A warning disposal method and system for hydraulic system pollution invasion

Abstract

The application discloses a kind of early warning disposal methods and systems for hydraulic system pollution invasion, belong to safety control technical field, include: through multiple sensors acquisition oil particle size, dielectric constant, gas content, moisture content and pressure pulsation signal, extract time series mutation characteristic factor, accurately identify four pollution types of water mist, dust particle, fuel mixing, bubble involvement by correlation identification model;Hydraulic system digital twin mirror image is constructed, the damage priority deviation coefficient is calculated by predicting pollution diffusion path;According to this, determine low, medium and high three levels of invasion risk;For medium and high risk, according to the pollution particle size level, trace the pollution source, match the failure of respirator, invasion path such as sealing wear;According to source, carry out oil line partition isolation verification, generate time stamp disposal data package and upload industrial block chain to store evidence;The application realizes accurate early warning, tracing and credible disposal of hydraulic system pollution invasion, improves system operation safety and reliability.

Description

Technical Field

[0001] This invention belongs to the field of safety management technology, specifically a method and system for early warning and handling of contamination intrusion into hydraulic systems. Background Technology

[0002] Hydraulic systems, as the core power unit of industrial equipment, are widely used in engineering machinery, aerospace, shipbuilding, and other fields. Their operational reliability directly determines the safety and efficiency of the entire machine. Oil contamination is a major cause of hydraulic component wear, valve jamming, seal failure, and system malfunctions. External contamination, such as water mist, dust, fuel mixing, and air bubble entrainment, is characterized by its suddenness, insidiousness, and rapid diffusion, easily triggering a chain reaction of failures. Existing technologies mostly rely on offline oil detection or single-parameter monitoring, which suffers from drawbacks such as strong lag, inaccurate identification of contamination types, and inability to predict diffusion paths, making early warning of contamination difficult. Furthermore, traditional disposal methods lack precise source tracing means, isolation strategies rely on manual experience, easily leading to inappropriate isolation scope or delays in disposal, and the disposal process lacks reliable evidence, making accountability difficult. In summary, existing hydraulic system contamination prevention and control technologies have significant shortcomings in real-time monitoring, accurate identification, risk prediction, source tracing, and reliable disposal. There is an urgent need for an integrated technical solution that can achieve accurate early warning of contamination intrusion, intelligent source tracing, efficient disposal, and blockchain evidence storage to ensure the safe and stable operation of hydraulic systems. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention proposes a method and system for early warning and handling of contamination intrusion in hydraulic systems. It collects signals from multiple sensors, including oil particle size, dielectric constant, gas content, moisture content, and pressure pulsation, extracts temporal abrupt change characteristic factors, and accurately identifies four types of contamination—water mist, dust particles, fuel mixing, and air bubble entrainment—using an association recognition model. A digital twin image of the hydraulic system is constructed to predict contamination diffusion paths and calculate damage priority deviation coefficients, thereby determining low, medium, and high intrusion risks. For medium and high risks, the source of contamination is traced according to particle size, matching intrusion paths such as breather failure and seal wear. Based on the source, oil circuit zoning and isolation verification is performed, generating a timestamped handling data package and uploading it to an industrial blockchain for evidence storage. This invention achieves accurate early warning, source tracing, and reliable handling of contamination intrusion in hydraulic systems, improving system operational safety and reliability.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for early warning and handling of contamination intrusion in hydraulic systems, comprising:

[0006] Collect time-series signals of oil particle size, oil dielectric constant, air content, moisture content and pressure pulsation of the hydraulic system, extract time-series abrupt change characteristic factors of external pollution intrusion, input the time-series abrupt change characteristic factors into a pre-constructed association identification model, and output the pollution type;

[0007] Construct a digital twin image of the hydraulic system, predict the pollution diffusion path sequence by combining the pollution type, and calculate the damage priority deviation coefficient between the actual pollutant diffusion state and the pollution diffusion path sequence.

[0008] By comparing the damage priority deviation coefficient with the preset grading threshold, the risk level of external pollution intrusion into the hydraulic system is determined. For hydraulic systems with external pollution intrusion, the corresponding intrusion path is matched according to the particle size level of the pollutants to trace the source of pollution.

[0009] Based on the pollution source, the oil pipeline is isolated and verified, a pollution disposal instruction summary is generated, and the verification timestamp is binary encoded and embedded into the disposal instruction summary to generate an early warning disposal data package and upload it to the industrial blockchain for evidence storage.

[0010] Specifically, the extraction of temporal mutation characteristic factors of exogenous pollution invasion includes:

[0011] Based on the cubic spline interpolation method, the timestamps of the oil particle size time series signal, oil dielectric constant, air content, moisture content and pressure pulsation are unified to the same time reference sequence to obtain a time-aligned multidimensional physical quantity sequence.

[0012] A sliding window differential operation is performed on the time-aligned oil particle size time series signal to obtain the particle concentration change rate sequence. The 3σ criterion is used to detect outliers in the particle concentration change rate sequence, and the time position corresponding to the particle concentration change rate exceeding three standard deviations is marked as the particle size mutation start point.

[0013] Based on the particle size mutation start point, the oil dielectric constant segment, air content segment, moisture content segment, and pressure pulsation segment within a preset time window are extracted from the time-aligned multidimensional physical quantity sequence, and first-order difference operations are performed on each segment to obtain the corresponding dielectric constant change rate sequence, air content change rate sequence, moisture content change rate sequence, and pressure pulsation change rate sequence. These sequences are then combined with the particle concentration change rate sequence to form a multidimensional mutation feature vector, which is used as the temporal mutation feature factor.

[0014] Specifically, the step of inputting time-series mutation feature factors into a pre-constructed association identification model and outputting contamination types includes:

[0015] The multidimensional mutation feature vector is subjected to dimensionality reduction processing based on principal component analysis. Principal components with a cumulative variance contribution rate greater than a preset contribution rate threshold are selected to obtain the dimensionality-reduced time-series mutation feature factors.

[0016] Load a pre-built association recognition model; the association recognition model is a support vector machine classifier based on radial basis function kernel function;

[0017] The dimensionality-reduced temporal mutation feature factors are used as input vectors and arranged in chronological order to form a feature input sequence. Each trained support vector is used as a reference vector, and the radial basis function kernel is used to calculate the kernel function value between the input vector and each reference vector. The support vectors are obtained by training on historical pollution intrusion samples using a sequence minimum optimization algorithm. A one-to-many classification strategy is used to calculate the weighted sum values ​​corresponding to the four types of pollution. The weighted sum values ​​of the four types are compared, and the pollution type identifier corresponding to the maximum value is output. The pollution type identifier is one of the following: water mist intrusion, dust particle intrusion, fuel oil mixing, or bubble entrainment.

[0018] Specifically, the construction of the digital twin mirror of the hydraulic system includes:

[0019] Collect the topological parameters of the physical entities of the hydraulic system and construct a geometric model of the hydraulic system;

[0020] A sub-model of oil flow mechanism was constructed based on the Navier-Stokes equations, and a sub-model of pollutant transport mechanism was constructed based on the principles of mass conservation and momentum conservation.

[0021] The oil flow mechanism sub-model and the pollutant transport mechanism sub-model are coupled through a data interface to obtain the output value of the joint mechanism model;

[0022] The measured values ​​of various flow field physical quantities corresponding to key measuring points of the hydraulic system physical entity are collected. Simultaneously, the simulated physical quantities output by the joint mechanism model at the corresponding measuring points and at the same time are obtained. For the same measuring point and the same physical quantity, the difference between the simulated physical quantity and the measured value is calculated to obtain the deviation of a single physical quantity at a single measuring point. After performing min-max normalization on the deviations of a single physical quantity at all measuring points and all physical quantities, the deviations are spliced ​​according to the spatial order and temporal order of the measuring points to form a deviation sequence.

[0023] A deviation correction sub-model is constructed based on a long short-term memory network. The deviation sequence is used as input to train the deviation correction sub-model to output the deviation prediction value in a preset time domain.

[0024] The predicted deviation value is added to the output value of the joint mechanism model to obtain the twin output value;

[0025] The hydraulic system geometric model, the oil flow mechanism sub-model, the contaminant transport mechanism sub-model, and the deviation correction sub-model are coupled together through a data interface, and the twin output value is used as the output end of the digital twin mirror to form a digital twin mirror of the hydraulic system.

