time series tensor, the dynamic response gradient factor, disturbance coupling factor, and abnormal activity factor are extracted to construct the state feature vector of the time window. The state evolution trend is modeled based on the gated loop structure to identify the operating condition category of the coal mining machine in the future time period. Based on the operating condition category and historical equipment response experience, a correspondence is established between the operating condition category and unit time output loss, energy consumption increase, and scheduling delay factor. A set of economic impact factors is generated and bound to the forecast time node. Calculate the total economic loss value for each time window based on three types of economic factors, and summarize them to form an economic loss trend sequence, identify high-risk windows, and mark potential intervention areas based on the working condition category; Based on the economic loss trend sequence and high-risk windows, a scheduling dependency graph is constructed to determine the synchronous mutation probability nodes, screen out intervention points, generate intervention strategy paths, and evaluate the intervention strategy paths; Visualize intervention strategy results, status trend charts, and economic losses, and deploy and execute strategic recommendations as work orders.

[0020] Step 1: Collect underground coal mining machine operating data and construct a time series window tensor. This involves collecting multi-channel operating status parameters of the underground coal mining machine's cutting motor, hydraulic propulsion cylinder, drum vibration unit, and the entire machine's motion system. The collected signals are then time-synchronized and nonlinearly normalized to construct a structured time window tensor. The specific process is as follows: A sensor group is formed by placing various types of industrial sensors on the key components of the coal machine, and multi-channel data collection is performed on the operating status of the coal machine with a fixed sampling period. The sensor group includes but is not limited to: a three-phase current acquisition unit installed at the output end of the cutting motor, used to collect the cutting current fluctuation parameters; a pressure transmitter and temperature sensor placed in the front and rear chambers of the hydraulic propulsion cylinder, used to obtain the propulsion oil pressure and return oil temperature respectively; a three-axis vibration sensor installed on the drum structure, used to obtain the mechanical impact acceleration value during cutting; a speed encoder integrated in the propulsion mechanism, used to collect the forward speed and displacement change information of the coal machine. The acquisition frequency of all sensor channels is uniformly set to a 50-millisecond period synchronized with the clock signal of the coal machine control host, that is, one data point is collected for each channel every 50 milliseconds, and the starting acquisition time is aligned with the rising edge of the control host clock signal; Using the unified clock signal provided by the coal machine master control system as the reference time source, all channel data is timestamped and time-aligned. This alignment involves mapping the timestamps collected from different channels to the sampling time of the master control system's clock signal, ensuring the timing consistency of all channel data at the same time point. In the event of timestamp offset, time interpolation is used to complete and synchronize the offset.

[0021] The specific method of time interpolation is as follows: if a channel fails to complete effective collection at the target time point, the numerical change trend of the two valid data points before and after the channel is used to perform linear extrapolation according to the sampling period ratio to interpolate the fitting value at the target moment. For example, in the propulsion oil pressure channel, if the oil pressure data at a certain target moment is missing, the collected data within 100 milliseconds before and after can be used to fit the pressure value at that point to ensure that all channels have no gaps under the same time axis.

[0022] After time alignment, the data still have physical dimension differences, such as current expressed in amperes, oil pressure in megapascals, vibration in meters per second squared, temperature in degrees Celsius, and speed in meters per second. To eliminate the influence of physical scale and improve the comparability of multi-channel data, a nonlinear amplitude normalization strategy with interval compression is used to map all channels. The normalization rule is as follows: a central value and sensitive interval are set based on the steady-state operating range of each channel under standard operating conditions, and a nonlinear S-shaped mapping function is constructed to map the original value of each channel to the interval (0, 1). The mapping process retains high resolution for edge values ​​and maintains symmetrical buffering for the median interval. For example, in the cutoff current channel, if the standard operating range is 400 to 800 amperes, the central value is set to 600 amperes. The boundary values ​​are compressed and mapped to bilateral values ​​close to 0 and 1. After amplitude normalization, a state tensor sequence is constructed using a sliding window of fixed time length. The sliding window length is set to 5 seconds, and the sliding step is set to 1 second. Each window contains 100 consecutive sampling points (based on a 50-ms sampling period). Within each window, all channels are arranged in chronological order to form a state matrix. The rows of the matrix represent the channel number, and the columns represent the time series position. The matrix is ​​the time series tensor corresponding to that window. For example, in the nth window, if there are 10 acquisition channels, the window tensor will have 10 rows and 100 columns, representing the complete dynamic change process of each channel within the 5-second period.

[0023] The periodic acquisition cycle and sliding window length are set based on the dynamic response characteristics of the coal machine's operating conditions. The acquisition cycle (50 milliseconds) covers the high-frequency vibration changes in the typical coal and rock cutting process, while the sliding window length (5 seconds) covers the typical propulsion rhythm and hydraulic stability time constant range. This can fully capture the data changes during the entire process of the coal machine transitioning from low load to high load and from stable cutting to sudden impact. Each constructed time series tensor will serve as the basic input unit for the subsequent state recognition module and will be arranged in chronological order to form a multi-channel time series tensor sequence for the operation process. All time series tensor data is cached in the control system's local storage area in the form of a three-dimensional array, where the first dimension is the time window sequence number, the second dimension is the channel number, and the third dimension is the sequence of time points within the window. For example, the storage structure is: window index, channel number, time point sequence number. All tensors are arranged in ascending order of acquisition time, supporting subsequent fast access by time index.

[0024] It's important to note that sensor calibration is performed daily and periodically, depending on underground operating conditions. For example, the cutting current sensor is benchmarked daily before mining by a fixed current output from the control cabinet. Hydraulic sensors undergo weekly on-site full-scale calibration tests, and vibration sensors undergo monthly amplitude-frequency response verification on a standard vibration table. If the sensor detection error exceeds the set tolerance (e.g., 2%), the system automatically issues an alarm and switches to the reference channel or implements interpolation redundancy compensation.