[0026] Specifically, the prediction of pollution diffusion path sequences based on pollution type includes:

[0027] The corresponding physical property parameters are retrieved from a preset pollutant attribute database based on the pollution type identifier; the physical property parameters include average particle size, particle density, particle shape coefficient, oil miscibility index, and bubble surface tension coefficient.

[0028] The physical property parameters are assigned to the pollutant transport mechanism sub-model, and the oil velocity distribution and pressure distribution output by the oil flow mechanism sub-model at the current moment are used as initial boundary conditions to input the pollutant transport mechanism model.

[0029] After receiving the physical property parameters and the initial boundary conditions, the pollutant transport mechanism sub-model calculates the mass concentration of pollutants at each node of the hydraulic system over time based on the convection-diffusion equation, thus obtaining the concentration time series of each node.

[0030] For each node, the time when the mass concentration first reaches the preset detection threshold in the corresponding concentration time series is marked as the pollution arrival time of that node. The corresponding nodes are sorted in order from early to late according to the pollution arrival time, and the path of pollutant propagation between adjacent nodes is used as the edge to generate a tree-structured pollution diffusion path sequence.

[0031] Specifically, the calculation of the damage priority deviation coefficient between the actual pollutant diffusion state and the pollution diffusion path sequence includes:

[0032] The actual pollutant concentration, actual pollutant particle size distribution, and actual pollutant accumulation rate of each hydraulic component are extracted in real time from the twin output value output by the digital twin mirror of the hydraulic system and combined into an actual pollutant diffusion state vector.

[0033] The predicted pollutant concentration, predicted pollutant particle size distribution, and predicted arrival time carried by each node are extracted from the pollution diffusion path sequence and combined into a predicted diffusion state vector.

[0034] The state deviation between the actual pollutant diffusion state vector and the predicted diffusion state vector is calculated based on Mahalanobis distance; each component of the state deviation corresponds to the deviation value of a hydraulic component.

[0035] Look up the damage sensitivity coefficient of each hydraulic component from the preset component damage sensitivity coefficient table;

[0036] Each component of the state deviation is multiplied by the damage sensitivity coefficient of the corresponding hydraulic component to obtain a weighted deviation component. All weighted deviation components are then accumulated in the order of nodes in the pollution diffusion path sequence and normalized by min-max to output the damage priority deviation coefficient.

[0037] Specifically, by comparing the damage priority deviation coefficient with a preset grading threshold, the risk level of external contamination intrusion into the hydraulic system is determined, including:

[0038] Obtain a preset first grading threshold and a second grading threshold, and simultaneously obtain a damage priority deviation coefficient; the first grading threshold is less than the second grading threshold;

[0039] When the damage priority deviation coefficient is less than the first classification threshold, it is determined to be a low-risk level, and a corresponding low-risk level code is generated.

[0040] When the damage priority deviation coefficient is greater than or equal to the first grading threshold and less than the second grading threshold, it is determined to be of medium risk level, and a corresponding medium risk level code is generated.

[0041] When the damage priority deviation coefficient is greater than or equal to the second classification threshold, it is determined to be a high-risk level, and a corresponding high-risk level code is generated.

[0042] The generated low-risk, medium-risk, and high-risk level codes are output as the results of the determination of the risk level of external pollution intrusion.

[0043] Specifically, for hydraulic systems susceptible to external contamination, matching corresponding intrusion paths according to contaminant particle size levels to trace the source of contamination includes:

[0044] Obtain the risk level of external pollution intrusion. When the risk level of external pollution intrusion is medium risk or high risk, execute:

[0045] The time-series signal of oil particle size is acquired, and the time-series signal of oil particle size is analyzed in layers according to the preset particle size interval boundary to obtain sub-concentration sequences corresponding to multiple particle size levels.

[0046] Mutation point detection is performed on the sub-concentration sequences corresponding to each particle size level. The cumulative sum and statistics of each sub-concentration sequence are calculated. The time point when the cumulative sum and statistics exceed the preset mutation judgment threshold is marked as the mutation start time of the corresponding particle size level. The mutation start times of all particle size levels are arranged in ascending order of particle size to construct a particle size-time mutation matrix.

[0047] The particle size-time mutation matrix is ​​matched with each invasion path feature template in the preset invasion path feature library by cosine similarity, and the invasion path corresponding to the invasion path feature template with the largest cosine similarity is selected as the pollution source output; the invasion path feature library stores the standard matrices corresponding to the respirator filter failure path, piston rod seal wear path, pipeline joint loosening path, and new oil filling pollution path, respectively.

[0048] Specifically, the step of performing oil pipeline zoning isolation verification based on the pollution source and generating a pollution disposal instruction summary includes:

[0049] Based on the pollution source, the corresponding oil circuit partition boundary to be isolated is queried from the preset isolation strategy library. The oil circuit partition boundary to be isolated includes the isolation valve tag number, the isolation pipeline section identifier, and the list of associated components.

[0050] In the digital twin mirror of the hydraulic system, the opening parameter of the valve element corresponding to the isolation valve position number is set to zero, the valve closing operation is simulated, and based on the simulated closing conditions, the oil flow mechanism sub-model and the contaminant transport mechanism sub-model are rerun to calculate the predicted decay curve of the oil contaminant concentration in the non-isolation zone.

[0051] Obtain a safety threshold; when all concentration values ​​of the predicted decay curve within a preset time window are lower than the safety threshold, generate an oil circuit partition isolation verification pass flag.

[0052] The pollution source, the boundary of the oil pipeline to be isolated, and the oil pipeline isolation verification mark are combined according to the preset instruction template to generate a pollution disposal instruction summary.

[0053] An early warning and response system for contamination intrusion into hydraulic systems, comprising:

[0054] The feature recognition module is used to collect time-series signals of oil particle size, oil dielectric constant, air content, moisture content and pressure pulsation, extract time-series abrupt change feature factors of external pollution intrusion, and input them into a radial basis kernel function support vector machine classifier after dimensionality reduction by principal component analysis, and output pollution type identifier.

[0055] The digital twin modeling module is used to construct a digital twin image of the hydraulic system, query the physical property parameters of pollutants by combining the pollution type, predict the pollution diffusion path sequence, calculate the damage priority deviation coefficient between the actual and predicted diffusion states, and compare the classification threshold to determine the risk level of external pollution intrusion.

[0056] The pollution source tracing module is used to analyze the time-series signal of oil particle size according to particle size intervals under medium and high risk levels, construct a particle size-time mutation matrix, and perform cosine similarity matching with the intrusion path feature library to trace the pollution source.

[0057] The isolation and disposal module is used to query the isolation strategy library based on the pollution source, simulate the oil circuit zoning isolation operation in the digital twin mirror and verify the isolation effectiveness, generate a pollution disposal instruction summary, embed the verification timestamp binary code into the pollution disposal instruction summary and verify it, generate an early warning disposal data package and upload it to the industrial blockchain to complete transaction verification and evidence storage.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] 1. This invention proposes an early warning and handling method for contamination intrusion in hydraulic systems. By using multi-sensor fusion monitoring and extraction of temporal mutation features, combined with a radial basis function kernel function support vector machine classifier, it can accurately identify four types of exogenous pollution, overcoming the shortcomings of traditional single-parameter monitoring in terms of lag and inaccurate identification. Relying on digital twin mirrors, it realizes pollution diffusion path prediction and risk classification, improving the accuracy and timeliness of early warning of pollution intrusion, and effectively reducing the risk of component wear and system failure caused by sudden pollution.

[0060] 2. This invention proposes an early warning and handling method for contamination intrusion in hydraulic systems. It traces the source of contamination according to the particle size hierarchy, achieves precise matching of intrusion paths, and performs oil circuit zoning isolation verification based on the source to ensure the scientific reliability of the handling strategy. At the same time, the handling instructions are embedded with timestamps and uploaded to the industrial blockchain for evidence storage, ensuring that the handling process is traceable and tamper-proof. This achieves integrated management and control of hydraulic system contamination early warning, source tracing, handling and evidence storage, and comprehensively improves the system's operational safety, reliability and management standardization. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of an early warning and handling method for contamination intrusion in hydraulic systems according to the present invention;

[0062] Figure 2 This is a flowchart illustrating the principle of an early warning and handling method for contamination intrusion in hydraulic systems according to the present invention.

[0063] Figure 3 This is a diagram of an early warning and response system architecture for contamination intrusion into hydraulic systems according to the present invention. Detailed Implementation

[0064] Example 1

[0065] Please see Figures 1-2 The present invention provides an embodiment of a method for early warning and handling of contamination intrusion in hydraulic systems, the method comprising S1 to S4, including the following steps:

[0066] S1: Collect the time-series signal of oil particle size, oil dielectric constant, air content, moisture content and pressure pulsation of the hydraulic system, extract the time-series abrupt change characteristic factors of external pollution intrusion, input the time-series abrupt change characteristic factors into the pre-constructed association identification model, and output the pollution type;

[0067] During long-term operation of industrial hydraulic systems, external impurities, aging and wear of seals, impurities introduced during new oil filling, and failure of the breather filter element can all lead to oil contamination. Contaminants such as particles, moisture, and air can accelerate the wear of hydraulic components, clog precision valve ports, and impair the lubricating properties of the oil, potentially causing system failures or even shutdowns. This embodiment uses a medium-to-high pressure hydraulic system commonly used in construction machinery as a specific application scenario. This system includes core components such as hydraulic pumps, multi-way valves, hydraulic cylinders, hydraulic motors, oil tanks, filters, various pipelines, and seals. Its operating pressure range is 16-31.5 MPa, and its operating oil temperature is 30-60℃. During daily operation, it is susceptible to external contamination from water mist, dust particles, fuel mixing, and air bubbles. This method can accurately identify the type of contamination and provide early warning and response.

[0068] Furthermore, the acquisition of the hydraulic system's oil particle size timing signal, oil dielectric constant, air content, moisture content, and pressure pulsation includes:

[0069] (1) The oil particle size time sequence signal is obtained based on the first online particle size monitoring sensor installed in the return oil pipeline of the hydraulic system, and the oil particle size time sequence signal is obtained based on the second online particle size monitoring sensor installed in the main oil circuit. The two signals are aligned by timestamp and the arithmetic mean is taken to obtain the fused oil particle size time sequence signal.

[0070] In the hydraulic system of construction machinery, the oil flowing through the return oil line carries particles generated by wear within the system and external contaminants, making it an area with a high concentration of contaminant particles. The main oil line directly delivers high-pressure oil to the actuators. The cleanliness of the oil directly affects the working accuracy of the actuators. Therefore, deploying particle size sensors at these two key locations can comprehensively capture the oil particle contamination situation. In this embodiment, the first online particle size monitoring sensor is an industrial-grade laser scattering online particle counter, installed on the straight return oil pipe section at the front end of the hydraulic oil tank, 50cm away from the return oil port. At this location, the oil flow is stable and there is no eddy current interference. The sensor probe is vertically inserted into the center of the pipe, which can collect the quantity and concentration data of particles of different sizes in the oil in real time, with an output frequency of 1 time / second, generating a time-series signal containing particle quantity, particle size distribution, and concentration value. The second online particle size monitoring sensor is the same model of laser scattering online particle counter as the first sensor, installed on the straight main oil pipe section between the hydraulic pump outlet and the multi-way valve inlet, 30cm away from the hydraulic pump outlet. At this location, the oil is in a high-pressure steady-state flow state. The sensor probe is also vertically inserted into the center of the pipe, collecting particle contamination data in the high-pressure oil in real time, with an output frequency consistent with the first sensor, generating a corresponding oil particle size time-series signal.

[0071] Both sensors collect oil particle size time-series signals containing timestamp information, generated by the sensor's built-in clock with millisecond-level accuracy. To eliminate signal errors caused by local eddies, installation deviations, and instantaneous flow fluctuations in a single sensor, the two signals need to be fused: First, using the system's unified clock as a reference, the timestamps of the two time-series signals are calibrated and aligned, and abnormal signals with time deviations exceeding 20 milliseconds are removed; then, the arithmetic average of the particle concentration values ​​of the two signals at the same time point after alignment is calculated, and particle size distribution data is simultaneously integrated to finally generate a fused oil particle size time-series signal, which can more realistically and stably reflect the real-time status of oil particle contamination in the entire hydraulic system.

[0072] (2) The dielectric constant of the oil is obtained based on a coaxial capacitive dielectric constant sensor installed in the return oil zone in the middle of the hydraulic oil tank;

[0073] In this embodiment, the hydraulic oil tank is a core component for storing, cooling, degassing, and settling impurities in the hydraulic system. The central return oil zone is where the return oil mixes thoroughly with the oil already in the tank. The oil in this zone is uniform, representing the overall dielectric properties of the system oil. Furthermore, it is far from the deposited impurities at the bottom of the tank and the air interface at the top, preventing interference from impurity deposition and oil evaporation on the dielectric constant detection results. In this embodiment, a high-precision coaxial capacitive dielectric constant sensor is selected. This sensor employs a coaxial cylindrical structure design, offering strong anti-electromagnetic interference capabilities, high detection accuracy, and fast response speed. It is vertically installed in the central return oil zone of the hydraulic oil tank, with the sensor probe completely submerged in the oil, positioned 15cm from the bottom of the tank and 10cm below the oil surface. This ensures sufficient oil flow around the probe and prevents the accumulation of impurities.

[0074] Furthermore, the dielectric constant of hydraulic oil is a key physical quantity reflecting the oil's insulation performance and composition purity. The dielectric constant of pure hydraulic oil is within a stable range. When contaminants such as water, fuel, and dust particles are mixed into the oil, the dielectric constant will change significantly: an increase in water content will significantly increase the dielectric constant, the presence of fuel will slightly decrease the dielectric constant, and the presence of dust particles will cause slight fluctuations in the dielectric constant. This coaxial capacitive dielectric constant sensor detects changes in the oil capacitance value and converts it into a dielectric constant value in real time, with an output frequency of 1 time / 2 seconds, enabling continuous and stable acquisition of oil dielectric constant data.

[0075] (3) The air content is obtained based on a differential pressure air content sensor connected in series between the high pressure pipeline between the hydraulic pump outlet and the filter inlet;

[0076] Furthermore, the high-pressure pipeline between the hydraulic pump outlet and the filter inlet is a crucial channel for the oil to flow from the power source to the purification device. Here, the oil is under high pressure and high-speed flow. If the system has problems such as poor sealing, low oil level, or severe return oil impact, outside air can easily mix into the oil, forming bubbles. These bubbles, carried by the oil flow, will compress under high pressure and expand under low pressure, causing cavitation, vibration, and noise, reducing system efficiency. Therefore, deploying an air content sensor at this location can accurately detect the air mixing in the oil. In this embodiment, a high-pressure differential pressure air content sensor is selected, capable of withstanding a maximum working pressure of 40MPa, suitable for high-pressure conditions in hydraulic systems. It is installed in series on the high-pressure straight pipe section between the hydraulic pump outlet and the high-pressure filter inlet, with both ends sealed to the pipeline via high-pressure flanges to ensure no leakage. The sensor integrates a high-precision differential pressure detection element, which can detect the pressure difference caused by the presence of bubbles as the oil flows through in real time.

[0077] Furthermore, the air content in the oil directly affects its compressibility and flow characteristics. Pure oil has an extremely low air content. When air bubbles are entrained and contamination occurs, the air content will rise rapidly. This differential pressure air content sensor detects the pressure difference before and after the oil flows through the sensor in real time, and converts it into the air volume ratio in the oil, i.e., the air content, using a preset algorithm. The output frequency is 1 time / second, which can quickly respond to air bubble entrainment contamination.

[0078] (4) The moisture content is obtained based on the resistive-capacitive moisture sensor installed in the low-temperature area at the bottom of the hydraulic oil tank;

[0079] Furthermore, moisture is one of the most harmful contaminants in hydraulic system fluids. Moisture contamination can lead to emulsification, decreased lubrication, corrosion of metal components, and short circuits in electrical components, severely impacting system reliability. The low-temperature region at the bottom of the hydraulic tank has a relatively low temperature, and moisture, being denser than oil, tends to accumulate there. This area offers the highest sensitivity for moisture detection and is far from the oil surface, avoiding interference from oil evaporation and condensation at the top. In this embodiment, a high-sensitivity resistive-capacitive moisture sensor is selected. It detects moisture by utilizing the impedance and capacitance changes caused by variations in oil moisture content. The detection range is 0-1%, with an accuracy of 0.01%. The sensor is horizontally installed in the low-temperature region at the bottom of the hydraulic tank, with the probe facing upwards, 5cm from the bottom of the tank. This prevents deposited impurities from covering the probe, ensuring accurate detection results.