[0025] Step 2: Extract characteristic pattern information during coal machine operation based on the constructed multi-channel time series tensor sequence and identify the operating condition category in the future time period. The specific process is as follows: In each time series tensor generated in step 1, the structural features and dynamic behavior of the data change trajectories of all acquisition channels are extracted and characterized. A multidimensional state feature vector of the coal machine within the window time period is constructed. This feature vector is used to describe the coal machine's operating status, including load response, disturbance coupling, vibration fluctuation, and structural coordination within the time window, providing a highly timely and interpretable feature basis for subsequent operating condition identification. The first category is the dynamic response gradient factor, which measures the maximum change intensity of a single channel within a time window and indicates the degree of sudden change in the operating state of the component corresponding to the channel. For example, in the cutting current channel, if a significant load increase occurs within a 5-second window, the current value fluctuation amplitude of the channel within the window will increase significantly, and the corresponding dynamic response gradient factor value will also increase, indicating the possibility of sudden changes in coal seam hardness and increased cutting resistance. The dynamic response gradient factor is calculated as follows: the continuous signal data of the selected channel is scanned, the numerical changes between all two adjacent data points in the entire time window are calculated, and the absolute change amplitude between each pair of adjacent data points is recorded as an instantaneous gradient value. Within the time window, the maximum value of all instantaneous gradient values ​​is selected as the dynamic response gradient factor of the channel. If the maximum change value exceeds the preset dynamic fluctuation threshold, it means that the channel has experienced sudden load, energy shock or fault precursor behavior within the current window period.

[0026] For example, if a current value is collected every 50 milliseconds in a 5-second window in the cutoff current channel, then the change between two adjacent values ​​is calculated among 100 consecutive current values, and a total of 99 gradient values ​​are obtained. The one with the largest change is taken as the gradient factor of the channel in this window.

[0027] The second category is the disturbance coupling factor, which is used to measure the degree of synchronous changes between different channels and reflect the structural coordination and load conduction behavior between equipment. Multiple channel pairs with physical coupling relationships are selected, such as propulsion oil pressure and forward speed, cutting current and drum vibration, and their change directions and rates within the window are compared to calculate the consistency of their disturbance changes. If multiple coupled channels show an upward or downward trend at the same time period, it indicates that their behavior has the characteristics of structural coupling response; otherwise, it indicates loose coupling or the existence of local anomalies. A specific calculation method for the disturbance coupling factor is as follows: select any two channels that are physically linked, such as propulsion oil pressure and coal machine forward speed, or cutting current and drum vibration; within the time window, judge the change trend of the sampling value sequences of the two channels respectively, that is, judge whether the change direction of each pair of adjacent data points is rising, falling or unchanged, compare whether the two channels show the same change trend at each sampling moment (for example, rising or falling at the same time), count the number of data point pairs with consistent change trends within the entire window period, and use their proportion as the disturbance coupling factor of the channel pair. If the proportion is higher than the set coupling threshold (for example, more than 70%), it is considered that the channel pair has significant disturbance coupling characteristics within the time window.

[0028] For example, if a total of 100 data points are collected for propulsion oil pressure within a 5-second time window, and the same number of data points are collected for forward speed, and trend comparison shows that the two channels change in the same direction at 78 moments, then the disturbance coupling factor of this channel pair is 78%, which can be judged as a strong coupling response state.

[0029] The third category is the abnormal activity factor, which measures the proportion of each channel in the high-frequency disturbance area within the window, reflecting the degree of instability during equipment operation. By setting the disturbance threshold for each channel, the frequency of signal changes exceeding the threshold per unit time is determined, thereby identifying operating condition characteristics such as increased cutting vibration, frequent hydraulic fluctuations, and gear meshing shock. When the abnormal activity factor maintains a high value on certain channels for a long time, it may indicate potential problems such as component aging or poor lubrication. A specific calculation method for the abnormal activity factor is as follows: first, a disturbance threshold is set for each channel. The disturbance threshold is obtained based on the empirical statistics of the change rate or fluctuation amplitude of the channel under normal operating conditions. Within the time window, the numerical change between two adjacent sampling points is calculated one by one to determine whether it exceeds the disturbance threshold set for the channel. All data points that meet the condition that the change exceeds the disturbance threshold are regarded as a high-frequency disturbance event. The total number of high-frequency disturbance events in the time window is counted and proportionally converted to the total number of sampling points in the window. The proportional value is used as the final result of the abnormal activity factor. The higher the value, the more unstable the channel is within the window period. For example, if the drum vibration channel collects a total of 100 data points in a window, and each change amplitude exceeding a certain acceleration threshold is set as a high-frequency disturbance, and a total of 24 high-amplitude vibration events occur, then the abnormal activity factor of the drum vibration channel in this window is 24%.

[0030] In each time window, the above three types of features are calculated in each channel or channel pair and then concatenated to form the state feature vector of the window. The length of the feature vector depends on the number of channels, the number of channel pair combinations and the feature type. For example, for 10 channels and 15 channel pairs, the feature vector may contain more than 100 dimensions of information. After completing the construction of the feature vectors of all time windows, they are combined in chronological order to form a continuous feature sequence. This sequence is the dynamic behavior trajectory input during the operation of the coal machine and enters the working condition classification model for classification prediction processing.