[0080] Furthermore, the water content of pure hydraulic oil is usually below 0.03%. When water mist intrusion, water seepage from seals, or condensation of oil occurs, the water content will rise rapidly. This resistive-capacitive water sensor collects oil water content data in real time, with an output frequency of 1 time / 2 seconds, continuously monitoring the water contamination of the oil and providing core data for the identification of water mist intrusion-type contamination.

[0081] (5) Pressure pulsation is obtained based on a high-frequency dynamic pressure sensor installed at the connection between the hydraulic pump outlet and the main pipeline of the system.

[0082] Furthermore, the connection between the hydraulic pump outlet and the main system pipeline is a critical node for the diversion of high-pressure oil to various actuators. The oil flow is complex and susceptible to pressure pulsations caused by contaminants, component wear, and flow fluctuations. The amplitude and frequency of these pressure pulsations directly reflect the state of contamination intrusion into the system: dust particles clogging valve ports and moisture causing component corrosion can increase the amplitude and disrupt the frequency of pressure pulsations, while air bubbles can cause periodic fluctuations. In this embodiment, a high-frequency dynamic pressure sensor is selected. This sensor has a response frequency of up to 10kHz, can capture high-frequency pressure fluctuations, and has a measurement range of 0-40MPa. It is suitable for high-pressure hydraulic systems and is vertically installed on the pipeline at the connection between the hydraulic pump outlet and the main system pipeline. The probe is flush with the inner wall of the pipeline to avoid interfering with the oil flow. The sensor is fixed to the pipeline through a sealing joint to ensure no leakage under high-pressure conditions.

[0083] Furthermore, this high-frequency dynamic pressure sensor collects dynamic pressure data of the oil in the pipeline in real time, with an output frequency of 10 times / second, generating a pressure pulsation timing signal that includes pressure peaks, valleys, and fluctuation frequencies, which can accurately capture abnormal pressure fluctuations caused by contamination intrusion.

[0084] The extraction of temporal mutation characteristic factors of exogenous pollution invasion includes:

[0085] S1.1: Based on the cubic spline interpolation method, the timestamps of the oil particle size time series signal, oil dielectric constant, air content, moisture content and pressure pulsation are unified to the same time reference sequence to obtain a time-aligned multidimensional physical quantity sequence. The cubic spline interpolation method is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0086] S1.2: Perform sliding window differential operation on the time-aligned oil particle size time series signal, that is, divide the difference between the particle number concentration at the current time and the particle number concentration at the previous time by the sampling interval to obtain the particle concentration change rate sequence, and use the 3σ criterion to detect outliers in the particle concentration change rate sequence, and mark the time position corresponding to the particle concentration change rate exceeding three standard deviations as the particle size mutation start point.

[0087] Furthermore, the 3σ criterion was used to detect outliers in the particle concentration change rate sequence. Time points corresponding to particle concentration change rates exceeding three standard deviations were marked as particle size mutation initiation points, including:

[0088] (1) Obtain the particle concentration change rate sequence. The particle concentration change rate sequence is a one-dimensional time series data obtained by processing the oil particle size time series signal through sliding window differential operation. All data in the particle concentration change rate sequence are actual values ​​obtained by continuous sampling. The data order is strictly arranged according to the time sequence. Each data point corresponds to a sampling time in a unified time reference sequence. The sequence length is determined by the actual sampling time. The original sequence output by differential operation is directly used.

[0089] (2) Perform statistical calculations on the obtained particle concentration change rate sequence. First, calculate the average value of all data in the particle concentration change rate sequence. When calculating the average value, the values ​​of all data in the particle concentration change rate sequence need to be added together in sequence. After obtaining the sum, divide it by the total number of data points in the particle concentration change rate sequence. The result is the average value of the particle concentration change rate sequence.

[0090] (3) Continue to calculate the standard deviation of the particle concentration change rate sequence. When calculating the standard deviation, first calculate the difference between each data point in the particle concentration change rate sequence and the average value, then square each difference, then add all the squared differences to get the sum of squares, then divide the sum of squares by the total number of data points in the particle concentration change rate sequence to get the variance, and finally take the square root of the variance. The result is the standard deviation of the particle concentration change rate sequence.

[0091] (4) Determine the detection threshold corresponding to the three-standard-deviation criterion. The detection threshold is directly calculated from the standard deviation. The specific setting method is to multiply the calculated standard deviation value by 3, and the result is the outlier detection threshold.

[0092] (5) Traverse each data point in the particle concentration change rate sequence and compare the particle concentration change rate value corresponding to each data point with the predetermined detection threshold of three times the standard deviation.

[0093] (6) Select data points in the sequence whose values ​​exceed three times the standard deviation detection threshold. These selected data points are the outliers in the particle concentration change rate sequence. The selection process is based solely on the magnitude of the values.

[0094] (7) Locate the time position of each outlier in the particle concentration change rate sequence. This time position corresponds one-to-one with the sampling time in the unified time reference sequence. Each outlier corresponds to only one unique time position. The location process directly uses the original timestamp information of the sequence.

[0095] (8) Mark the earliest time position among all outlier values ​​as the particle size mutation start point. If multiple outlier values ​​appear in the particle concentration change rate sequence at the same time, only the earliest time position is selected as the mutation start point, and the time positions corresponding to the other outlier values ​​are no longer used as mutation start points. After completing this marking operation, the initial time of the external pollution invasion can be determined.

[0096] S1.3: Based on the particle size mutation start point, extract the oil dielectric constant segment, air content segment, moisture content segment, and pressure pulsation segment within a preset time window from the time-aligned multidimensional physical quantity sequence, and perform first-order difference operations on each to obtain the corresponding dielectric constant change rate sequence, air content change rate sequence, moisture content change rate sequence, and pressure pulsation change rate sequence. Then, combine these with the particle concentration change rate sequence to form a multidimensional mutation feature vector, and use the multidimensional mutation feature vector as the time-series mutation feature factor.

[0097] Furthermore, the realization process of the dielectric constant change rate sequence, gas content change rate sequence, moisture content change rate sequence, and pressure pulsation change rate sequence includes:

[0098] (1) Determine the time position of the starting point of particle size mutation. This starting point is the only time marker point previously detected by sliding window difference operation and three times standard deviation criterion. It represents the earliest moment of the invasion of exogenous pollution. This time point corresponds to a precise time number in the unified time reference sequence. All truncation operations are positioned with this number as the reference center without any offset or correction.

[0099] (2) Set the preset time window parameters for extracting physical quantity segments. The time window adopts a symmetrical extraction method, that is, extract a fixed number of historical sampling points before the start point of granularity change and extract a fixed number of sampling points after the start point. The number of sampling points before and after the start point is consistent to ensure that the center of the window falls on the start point of the change. The window length is set to 40 consecutive sampling points, of which 20 sampling points are taken before the start point of the change and 20 sampling points are taken after the start point of the change. The time interval corresponding to each sampling point is uniformly 100 milliseconds. The actual coverage time of the entire time window is 4 seconds. This parameter is a fixed setting and is not adjusted with changes in working conditions.

[0100] (3) Based on the determined particle size mutation start point and the set time window parameters, locate and extract the time series segments of four independent physical quantities in the multidimensional physical quantity sequence that has been time aligned. The specific extraction objects are the oil dielectric constant time series segment, the air content time series segment, the moisture content time series segment and the pressure pulsation time series segment. Each type of physical quantity segment is strictly extracted according to the same time window boundary to ensure that the four types of segments are completely aligned in the time dimension, and the start time, end time and number of sampling points are completely consistent to avoid time sequence misalignment or inconsistent length.

[0101] (4) Perform first-order difference operation on each type of physical quantity time series segment. The first-order difference operation method is to calculate the difference between adjacent sampling points. That is, for any physical quantity segment, starting from the second sampling point, subtract the physical quantity value corresponding to the previous sampling point from the physical quantity value corresponding to the current sampling point in turn. The difference obtained is the first-order difference result at the current position. No smoothing, filtering or normalization processing is performed on the original data during the operation. The original sampling values ​​are used directly for calculation to ensure that the difference result truly reflects the actual change of the physical quantity at adjacent time points.

[0102] (5) Following the same first-order difference operation rules, complete all difference calculations for the four types of physical quantity segments in sequence. For the oil dielectric constant segment, calculate the dielectric constant change rate sequence; for the air content segment, calculate the air content change rate sequence; for the moisture content segment, calculate the moisture change rate sequence; and for the pressure pulsation segment, calculate the pressure pulsation change rate sequence. The length of each change rate sequence is equal to the number of sampling points of the corresponding original physical quantity segment minus 1. The elements within the sequence are strictly arranged in chronological order to maintain temporal continuity and consistency.

[0103] The step of inputting time-series mutation feature factors into a pre-constructed association identification model and outputting contamination types includes:

[0104] S1.4: Perform dimensionality reduction processing based on principal component analysis on the multidimensional mutation feature vector, select principal components whose cumulative variance contribution rate is greater than a preset contribution rate threshold, and obtain the dimensionality-reduced time-series mutation feature factors.

[0105] Furthermore, the specific steps in S1.4 include:

[0106] (1) Obtain the multidimensional mutation feature vector to be processed. The multidimensional mutation feature vector is formed by combining the particle concentration change rate sequence, dielectric constant change rate sequence, gas content change rate sequence, moisture content change rate sequence and pressure pulsation change rate sequence. The vector contains feature data in 5 dimensions. Each dimension corresponds to the change rate information of a physical quantity. The data in each dimension are time series values ​​obtained by continuous sampling. The data order is strictly arranged according to the time sequence. The length of the vector is consistent with the number of sampling points in the time window.

[0107] (2) Standardize the multidimensional mutation feature vector. The standardization process adopts the zero mean standardization method. The purpose of the process is to eliminate the influence of the difference in the dimensions of the data. Specifically, for each dimension in the vector, first calculate the mean of all data in that dimension, then calculate the standard deviation of all data in that dimension, then subtract the mean of the corresponding dimension from each original data in that dimension, and then divide by the standard deviation of the corresponding dimension. The standardization calculation of the five dimensions is completed in sequence to obtain the standardized multidimensional mutation feature vector.

[0108] (3) Calculate the covariance matrix of the standardized multidimensional mutation feature vector. The covariance matrix is ​​used to reflect the correlation between features of each dimension. The number of rows and columns of the matrix is ​​consistent with the number of vector dimensions, that is, 5×5. The specific operation is to first calculate the covariance between any two dimensions. When calculating the covariance, first calculate the average of the product of the corresponding data of the two dimensions, and then subtract the product of the average values ​​of the two dimensions. Calculate the covariance between all dimensions pairwise in turn, and arrange them in the order of dimensions to form the covariance matrix. The diagonal elements of the matrix are the variance of each dimension itself, and the off-diagonal elements are the covariance between the corresponding two dimensions.

[0109] (4) Perform eigenvalue decomposition on the covariance matrix. Specifically, solve the characteristic equation of the covariance matrix to obtain a set of eigenvalues ​​and corresponding eigenvectors. Each eigenvalue corresponds to an eigenvector. The magnitude of the eigenvalue represents the magnitude of the data variance in the direction of the corresponding eigenvector, and the eigenvector represents the main direction of the data variance distribution.

[0110] (5) Sort the eigenvalues ​​in descending order and simultaneously arrange the eigenvectors corresponding to each eigenvalue in the same order. The purpose of sorting is to select the principal components with the largest variance contribution first. After sorting, the first eigenvalue is the largest variance contribution value and the last eigenvalue is the smallest variance contribution value.

[0111] (6) Calculate the variance contribution rate corresponding to each feature value. The variance contribution rate is used to measure the contribution ratio of a single principal component to the total variance of the original data. The specific operation is to first calculate the sum of all feature values, then divide the single feature value by the sum of feature values ​​to obtain the variance contribution rate corresponding to the feature value. Calculate the variance contribution rate of all feature values ​​in turn. The value range of the variance contribution rate is between 0 and 1. The sum of all variance contribution rates is 1.

[0112] (7) Calculate the cumulative variance contribution rate. The cumulative variance contribution rate is used to measure the cumulative contribution ratio of the first N principal components to the total variance of the original data. The specific operation is to start from the first variance contribution rate after sorting and add each variance contribution rate in turn to obtain a set of cumulative variance contribution rates. The first cumulative variance contribution rate is equal to the first variance contribution rate, the second cumulative variance contribution rate is equal to the sum of the first two variance contribution rates, and so on, until the last cumulative variance contribution rate is equal to one.

[0113] (8) Set a preset contribution rate threshold, which is set to 95% in this embodiment;

[0114] (9) Screen the principal components that meet the conditions. The specific operation is to start from the first principal component after sorting and check the corresponding cumulative variance contribution rate one by one. When the cumulative variance contribution rate is greater than 95%, stop screening and keep all the principal components that have been checked. Discard the principal components that have not been checked.

[0115] (10) Extract the eigenvectors corresponding to the principal components to be selected and retained. These eigenvectors constitute the principal component transformation matrix. The number of rows in the transformation matrix is ​​equal to the dimension of the original eigenvectors, and the number of columns is equal to the number of principal components to be selected and retained. Each column vector in the matrix corresponds to a retained principal component direction.

[0116] (11) Generate the time-series mutation feature factor after dimensionality reduction. When generating, the standardized multidimensional mutation feature vector is multiplied with the principal component transformation matrix. The multiplication process is carried out according to the matrix multiplication rules. The final result is the time-series mutation feature factor after dimensionality reduction. Its dimension is equal to the number of principal components that are selected and retained. The data order is strictly arranged according to the principal component sorting result, and more than 95% of the effective information in the original data is completely retained.

[0117] S1.5: Load the pre-built association recognition model; the association recognition model is a support vector machine classifier based on radial basis function kernel function; the support vector machine classifier is a discriminative classification model based on statistical learning theory; the dimensionality-reduced temporal mutation feature factors are used as input vectors and arranged in chronological order to form a feature input sequence; each trained support vector is used as a reference vector, and the radial basis function kernel function is used to calculate the distance between the input vector and each reference vector in turn, and then the distance is substituted into the kernel function formula to obtain the corresponding kernel function value. Each support vector corresponds to a kernel function value, and all kernel function values ​​are output in the order of the support vectors to form a kernel function value sequence. The bandwidth parameter of the radial basis function kernel function is set to 0.5, the penalty coefficient is set to 10, and the number of classification categories is fixed at 4. The support vectors are obtained by using a sequence minimum optimization algorithm from... The data was trained using historical pollution intrusion sample data. A one-to-many classification strategy was adopted, and independent sub-classification hyperplanes were constructed for each of the four pollution types. For the kernel function value corresponding to each type, each kernel function value was multiplied by the corresponding weight coefficient in sequence, and then all the multiplication results were accumulated to obtain the comprehensive weighted sum value corresponding to each of the four pollution types. The comprehensive weighted sum value is used as the discrimination index for each type. The larger the value, the higher the matching degree of the sample with the pollution type. The comprehensive weighted sum value corresponding to each type is input into a sign function. The output of the sign function is used to identify the offset direction of the sample relative to the hyperplane of that type. The magnitudes of the comprehensive weighted sum values ​​of the four types are compared, and the pollution type corresponding to the maximum value is selected as the final identification result. The corresponding pollution type identifier is output, which is one of the following: water mist intrusion, dust particle intrusion, fuel mixing, or bubble entrainment.

[0118] Furthermore, the training set of the association recognition model was collected from four types of exogenous pollution intrusion events that occurred during the actual operation of the hydraulic system. Each sample data contains a time-series mutation feature factor after dimensionality reduction and the corresponding pollution type label. The time-series mutation feature factor is fixed to 3 dimensions, corresponding to the number of principal components obtained after the previous dimensionality reduction. There are four types of pollution type labels, corresponding to water mist intrusion, dust particle intrusion, fuel mixing, and bubble entrainment. The total number of sample data is set to 1200, of which the number of samples corresponding to each type of pollution is 300, and the number of samples of each type is kept balanced. All sample data are collected under actual working conditions and no data augmentation processing is performed.

[0119] Furthermore, the sequential minimum optimization algorithm is used to train the support vector machine classifier. The upper limit of the number of iterations in the training process is set to 1000, and the convergence threshold is set to 0.001. In each iteration, two Lagrange multipliers are randomly selected for optimization, while the other multipliers remain fixed. The optimization is repeated until the model error is less than the convergence threshold or the upper limit of the number of iterations is reached. No early stopping mechanism is set in the training process, and no batch training method is used. All sample data are input into the model at one time to complete the training. Support vectors are automatically selected during the training process. The number of support vectors is automatically determined by the algorithm according to the sample distribution, without the need to manually set an upper limit. After the training is completed, the model weight parameters, radial basis kernel function parameters, and support vector data are saved to form a pre-built association recognition model file. The sequential minimum optimization algorithm is the prior art in this field and is not an inventive solution of this application. It will not be described in detail here.