[0031] The operating condition classification model uses a sequence modeling algorithm based on a gated loop structure, which has the ability to model state evolution trends in long time series. The model input is a sequence of state feature vectors of several continuous time windows, and the output is the predicted operating condition category of the coal machine in the future time period (operating condition category). The operating condition category is divided into several typical states, including but not limited to normal operation, light load operation, heavy load operation, high-frequency vibration, hydraulic fluctuation and shutdown warning. Each operating condition category corresponds to the key state characteristics in the actual operation behavior of the coal machine; For example, if the cutting current in multiple consecutive windows shows a clear upward trend, while the vibration channel activity increases significantly and the oil pressure stability decreases, the model may output a heavy-load operation or high-impact risk label, indicating that the coal machine is currently in a complex operating condition and requires attention to the risks of equipment thermal load and structural loss. During model training, supervised learning is performed using a set of operating condition samples consisting of historical coal mining machine operation data and field annotated records to ensure that the model can identify state change patterns in real production scenarios. After model training is completed, it is deployed in underground or surface edge servers for real-time inference. Each time a new sliding window data collection cycle is completed, the operating condition classification results are automatically updated and a corresponding prediction label sequence is generated. The generation cycle of the prediction label is consistent with the time window step, ensuring that the model output keeps pace with data updates.

[0032] Step 3: Based on the identified coal machine operating condition categories, the corresponding relationship between the operating condition and the economic impact factor is established, and a factor set for economic loss assessment is generated. The specific implementation is as follows: In step 2, each time window is classified into a specific coal-fired power plant operating condition category, which identifies the current operating state of the coal-fired power plant. Each operating condition category corresponds to varying degrees of energy intensity, production capacity fluctuation, and operation and maintenance intervention costs. These status labels need to be further mapped into measurable economic impact indicators. Based on historical equipment response experience (equipment historical data analysis and expert knowledge annotation), a one-to-one correspondence table is established between the operating condition categories and their corresponding three major economic factors. The three economic factors are: unit time output loss factor, unit time energy consumption increase factor, and unit time scheduling delay factor. The specific definitions are as follows: The first category is the unit-time production loss factor, which measures the degree of production capacity reduction under a specific operating condition compared to ideal operating conditions (i.e., normal operation). The value of this factor is obtained through statistical regression based on historical production records. For example, if the coal machine's propulsion speed drops significantly under heavy load operation, and the coal production per unit time is only 70% of the normal value, the production loss factor for this condition is set to 30%, indicating that the production capacity will be lost per hour equivalent to 30% of normal operation. The unit-time yield loss factor is generated as follows: Under normal operating conditions, the standard output value of the coal machine per unit time (e.g., per hour) is calculated and recorded as the baseline capacity. Historical data is extracted for the target operating condition (e.g., overload operation), and all window time periods under the corresponding operating condition are screened. The actual average output of the coal machine during these time periods is calculated. Then, the average output under the target condition is subtracted from the normal output to obtain the unit-time yield loss relative to the baseline. This yield loss value is then converted to the baseline capacity to obtain the unit-time yield loss factor for that condition. The result is usually expressed as a percentage. If factors such as a significant decrease in propulsion speed, drum jamming, and hydraulic instability are present in this condition, correction coefficients can be introduced based on characteristic indicators for manual optimization and compensation. For example, if the coal machine's hourly coal production under normal conditions is 500 tons, but under overload operation, it actually only produces an average of 350 tons per hour, the production loss is 150 tons, corresponding to a unit-time yield loss factor of 30%.

[0033] The second type is the energy consumption increase factor per unit time, which is used to reflect the increase in energy consumption of the power and hydraulic systems of the coal machine under specific operating conditions compared to normal conditions. For example, under abnormal vibration conditions, due to component resonance or cutting impact, the motor operation is unstable, and energy consumption usually increases significantly. If the current and hydraulic power in this state increase by an average of 20% compared to the normal state, the corresponding energy consumption increase factor is set to 20%; The energy consumption per unit time increase factor is generated as follows: First, the total energy consumption per unit time of the coal machine under normal conditions is calculated and used as a baseline value. This includes the sum of the power consumption of the cutting motor, traction system, and hydraulic system. For the target operating state, window data corresponding to the historical state is extracted, and the average values ​​of energy-related channels such as current, pressure, and temperature per unit time are calculated. Then, using the equipment's electrical and hydraulic characteristics, these state values ​​are converted into total energy consumption, resulting in the average energy consumption per unit time under the target state. The difference between the energy consumption under the target state and the normal baseline energy consumption is then compared and converted into a ratio of this difference to the baseline value. This ratio is used as the energy consumption increase factor for that state, expressing the additional burden that state places on energy consumption. For example, if the coal machine's energy consumption per unit time is 200 kWh under normal conditions and increases to 240 kWh under abnormal vibration conditions, the energy consumption increase is 40 kWh, corresponding to a 20% energy consumption per unit time increase factor.

[0034] The third category is the unit time scheduling delay factor, which is used to evaluate the temporal impact of the state on the work rhythm, task progress, and human resource scheduling. This factor is obtained by statistically analyzing the duration of the working condition and the offset time of subsequent tasks. For example, in the shutdown warning state, if it usually triggers a maintenance pause or fault handling time of 5 to 10 minutes, the corresponding scheduling delay factor can be set to an average task delay of 8 minutes per hour; The unit-time scheduling delay factor is generated as follows: During the normal operation cycle of the coal machine, the average advance distance per unit time, the number of cutting tasks, or the production rhythm are calculated and used as a reference for the standard scheduling time series. For the target operating state, historical operating condition records are extracted to calculate whether the persistent occurrence of this state causes advance stoppage, maintenance insertion, or remote intervention. The actual task pause time, delay time, or task interruption duration caused by this state is analyzed, and the average cumulative delay time per unit time (e.g., 1 hour) is calculated. This time is used as the unit-time scheduling delay factor for this state. The result is expressed in minutes or seconds, reflecting the time offset impact of the scheduling system caused by the abnormal state. If the state triggers a system-level shutdown warning or linkage protection strategy, a delay amplification weight can be introduced to make the delay factor reflect the actual decision cost. For example, if the coal machine needs to suspend operation and perform remote confirmation and manual maintenance operations under the shutdown warning state, resulting in an average pause time of 10 minutes, and this state occurs on average once an hour, the unit-time scheduling delay factor is 10 minutes.