[0120] S2: Construct a digital twin image of the hydraulic system, predict the pollution diffusion path sequence by combining the pollution type, and calculate the damage priority deviation coefficient between the actual pollutant diffusion state and the pollution diffusion path sequence;

[0121] The construction of the digital twin mirror image of the hydraulic system includes:

[0122] S2.1: Collect the topological parameters of the physical entities of the hydraulic system and construct a geometric model of the hydraulic system; the topological parameters include pipeline parameters, valve port diameter parameters, hydraulic pump displacement parameters, hydraulic motor displacement parameters, oil tank volume parameters, and filter accuracy parameters;

[0123] Furthermore, the specific steps of S2.1 include:

[0124] (1) Prepare the data collection tools and basic data. The data collection tools include a steel tape measure, a laser rangefinder, an angle measuring instrument, a vernier caliper, a volume measuring bucket, a precision testing template, a high-definition camera, and paper recording forms. All tools are calibrated before use to ensure that the measurement accuracy meets the industrial standard. The measurement accuracy of the laser rangefinder is set to 1mm, the measurement accuracy of the angle measuring instrument is set to 1°, the measurement accuracy of the vernier caliper is set to 0.02mm, and the scale accuracy of the volume measuring bucket is set to 1L. At the same time, prepare the original design drawings, assembly drawings, and equipment factory parameter manuals of the hydraulic system as the reference for parameter collection to avoid data collection omissions or parameter deviations.

[0125] (2) The hydraulic system is divided into five areas according to its function: power unit area, control unit area, execution unit area, oil storage area and filtration and purification area. The power unit area includes hydraulic pumps and connecting pipelines, the control unit area includes various control valves and connecting pipelines, the execution unit area includes hydraulic motors, hydraulic cylinders and connecting pipelines, the oil storage area includes hydraulic oil tanks, and the filtration and purification area includes various filters and connecting pipelines. Parameters are collected for each area to ensure coverage of all physical entities of the system and to ensure that no key components and pipelines are missed.

[0126] (3) Collect pipeline parameters, including pipeline length, pipeline inner diameter, and bend angle. When collecting parameters, first distinguish between high-pressure pipelines, low-pressure pipelines, and return oil pipelines. High-pressure pipelines refer to pipelines with a working pressure greater than 16MPa, low-pressure pipelines refer to pipelines with a working pressure less than or equal to 16MPa, and return oil pipelines refer to pipelines where oil flows back to the oil tank. Measure each independent pipeline section sequentially. The pipeline length is measured along the centerline of the pipeline using a laser rangefinder, representing the straight-line distance from one end of the pipeline interface to the other end. The measurement accuracy for high-pressure pipeline length is controlled within 5mm, while the measurement accuracy for low-pressure pipelines and return oil pipelines is controlled within 10mm. The inner diameter of the pipeline was measured using vernier calipers. The measurement location was selected in the middle of the straight section of the pipeline, avoiding deformation areas such as bends and joints. Three different locations were measured for each pipeline, and the average value was taken. The measurement accuracy of the inner diameter of the high-pressure pipeline was controlled within 0.1 mm, and the measurement accuracy of the inner diameter of the low-pressure pipeline and the return oil pipeline was controlled within 0.2 mm. The bending angle of the pipeline bend was measured using an angle measuring instrument, distinguishing between 90° bends, 45° bends, and bends of arbitrary angle. Each bend was measured twice and the average value was taken. The angle measurement accuracy was controlled within one degree. All pipeline parameters were recorded one by one according to pipeline number, region, type, length, inner diameter, and bend angle.

[0127] (4) Collect valve port diameter parameters. Valve port diameter refers to the diameter of the oil flow port inside various control valves, covering all valves such as multi-way valves, relief valves, check valves, throttle valves, and pressure reducing valves. When collecting data, the valves should be safely removed from the system to avoid operation under pressure. Use vernier calipers to measure the effective flow diameter of the valve port. The measurement position is the minimum flow cross section of the inner wall of the valve port. Measure three different positions for each valve port and take the average value. The measurement accuracy is controlled within 0.02mm. The main valve port diameter of a multi-way valve is usually 20mm, the valve port diameter of a relief valve is usually 10mm, the valve port diameter of a check valve is usually 8mm, and the valve port diameter of a throttle valve is usually 4mm. Record the valve port diameter parameters one by one according to the valve number, type, installation position, and diameter value.

[0128] (5) Collect hydraulic pump displacement parameters. Hydraulic pump displacement refers to the volume of oil discharged by the hydraulic pump per revolution. The volumetric method is used for actual measurement. During measurement, connect the hydraulic pump outlet to a standard volume measuring bucket, control the hydraulic pump to run stably at the rated speed of 1500 rad / min, record the volume of oil discharged within 60 seconds, calculate the displacement per revolution, and control the measurement accuracy to 0.1 mL / rad. The displacement of common engineering machinery hydraulic pumps is 40 mL / rad to 100 mL / rad. For example, the displacement of the main hydraulic pump is set to 63 mL / rad, and the displacement of the auxiliary hydraulic pump is set to 40 mL / rad. At the same time, check against the hydraulic pump factory parameter manual to ensure that the deviation between the measured value and the rated value does not exceed 2%. Record the hydraulic pump displacement parameters one by one according to the pump body number, type, rated speed, measured displacement, and rated displacement.

[0129] (6) Collect hydraulic motor displacement parameters. Hydraulic motor displacement refers to the volume of oil required for each revolution of the hydraulic motor. The measurement method is the same as that for hydraulic pump displacement. The volumetric method is used for actual measurement. Connect the hydraulic motor inlet to a standard volume measuring barrel, control the hydraulic motor to run stably at the rated speed of 1200 rad / min, record the volume of oil input within 60 seconds, and calculate the displacement per revolution. The measurement accuracy is controlled within 0.1 mL / rad. The displacement of common hydraulic motors is 30 mL / rad to 80 mL / rad. For example, the displacement of the walking hydraulic motor is set to 50 mL / rad, and the displacement of the rotary hydraulic motor is set to 40 mL / rad. Check against the hydraulic motor's factory parameter instruction manual to ensure that the deviation between the measured value and the rated value does not exceed 2%. Record the hydraulic motor displacement parameters one by one according to the motor number, type, rated speed, measured displacement, and rated displacement.

[0130] (7) Collect oil tank volume parameters. The oil tank volume refers to the effective volume of oil that can be contained inside the hydraulic oil tank. The actual measurement is carried out by water injection. After cleaning the hydraulic oil tank, seal all oil outlets and oil return ports. Slowly inject clean water from the oil tank filler port until the clean water reaches the rated oil level mark of the oil tank. Record the total volume of the injected clean water, which is the effective volume of the oil tank. The measurement accuracy is controlled within 1 liter. The volume of hydraulic oil tanks of common engineering machinery is 200 liters to 500 liters. For example, the main oil tank volume is set to 300 liters and the spare oil tank volume is set to 100 liters. At the same time, measure the external length, width and height of the oil tank. The length is set to 1000 mm, the width is set to 600 mm and the height is set to 800 mm for geometric model construction. Record the oil tank volume parameters one by one according to the oil tank number, effective volume, external dimensions and rated oil level height.

[0131] (8) Collect filter accuracy parameters. Filter accuracy refers to the smallest impurity particle diameter that the filter element can filter. It covers high pressure filter, return oil filter and suction oil filter. The high pressure filter accuracy is set to 10 micrometers, the return oil filter accuracy is set to 20 micrometers, and the suction oil filter accuracy is set to 100 micrometers. When collecting, check the filter element's certificate of conformity and technical parameter table for confirmation. No actual measurement is required. At the same time, record the filter's external dimensions, interface specifications and installation position. The filter shell length is set to 200mm and the diameter is set to 80mm. Record the filter accuracy parameters one by one according to filter number, type, filtration accuracy, external dimensions and installation position.

[0132] (9) Summarize the collected parameters such as pipe length, pipe inner diameter, elbow angle, valve port diameter, hydraulic pump displacement, hydraulic motor displacement, oil tank volume and filter accuracy, remove duplicate data and obvious abnormal data. Abnormal data refers to values ​​that exceed the normal parameter range of the hydraulic system, such as pipe inner diameter less than 4mm and oil tank volume less than 100 liters. Perform a second on-site retest on the questionable parameters to ensure that all parameters are true, accurate, complete and consistent. After sorting, form a general table of hydraulic system topology parameters, mark the parameter collection time, collection personnel and verification results, and use it as the sole data basis for geometric model construction.

[0133] (10) Three-dimensional modeling software was used for modeling. The modeling scale was strictly set to 1:1. The modeling sequence was from basic components to overall assembly. First, a three-dimensional model of the hydraulic oil tank was constructed. A cuboid structure was drawn according to the outer dimensions of the oil tank. The positions and dimensions of the oil tank filling port, oil outlet, oil return port and sensor mounting holes were marked. Second, a three-dimensional model of various pipelines was constructed. According to the pipeline length, inner diameter and bend angle, straight pipe sections and bend sections were drawn and connected according to the actual route. High pressure pipelines, low pressure pipelines and return oil pipelines were distinguished and marked with different colors. Next, a three-dimensional model of various valves was constructed. According to the valve port diameter and outer dimensions, the valve body, valve core and interface structure were drawn and assembled to the corresponding nodes of the pipeline according to the actual installation position. Then, a three-dimensional model of the hydraulic pump and hydraulic motor was constructed. According to the displacement parameters and outer dimensions, the pump body, motor housing, input shaft and output shaft structure were drawn and connected to the power unit and execution unit pipelines. Finally, a three-dimensional model of various filters was constructed. According to the outer dimensions and interface specifications, the filter shell and filter element cavity structure were drawn and connected in series to the filtration and purification area pipelines.

[0134] (11) According to the actual assembly relationship of the hydraulic system, all components such as oil tank, pipeline, valve, hydraulic pump, hydraulic motor, and filter are assembled as a whole to ensure that the interfaces between components are matched, the positions are accurate, and the connection relationships are correct. The original size and topological connection of the components are not changed during the assembly process. After the assembly is completed, the model details are optimized, redundant lines and redundant structures are removed, and all key nodes are marked, including the sensor installation position, valve port position, pump outlet position, motor inlet position and filter interface position. Each key node is marked with a unique number to facilitate data docking and simulation positioning of the digital twin mirror. Finally, a complete geometric model of the hydraulic system is formed, which intuitively restores the physical entity's structural form, size parameters, component layout and pipeline routing.

[0135] S2.2: Construct a sub-model of oil flow mechanism based on the Navier-Stokes equations, and construct a sub-model of pollutant transport mechanism based on the principles of mass conservation and momentum conservation.

[0136] Furthermore, the oil flow mechanism sub-model is used to accurately simulate the steady-state and transient flow processes of oil within the hydraulic system, outputting key flow field parameters such as oil velocity, pressure, flow rate, and temperature distribution. It is applicable to hydraulic system operating pressures of 16MPa to 31MPa, oil temperatures of 30℃ to 60℃, and oil kinematic viscosity of 20mm. 2 / s to 40mm 2 High-pressure viscous flow scenario with a flow rate of / s.

[0137] Furthermore, the pollutant transport mechanism sub-model is used to simulate the transport, diffusion, and deposition processes of exogenous pollutants such as solid particles, water mist, fuel oil, and bubbles in oil as they flow with the oil. It outputs parameters such as pollutant concentration distribution, particle size distribution, and deposition location. It is applicable to hydraulic system flow field environments with pollutant particle size ranging from 0.1 micrometers to 25 micrometers, pollutant concentration ranging from 0 to 50 mg / L, and oil flow velocity ranging from 0.5 m / s to 5 m / s. The model construction strictly follows the principles of mass conservation and momentum conservation, and the flow field data output by the oil flow mechanism sub-model is coupled to achieve bidirectional linkage.

[0138] Furthermore, a sub-model of the oil flow mechanism is constructed based on the Navier-Stokes equations, including:

[0139] (1) Determine the physical properties of the hydraulic fluid, wherein the density of the hydraulic oil is set to 870 kg / m³. 3 The dynamic viscosity of the oil is set to 0.028 Pa·s, corresponding to the viscosity value at an oil temperature of 40℃. The compressibility coefficient of the oil is set to 0.0001 MPa, which is used to describe the small compressibility characteristics of the oil under high pressure. The specific heat capacity of the oil is set to 2100 J / (kg·℃). The thermal conductivity of the oil is set to 0.14 W / (m·℃).

[0140] (2) Import the completed hydraulic system geometric model into the fluid simulation software and adopt the structured mesh generation method. The mesh size of the straight pipe section is set to 2mm, the mesh size of the complex flow areas such as elbows, valve ports, and pump outlets is densified to 0.5mm, and the mesh size of the large space area of ​​the oil tank is widened to 5mm. After the mesh generation is completed, the total number of meshes is controlled between 1.5 million and 2 million to ensure that the mesh resolution of the complex flow areas is sufficient and the calculation efficiency is reasonable. When checking the mesh quality, the orthogonal quality threshold is set to 0.85. Meshes below the orthogonal quality threshold are regenerated to avoid distorted meshes affecting the solution stability.

[0141] (3) Set the boundary conditions for solving the Navier-Stokes equations. The boundary conditions are divided into four categories: inlet boundary, outlet boundary, wall boundary and internal coupling boundary. The inlet boundary is set as the hydraulic pump outlet, and the velocity inlet condition is adopted. The oil flow velocity is set to 2m / s according to the pump displacement, and the oil temperature is set to 40℃. The outlet boundary is set as the return oil pipeline inlet, and the pressure outlet condition is adopted. The outlet pressure is set to 0.2MPa to simulate the return oil back pressure. The wall boundary includes the inner wall of the pipeline, the inner wall of the valve, the inner wall of the pump cavity, and the inner wall of the oil tank. All of them are set as non-slip thermal insulation walls, and the wall roughness is set to 0.015mm, which corresponds to the roughness standard of the inner wall of industrial steel pipe. The internal coupling boundary is set as a filter element, throttling port and other porous or throttling structure. The viscous resistance coefficient of the filter element is set to 15000 per meter, and the inertial resistance coefficient is set to 200 per meter.

[0142] (4) A pressure-based transient solver is used with a time step of 0.001 seconds to ensure the capture of transient flow characteristics such as oil pressure pulsation. The solution convergence residual threshold is set to 0.0001, the upper limit of the number of iterations is set to 500, the second-order upwind scheme is used for the convection term, the central difference scheme is used for the diffusion term, and the pressure-velocity coupling algorithm adopts the semi-implicit method for pressure-linked equations (SIMPLE) to adapt to the high-speed flow scenario of incompressible fluids. It has high solution stability and fast convergence speed. The standard two-equation turbulence model is selected for the turbulence model. The turbulence intensity is set to 5%, and the turbulence viscosity ratio is set to 10 to simulate the turbulence flow effect in the high-pressure pipeline.

[0143] Furthermore, the convection term discretization method adopts a second-order precision upwind scheme. When calculating the changes in physical quantities caused by fluid flow and transport, the second-order precision upwind scheme uses the data of two adjacent nodes upstream of the current calculation point for linear interpolation. It has second-order calculation precision, which can accurately capture the convection effect when high-pressure oil flows at high speed. At the same time, it has good numerical stability and is not prone to calculation oscillation. The second-order precision upwind scheme is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0144] Furthermore, the diffusion term is discretized using a central difference scheme. When calculating the changes in physical quantities caused by the viscosity of the fluid, the central difference scheme uses the average of the data from the adjacent nodes on the left and right sides of the current calculation point to perform differential calculation. It has second-order calculation accuracy and can accurately simulate the momentum and energy diffusion caused by the viscosity of oil. The calculation results are smooth and highly accurate. The central difference scheme is prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0145] Furthermore, the SIMPLE algorithm is used to solve the coupling relationship between pressure and velocity. The SIMPLE algorithm first estimates the velocity field, then solves the pressure field correction, and finally corrects the velocity field through a step-by-step iterative approach. It implicitly handles the coupling relationship between pressure and velocity in a step-by-step manner, making it suitable for flow calculations of weakly compressible fluids such as hydraulic oil under high pressure and high speed conditions. It has the characteristics of high computational stability, fast convergence speed, and low divergence. The SIMPLE algorithm is prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0146] Furthermore, the standard two-equation turbulence model is selected as the turbulence model. The standard two-equation turbulence model calculates the turbulent pulsating kinetic energy and the turbulent pulsating kinetic energy dissipation rate by solving two independent transport equations, thereby describing the complex pulsating characteristics of oil turbulence flow in high-pressure pipelines. In the model, the turbulence intensity is fixed at 5% to characterize the ratio of turbulent pulsating velocity to average flow velocity in oil flow; the turbulent viscosity ratio is fixed at 10 to characterize the ratio of turbulent viscosity coefficient to fluid molecular viscosity coefficient, which can accurately simulate the mixing, diffusion and energy loss effects brought about by oil turbulence in high-pressure pipelines. The transport equation is prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0147] (5) Start the solver to begin iterative calculation. After each iteration, the residual convergence is automatically checked. When all physical quantity residuals are below 0.0001 and the inlet and outlet flow difference is stable within 1%, it is determined to be converged. Output the oil flow field calculation results, including the flow velocity, pressure, flow rate and temperature distribution data of each node. Then, the model is verified using the measured flow field data of the hydraulic system. The measured pressure and flow velocity data of 10 key measuring points are compared with the model calculation results. The error is required to be no more than 5%. If the error exceeds the range, check the mesh quality, boundary conditions or physical property parameters, correct and recalculate until the verification is qualified. Finally, an oil flow mechanism sub-model that can stably output flow field parameters is formed.