[0035] To ensure the stability and traceability of the mapping relationship, the setting of all factor values ​​must be comprehensively confirmed in combination with the following three types of information sources: On-site operation data comes from the historical operating condition sequence and corresponding production capacity and power consumption records during actual coal mining machine operation; Maintenance system records, including data from the equipment management system on fault response time, processing cycle, and repair time; Expert experience knowledge, which is the manual operation experience correction value given by coal mine operation and maintenance experts, dispatching supervisors, etc. based on long-term operation experience; After the above factors are constructed, they are bound to the working condition categories of the working condition forecast output in step 2. In each time window, the working condition category corresponding to the time window and the three types of economic factors obtained by mapping are recorded to generate a set of economic impact factors. The structure of the economic impact factor set sequence is a three-field data set arranged by time window number. Each record contains the operating condition category, output loss factor, energy consumption increase factor, and scheduling delay factor for that window. For example, for a window numbered W13, if its operating condition category is heavy load operation, the corresponding economic factor set for this window might be [heavy load operation, 30%, 20%, 5 minutes].

[0036] Step 4: Based on the constructed set of economic impact factors, a loss valuation function is established to quantify the cumulative economic losses of coal mining machinery in the future time period, and the high-risk economic sections during operation are identified in combination with the time distribution characteristics. The specific steps are as follows: For each time window, based on the three types of economic factors bound to it, losses are estimated from three dimensions: capacity loss, power cost, and dispatch cost. The sum of these is used as the economic loss value of the window. The specific implementation is as follows: Based on the unit time production loss factor and the normal coal mining quota, calculate the theoretical output value loss caused by the production capacity reduction in this window, and express the result in monetary units; based on the unit time energy consumption increase factor and the equipment operating electricity price per kilowatt-hour, calculate the additional energy cost incurred in this window; based on the unit time scheduling delay factor and the scheduling labor cost benchmark, calculate the labor and equipment idling costs caused by scheduling delays in this window. Add the above three results to form the total economic loss value in this window; For example, if a window is operating under heavy load, its production loss factor is 30%, its energy consumption increase factor is 15%, and its scheduling delay is 5 minutes per hour, combined with the sales price per ton of coal, the kilowatt-hour electricity price, the unit price of labor hours and other standards, it can be concluded that the total loss corresponding to the window is a specific monetary amount (such as the loss amount per hour).

[0037] Taking the window number as the time axis, summarize the economic loss values ​​of all time windows in order to build a complete time series structure as the economic loss trend sequence. The specific implementation is as follows: Arrange the economic loss values ​​of each time window in the order of collection to form a data series with time as the horizontal axis and economic loss as the vertical axis; Smoothing the sequence can eliminate the accidental interference caused by single-point spikes and make it more trend-stable. Connect the economic loss values ​​over time to generate an economic loss curve, showing the economic burden level of the coal machine at each stage during the entire operation cycle; By analyzing the trend curve, we can further identify high-loss sections, stable sections and inflection points.

[0038] A risk threshold is set based on the loss trend sequence. Typically, the high quantile of the economic loss trend sequence (e.g., 85%) is selected as the high-risk criterion. The entire economic loss trend sequence is traversed, and the window numbers of all windows with economic loss values ​​above the threshold are extracted. These windows are identified as high-risk windows. All high-risk windows are timestamped and recorded together with the corresponding working condition category and state feature vector. For example, during coal mining on a certain working day, if the economic loss between 9:00 and 9:45 in the morning is continuously higher than the set baseline threshold, and is accompanied by frequent high vibration and propulsion delay tags, the system will mark this time period as a potential intervention area.

[0039] Step 5, according to the generated economic loss trend sequence and the identified high-risk window, a multi-component operation state scheduling dependency graph is constructed, and a scheduling intervention strategy candidate set is generated combining with the working condition category and historical response behavior, and the adjustable task path and maintenance window cluster are output, the specific implementation is as follows: The scheduling dependency graph is used to represent the dependency structure between each running subsystem of the coal machine in task execution and working condition influence, which is established according to the coal machine structure topology, task execution sequence and working condition influence flow direction; The node set of the scheduling dependency graph is defined, including the key components and system modules of the coal machine, such as cutting motor, drum assembly, push oil cylinder, hydraulic pump station, traction system, cooling module, etc., each node corresponds to an independent device unit; The edge set of the graph is defined, which is used to represent the scheduling dependency or operation linkage relationship between two nodes, the direction of the edge represents the influence flow direction, for example, the current fluctuation of the cutting motor can affect the load of the hydraulic system, and the change of the hydraulic oil temperature can be fed back to the cooling system; The weight on the edge is set as the state correlation degree, which is calculated according to the synchronous variation probability of two channels in the high-risk window in the historical data, a multi-linkage structure graph is generated, forming a complete scheduling dependency graph, which is used to guide the chain design of intervention strategy.

[0040] After the scheduling dependency graph is constructed, the state change path in the high-risk window needs to be further tracked to identify the propagation chain that causes the economic loss to expand, and the controllable control points are determined; For each window marked as high risk, analyze its corresponding state feature vector and working condition category sequence to determine the main working condition type that causes economic loss; In the scheduling dependency graph, take the rise of the economic loss factor as the signal source, and expand outward in the graph to track the associated scheduling dependency graph nodes; Determine which paths have continuous abnormal response node groups through path analysis algorithm, i.e. there is a state propagation chain; Select the nodes with high state propagation priority score and large state influence degree index on the state propagation chain as intervention points to determine whether they have adjustment ability or maintenance possibility; All key points are summarized to form an operable intervention entry set.

[0041] The process of obtaining the synchronous variation probability is implemented as follows: First, select the time period set that has been identified as high economic loss window in the historical record, denoted as high-risk window set. These windows correspond to the time periods when the coal machine system appears abnormal economic indicators such as capacity decline, energy consumption increase or scheduling delay in the past operation, which have strong state disturbance characteristics, so they can be used as key observation samples for state linkage relationship identification; For any two channel parameters to be analyzed (such as cutting motor current and roller temperature), the following steps are performed: State variation marking is performed to preprocess the time series signal of each channel. First, the direction of parameter value change in each high-risk window is calculated, i.e., the local change sign of each window. If the parameter shows a continuous rising or falling trend in the current window, it is marked as a variation; if the change amplitude is small or the trend is not obvious, it is marked as stable; Synchronous variation judgment is performed. In each high-risk window, it is judged whether the two channels have occurred variation at the same time. If both channels are marked as variation in the same window, it is considered as a synchronous variation event of the channel pair in that window; All high-risk windows are traversed, and the number of synchronous variations of the two channels in all windows is counted. The number is divided by the total number of high-risk windows to obtain the synchronous variation probability value of the channel pair.

[0042] The higher the synchronous variation probability as the edge weight, the stronger the linkage between the two state channels in the abnormal state, which is suitable for being a candidate target for coordinated intervention in subsequent scheduling intervention, for example, if the synchronous variation probability of hydraulic propulsion pressure and vibration acceleration is significantly higher than that of other channel pairs, the hydraulic parameters are adjusted as an indirect control method when vibration anomaly occurs.

[0043] After identifying the intervention points, the system automatically generates multiple feasible intervention strategy paths according to the operation control rules, device adjustment parameters and historical intervention effect records, which are used for decision-making and deployment by the scheduling platform; The intervention strategy generation method is as follows: For each key intervention node, check the device control protocol to determine whether automatic adjustment, operation speed limit, parameter amplitude limit or temperature control switching functions are supported; Call the historical execution library to extract the successful intervention actions and their feedback effects in similar states, such as "limiting the flow of hydraulic pump pressure difference by 5% to alleviate oil temperature rise"; According to the current working condition trend prediction results, combine multiple intervention actions to form different strategy paths, and build a candidate strategy set; Each strategy path includes adjustment object, execution time, duration, expected effect and rollback condition to ensure controllability and rollback mechanism; Evaluate the implementation cost, response delay and historical success rate of all candidate strategies, and mark the preferred path.

[0044] Select the nodes with high state propagation priority score and large state impact degree index on the state propagation chain as intervention points, and judge whether they have adjustment ability or maintenance possibility. Two quantitative indexes are used for sorting and screening: The state propagation priority score is used to measure the leading role of a node in the state propagation path and is defined as follows: If a node is at the starting or middle position in multiple propagation paths and its state mutation occurs more frequently than other nodes, then its state propagation priority score is higher. The state propagation priority score is calculated by counting the number of times the node appears as the first mutation node or the intermediate transmission node in all observed propagation chains; The state impact index is used to measure the contribution of a node mutation to the final economic loss and is defined as follows: In all historical high-risk windows, the average cumulative loss caused by the occurrence of the node mutation is calculated, and this value is normalized and sorted among all nodes as the impact score.

[0045] The intervention criticality score is constructed by weighting the state propagation priority score and the state impact index. For example, the priority weight is set to 0.6 and the impact weight is set to 0.4. If the intervention criticality score exceeds the preset threshold (such as 0.65) and the node is currently in an abnormal state, it is identified as a critical node with priority intervention value.

[0046] After the intervention strategy pathways are generated, the following three quantitative assessments need to be performed on each pathway: The implementation cost is defined as the combined cost of labor costs, material costs, and equipment downtime losses required to complete all actions in the intervention strategy path. For example, if the path includes replacing a hydraulic pump, the average downtime and cost in its historical maintenance records are consulted. The response delay is defined as the average time interval from the issuance of the strategy to the feedback of the actual effect. The historical success rate is the ratio of the strategy path to successfully reducing economic losses to below the set threshold under similar historical working conditions. The records of the execution of the strategy path can be extracted from the database, and the number of successful attempts is counted as a percentage of the total number of attempts.

[0047] For example, in a high-risk window, if an abnormality in the hydraulic system causes increased vibration, three intervention plans can be output: one is to actively reduce the propulsion rate by 5%, the second is to switch the cooling mode to high-speed operation, and the third is to lower the pre-regulation pressure limit by 0.3 MPa, accompanied by an estimated effect improvement ratio and scheduling impact duration.

[0048] In addition to short-term interventions, for structural problems or component degradation trends, the system must combine task planning with maintenance plans, reserve maintenance windows, and output the optimal intervention path. The implementation method is as follows: Summarize the continuous high-risk window information and its corresponding equipment modules to determine whether there is a degradation trend; If it is determined to be a medium- to long-term performance degradation, the maintenance window priority is planned and matched with the subsequent task cycle; Inserting an adjustment path in the current scheduling plan, such as switching a standby propulsion path, adjusting a working face advancing sequence, or compressing a working cycle; Generating a maintenance candidate window time period and a task replacement path, and submitting the scheduling center for review and execution.

[0049] For example, if there are 100 high-risk windows in the history, and in 36 of them, the cutting motor current and the drum temperature both show significant rises or falls (i.e., consistent disturbance trends), then their synchronous variation probability is 36%, which can be used to assign a weight to the edge connection between them, indicating the degree of associated disturbance trends in the system under high-risk conditions.

[0050] Step 6: Based on the working condition classification results, economic loss trend sequences, and scheduling intervention strategies generated in the previous steps, a multi-level evaluation index atlas is constructed, and the structured visual expression of the evaluation results is realized. Through the scheduling platform interface, the task deployment and ground system linkage of the intervention strategy are realized, and the specific implementation is as follows: The working condition categories are expanded along the time axis, and the economic loss curve is formed by combining the economic loss calculation results. The state abnormal nodes corresponding to each scheduling dependent unit (such as the cutting motor, drum, and hydraulic propulsion system) are marked on the curve. On this basis, according to the distribution density of the loss value, the critical point of the speed increase, and the trend turning point, a multi-dimensional evaluation atlas is constructed, including but not limited to: dynamic working condition risk level graph (mapping working condition classification to risk level), economic loss intensity heat map (showing the economic pressure interval of each channel per unit time), and scheduling intervention feasibility grid graph (indicating the adjustment or maintenance strategies that can be triggered under the current working condition and their expected success), each evaluation atlas is aligned to the time reference, and is displayed in layers for combined display, facilitating cross-observation and comparative analysis of state evolution, risk outbreak, and strategy action timing; Further, combined with the key nodes marked in the visualization atlas and the preset strategy execution interface, the intervention strategy is converted into an executable task package through the scheduling platform interface and is sent to the surface management system. Each task package includes the identification number of the corresponding working condition, the recommended execution action (such as frequency reduction, hydraulic pressure limiting, shutdown inspection, etc.), the expected control period, and the operation feedback path. After accepting the task package, the system automatically deploys and transmits scheduling instructions based on the device connection state and operation permissions. During the deployment process, the platform will real-time feedback on the status of strategy execution, and embed the feedback signals into the state monitoring system to update the atlas display results, forming a closed-loop response process from state recognition, economic evaluation, strategy suggestion to scheduling execution, realizing the linkage of ground and underground management and control, and the cyclic driving of evaluation information.

[0051] It should be noted that the threshold values involved in the embodiments are determined and adjusted according to specific scenarios and needs.

[0052] The present invention constructs a high-precision time series tensor input foundation by performing sampling alignment and normalization processing on key state signals such as cutting motor current, voltage, vibration, hydraulic oil pressure and drum temperature under a unified time base. The characteristic vector sequence is constructed by combining indicators such as dynamic response gradient, disturbance coupling and abnormal activity, and the gated loop structure model is used to realize dynamic prediction of coal machine working conditions, improve the forward-looking recognition ability of abnormal operating conditions, and quantify the economic impact of future operating conditions by constructing a mapping relationship between working conditions and unit time production loss, energy consumption increase, and scheduling delay. The economic impact of future operating conditions is generated, and an economic loss trend sequence is generated to effectively identify high-loss risk windows. Further based on the scheduling dependency graph, the state propagation chain and synchronous mutation probability are mined, the intervention points are accurately screened and multi-path intervention strategies are formulated, the implementation cost and response effect are evaluated, and the feasibility and timeliness of the scheduling plan are ensured. Finally, the visualization platform is combined to realize the intuitive display and linkage deployment of monitoring results and strategies, significantly improving the intelligent perception, prediction and scheduling intervention capabilities of the coal machine system in complex underground environments.

[0053] Example 2: Underground coal machine condition monitoring system based on operating status, such as Figure 2 As shown, specifically including: The coal mining machine data acquisition module collects the operating status signals of underground coal mining machines and performs sampling alignment and nonlinear normalization mapping under a unified time base to construct a structured time series tensor; The operating condition classification and acquisition module extracts the dynamic response gradient factor, disturbance coupling factor, and abnormal activity factor based on the time series tensor, constructs the state feature vector of the time window, and models the state evolution trend based on the gated loop structure to identify the operating condition category of the coal mining machine in the future time period; The economic impact module establishes a correspondence between the operating condition category and the unit time output loss, energy consumption increase, and scheduling delay factor based on the operating condition category and historical equipment response experience, generates a set of economic impact factors, and binds them to the forecast time node; The risk identification module calculates the total economic loss value of each time window based on three types of economic factors, summarizes them into an economic loss trend sequence, identifies high-risk windows, and marks potential intervention areas based on the working condition category; The path intervention module builds a scheduling dependency graph based on the economic loss trend sequence and high-risk windows, determines the synchronous mutation probability nodes, screens out intervention points, generates intervention strategy paths, and evaluates the intervention strategy paths; The coal machinery adjustment module visualizes intervention strategy results, status trend charts, and economic losses, and deploys and executes strategy recommendations in the form of work orders.

[0054] The above formulas are all dimensionless and calculated numerically. Specific dimension removal can be achieved by various means such as standardization, which will not be elaborated here. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0055] The above embodiments can be implemented in whole or in part via software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in the embodiments of this application are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, ATA hard drives, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state ATA hard drive.

[0056] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0057] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0058] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0059] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0060] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0061] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. An underground coal mining machine operating condition monitoring method based on operating status, characterized in that: The steps include: Collect the operating status signals of underground coal mining machines, perform sampling alignment and nonlinear normalization mapping under a unified time base, and construct a structured time series tensor; Based on the time series tensor, the dynamic response gradient factor, disturbance coupling factor, and abnormal activity factor are extracted to construct the state feature vector of the time window. The state evolution trend is modeled based on the gated loop structure to identify the operating condition category of the coal mining machine in the future time period. Based on the operating condition category and historical equipment response experience, a correspondence is established between the operating condition category and unit time output loss, energy consumption increase, and scheduling delay factor. A set of economic impact factors is generated and bound to the forecast time node. Calculate the total economic loss value for each time window based on three types of economic factors, and summarize them to form an economic loss trend sequence, identify high-risk windows, and mark potential intervention areas based on the working condition category; Based on the economic loss trend sequence and high-risk windows, a scheduling dependency graph is constructed to determine the synchronous mutation probability nodes, screen out intervention points, generate intervention strategy paths, and evaluate the intervention strategy paths; Visualize intervention strategy results, status trend charts, and economic losses, and deploy and execute strategic recommendations as work orders.

2. The method for monitoring the working condition of an underground coal mining machine based on the operating status according to claim 1, characterized in that: Collect the operating status signals of underground coal mining machinery, perform sampling alignment and nonlinear normalization mapping under a unified time base, and construct a structured time series tensor. The specific process is as follows: The sensor group includes a three-phase current acquisition unit installed at the output end of the cutting motor, pressure transmitters and temperature sensors arranged in the front and rear chambers of the hydraulic propulsion cylinder, a three-axis vibration sensor installed on the drum structure, and a speed encoder integrated into the propulsion mechanism; The acquisition frequency of all sensor channels is uniformly set to a period synchronized with the clock signal of the coal machine control host, and time interpolation is used for completion and synchronization correction; All sensor channels are mapped and transformed using a nonlinear amplitude normalization strategy with interval compression; After completing the amplitude normalization process, a state tensor sequence is constructed using a sliding window of fixed time length. All sensor channels are arranged in chronological order to form a state matrix. The rows of the matrix represent the channel numbers, and the columns of the matrix represent the time series positions. The matrix is ​​the time series tensor corresponding to the sliding window.

3. The method for monitoring the working condition of an underground coal mining machine based on the operating status according to claim 2, characterized in that: The dynamic response gradient factor, disturbance coupling factor, and abnormal activity factor are extracted based on the time series tensor. The specific process is as follows: Scan the continuous signal data of the selected channel, calculate the numerical changes between all two adjacent data points in the entire time window, record the absolute change amplitude between each pair of adjacent data points as an instantaneous gradient value, and select the maximum value of all instantaneous gradient values ​​in the time window as the dynamic response gradient factor of the channel; Compare whether the two channels show the same change trend at each sampling moment, count the number of data point pairs with the same change trend within the entire window period, and use the proportion as the disturbance coupling factor of the channel pair; Within the time window, the numerical change between two adjacent sampling points is calculated one by one to determine whether it exceeds the disturbance threshold set for the channel. All data points that meet the condition that the change exceeds the disturbance threshold are regarded as a high-frequency disturbance event. The total number of high-frequency disturbance events in the time window is counted and compared with the total number of sampling points in the time window. The ratio is used as the abnormal activity factor; The dynamic response gradient factor is used to measure the maximum change intensity of a single channel within a time window, indicating the degree of sudden change in the operating state of the component corresponding to the channel; The perturbation coupling factor is used to measure the degree of synchronization change between different channels; The abnormal activity factor is used to measure the proportion of each channel in the high-frequency disturbance area within the time window.

4. The method for monitoring the working condition of an underground coal mining machine based on the operating status according to claim 3 is characterized in that: Construct the state feature vector of the time window, and model the state evolution trend based on the gated loop structure to identify the operating condition category of the coal mining machine in the future time period. The specific process is as follows: In each time window, the dynamic response gradient factor, disturbance coupling factor and abnormal activity factor characteristics are calculated in each channel or channel pair and then uniformly spliced ​​to form the state feature vector of the time window; After the state feature vectors of all time windows are constructed, they are combined in chronological order to form a continuous feature sequence, which is used as the dynamic behavior trajectory input during the operation of the coal machine. The operating condition classification model is used for operating condition classification prediction processing. The operating condition classification model uses a sequence modeling algorithm based on a gated loop structure. The model input is a sequence of state feature vectors in a continuous time window, and the output is the predicted result of the coal machine's operating condition category in the future time period. Working condition categories include normal operation, light load operation, heavy load operation, high frequency vibration, hydraulic fluctuation and shutdown warning categories.

5. The method for monitoring the working condition of an underground coal mining machine based on the operating status according to claim 4 is characterized in that: Based on the operating condition category and historical equipment response experience, a correspondence is established between the operating condition category and the unit time output loss, energy consumption increase, and scheduling delay factor. A set of economic impact factors is generated and bound to the forecast time node. The specific process is as follows: Equipment historical response experience includes equipment historical data analysis and expert knowledge annotation. The three types of economic factors include unit time output loss factor, unit time energy consumption increase factor, and unit time scheduling delay factor. Under normal operating conditions, the standard output value of the coal machine per unit time is recorded as the benchmark capacity. Historical data of the target state is extracted, and all window time periods under the corresponding operating condition category are screened out. The actual average output of the coal machine in all window time periods is calculated. The average output under the target state is subtracted from the benchmark capacity to obtain the output loss value. The output loss value is then converted into the ratio of the benchmark capacity to obtain the unit time output loss factor. The total energy consumption per unit time of the coal machine under normal conditions is calculated and used as a benchmark value. The time window data corresponding to the historical state is extracted for the target state. The average values ​​of the current, pressure, and temperature channels per unit time are calculated and converted into total energy consumption. The average energy consumption per unit time under the target state is obtained. The difference between the average energy consumption under the target state and the total energy consumption under normal conditions is compared, and the difference is calculated by comparing with the benchmark value. The ratio is used as the energy consumption increase factor. During the normal operation cycle of the coal machine, the average advance distance per unit time, the number of cutting tasks or the production rhythm are counted and used as a reference for the standard scheduling time series. For the target operating state, historical operating condition records are extracted, and the actual task pause time, delay time or task interruption time caused by the target operating state are counted. The average cumulative delay time per unit time is calculated and used as the unit time scheduling delay factor for the target operating state. The predicted operating condition categories are bound together, and within each time window, the operating condition category corresponding to the time window and the three types of economic factors obtained by mapping are recorded to generate a sequence of economic impact factor sets.

6. The method for monitoring the working condition of an underground coal mining machine based on the operating status according to claim 5, characterized in that: The total economic loss value of each time window is calculated based on the three types of economic factors, and summarized to form an economic loss trend sequence. High-risk windows are identified, and potential intervention areas are marked based on the working condition category. The specific process is as follows: For each time window, based on the three types of economic factors bound to it, losses are calculated from three dimensions: capacity loss, electricity cost, and dispatch cost, and the sum is taken as the total economic loss value of the time window; Taking the window number as the time axis, summarize the economic loss values ​​of all time windows in order to build a complete time series structure as the economic loss trend series; Arrange the economic loss values ​​of each time window in the order of collection to form a data sequence with time as the horizontal axis and economic loss as the vertical axis; A risk threshold is set based on the loss trend sequence, and the high quantile of the economic loss trend sequence is selected as the high-risk judgment standard. The entire economic loss trend sequence is traversed, and the window numbers of all economic loss values ​​higher than the risk threshold are extracted. They are determined as high-risk windows and marked as potential intervention areas.

7. The method for monitoring the working condition of an underground coal mining machine based on the operating status according to claim 6, characterized in that: For each time window, based on the three types of economic factors bound to it, losses are calculated from the three dimensions of capacity loss, power cost, and dispatch cost, and the sum is taken as the total economic loss value of the time window. The specific process is as follows: Calculate the theoretical output value loss caused by the decline in production capacity during the time window based on the unit time output loss factor and the normal coal mining quota; Calculate the additional energy costs incurred within the time window based on the energy consumption increase factor per unit time and the equipment's electricity price per kilowatt-hour; Calculate the labor and equipment idle costs caused by scheduling delays within the time window based on the unit time scheduling delay factor and the scheduling labor cost benchmark; Add the three results together to form the total economic loss value within the time window.

8. The method for monitoring the working condition of an underground coal mining machine based on the operating status according to claim 7, characterized in that: Based on the economic loss trend sequence and high-risk windows, a scheduling dependency graph is constructed to determine the synchronous mutation probability nodes, screen out intervention points, and generate intervention strategy paths. The specific process is as follows: Define a set of nodes in the scheduling dependency graph. Each node corresponds to an independent device unit. The edge set in the scheduling dependency graph is used to indicate the existence of a scheduling dependency or operation linkage relationship between two nodes. The direction of the edge indicates the impact flow. The weight of the edge is set to the degree of state correlation, which is calculated based on the probability of synchronous mutation of two channels in the high-risk window in the historical data; For each window marked as high-risk, the corresponding state feature vector and operating condition category sequence are analyzed to determine the operating condition type that causes economic loss. Using the rising economic loss factor as the signal source, the graph is expanded outward to track the associated scheduling dependency graph nodes. Identify the node groups with continuous abnormal responses and determine the intervention points based on the status propagation priority score and status impact index; After identifying the intervention points, an intervention strategy path is generated based on the operation control rules, equipment adjustment parameters and historical intervention effect records.

9. The method for monitoring the working condition of an underground coal mining machine based on the operating status according to claim 8, characterized in that: The intervention strategy path is evaluated. The specific process is as follows: The implementation cost is the composite cost of labor costs, material costs and equipment downtime losses required to complete all actions in the intervention strategy path; Response latency is the average time interval from the issuance of a strategy to the actual effect feedback; The historical success rate is the ratio of the intervention strategy path that successfully reduces economic losses to below the set threshold under historical operating conditions.

10. An underground coal mining machine operating condition monitoring system based on operating status, used to implement the underground coal mining machine operating condition monitoring method based on operating status according to any one of claims 1 to 9, characterized in that: include: The coal mining machine data acquisition module collects the operating status signals of underground coal mining machines and performs sampling alignment and nonlinear normalization mapping under a unified time base to construct a structured time series tensor. The operating condition classification and acquisition module extracts the dynamic response gradient factor, disturbance coupling factor, and abnormal activity factor based on the time series tensor, constructs the state feature vector of the time window, and models the state evolution trend based on the gated loop structure to identify the operating condition category of the coal mining machine in the future time period; The economic impact module establishes a correspondence between the operating condition category and the unit time output loss, energy consumption increase, and scheduling delay factor based on the operating condition category and historical equipment response experience, generates a set of economic impact factors, and binds them to the forecast time node; The risk identification module calculates the total economic loss value of each time window based on three types of economic factors, summarizes them into an economic loss trend sequence, identifies high-risk windows, and marks potential intervention areas based on the working condition category; The path intervention module builds a scheduling dependency graph based on the economic loss trend sequence and high-risk windows, determines the synchronous mutation probability nodes, screens out intervention points, generates intervention strategy paths, and evaluates the intervention strategy paths; The coal machinery adjustment module visualizes intervention strategy results, status trend charts, and economic losses, and deploys and executes strategy recommendations in the form of work orders.

Images