[0148] Furthermore, a sub-model of pollutant transport mechanism is constructed based on the principles of mass conservation and momentum conservation, including:

[0149] (1) Determine the physical property parameters of the pollutants to provide the physical property conditions for solving the transport equation. The density of solid dust particles is set to 2800 kg / m³. 3 The particle shape factor was set to 0.65 to describe the flow resistance characteristics of irregular particles; the water mist particle density was set to 1020 kg / m³. 3 The shape factor is set to 0.95, approximately spherical; the fuel density is set to 830 kg / m³. 3 The miscibility coefficient with oil was set to 0.9 to describe the mutual solubility and diffusion characteristics; the bubble density was set to 1 kg / m³. 3 The surface tension coefficient is set to 0.025 N / m to describe the bubble breakup and merging characteristics; the initial particle size distribution adopts a normal distribution, the average particle size is set to 10 micrometers, the standard deviation is set to 3 micrometers, and all pollutant physical property parameters are set according to the pollution type, and a parameter library is built for easy access;

[0150] (2) The fluid domain mesh completed by reusing the oil flow mechanism sub-model is added on the basis of the fluid domain mesh, and transport variables such as pollutant concentration and particle size distribution are added. The transport calculation domain is defined to completely overlap with the fluid domain, including all oil flow areas, to ensure that the pollutant transport simulation covers the entire path of oil flow. At the same time, the pollutant deposition boundary is defined. Pipe bends, valve ports, filter elements and other locations are set as depositable boundaries, the bottom of the oil tank is set as a strong deposition boundary, and the wall deposition rate coefficient is set to 0.001 to describe the deposition and adhesion characteristics of particles on the wall.

[0151] (3) A mass transport equation for pollutants is constructed based on the principle of mass conservation. The mass conservation equation describes the mass change law of pollutants in oil. The core is that the mass of pollutants flowing in minus the mass of pollutants flowing out minus the mass of pollutants deposited equals the rate of mass change. When constructing the equation, convection, diffusion, source and deposition terms are introduced. The convection term is driven by the velocity field output by the oil flow mechanism sub-model. The diffusion term includes molecular diffusion and turbulent diffusion. The molecular diffusion coefficient is set to 0.00001m. 2 / s, the turbulent diffusion coefficient is determined by the turbulence model parameters, the source term is set as the pollution intrusion source, the initial intrusion location is set as typical locations such as the breather interface, piston rod seal, and new oil filling port, and the deposition term is determined by the product of the wall deposition rate coefficient and the wall pollutant concentration to ensure that the pollutant mass is strictly conserved during the transport process;

[0152] (4) Based on the principle of momentum conservation, a momentum transport equation for pollutants is constructed. The momentum conservation equation describes the momentum change law of pollutant particles as they flow with the oil. The core is that the net force on the particles is equal to the mass multiplied by the acceleration. When constructing the equation, the drag force, gravity, buoyancy, lift force and virtual mass force on the particles are considered. The drag force coefficient is determined based on the particle Reynolds number, and the gravitational acceleration is set to 9.81 m / s². 2Buoyancy is calculated from the density difference between oil and contaminants. The lift coefficient is set to 0.01 and the virtual mass force coefficient is set to 0.5 to describe the additional inertial effect when particles accelerate. The equation is coupled with the pressure gradient and velocity gradient data of the oil flow field to realize the linkage transfer of contaminant momentum and oil momentum, and accurately simulate the dynamic process of particles following the oil flow, deviating and colliding.

[0153] (5) Set the solution parameters for the pollutant transport equation. The transient solver is the same as that used in the oil flow mechanism sub-model. The time step is kept at 0.001s. The solution convergence residual threshold is set to 0.0001. The upper limit of the number of iterations is set to 300. The pollutant concentration discrete scheme adopts the first-order upwind scheme to balance the calculation accuracy and efficiency. The particle size distribution adopts the multiphase flow discrete phase model. The number of discrete phase particles is set to 100,000. The tracking time is set to 2 seconds to cover the complete process of pollutant from intrusion to diffusion. The release frequency of the intrusion source is set to 1,000 particles per second to simulate the continuous pollution intrusion scenario.

[0154] (6) Couple the real-time flow field data output by the oil flow mechanism sub-model, start the iterative calculation of the pollutant transport equation, and update the flow field data synchronously after each iteration to ensure bidirectional coupling between the flow field and pollutant transport. After convergence, output the results such as pollutant concentration distribution, particle size distribution, deposition location, and transport path. Then, use hydraulic system pollution test data to verify the model. Select the measured pollutant concentration and particle size data of 5 key nodes and compare them with the model calculation results. The error is required to be no more than 6%. If the error exceeds the range, correct the pollutant physical property parameters or deposition coefficient, recalculate until the verification is qualified, and finally form a pollutant transport mechanism sub-model that can accurately simulate the pollutant transport process.

[0155] S2.3: Connect the oil flow mechanism sub-model and the pollutant transport mechanism sub-model through a standardized data interface. Set the interface data transmission frequency to 100Hz to ensure that the flow field data is transmitted to the pollutant transport sub-model in real time and the pollutant transport results are synchronously fed back to the oil flow mechanism sub-model, realizing bidirectional real-time coupling. After integration, a complete joint mechanism model of flow field-pollutant transport is formed, and the output value of the joint mechanism model is obtained.

[0156] S2.4: Collect measured values ​​of multiple flow field physical quantities corresponding to key measuring points of the hydraulic system's physical entity. These multiple flow field physical quantities include oil flow velocity, pressure, oil temperature, and contaminant concentration. Simultaneously acquire simulated physical quantities output by the joint mechanism model at corresponding measuring points and at the same time. For the same measuring point and the same physical quantity, calculate the difference between the simulated physical quantity and the measured value to obtain the single-measurement-point single-physical-quantity deviation. Finally, perform min-max normalization on the single-measurement-point single-physical-quantity deviations corresponding to all measuring points and all physical quantities, and then splice them into one-dimensional time-series data according to the spatial order and temporal order of the measuring points to form a deviation sequence.

[0157] S2.5: Construct a deviation correction sub-model based on a long short-term memory network, take the deviation sequence as input, train the deviation correction sub-model to output the deviation prediction value in a preset time domain;

[0158] Furthermore, the deviation correction sub-model is used to learn the temporal variation law of the deviation sequence between the simulated values ​​of the hydraulic system flow mechanism and the measured values ​​of the physical entity, so as to achieve accurate prediction of the deviation values ​​within the preset time domain and provide a basis for the correction of the simulation results of the digital twin mirror. The model input is a one-dimensional temporal deviation sequence, with each time step corresponding to a deviation value. The model output is a sequence of predicted deviation values ​​within the preset time domain. The preset time domain is set to 1 second, corresponding to 1000 time steps, with each time step corresponding to 1ms. The model as a whole is adapted to the temporal deviation prediction scenario under the high-pressure dynamic working conditions of the hydraulic system.

[0159] Furthermore, the specific steps of S2.5 include:

[0160] (1) Prepare the deviation sequence dataset required for model training. The deviation sequence dataset comes from the historical operation record of the digital twin mirror of the hydraulic system. The collection time is set to 1000 hours of continuous operation. The data sampling frequency is consistent with the flow mechanism simulation calculation frequency, that is, a deviation value is generated every 1ms, forming an original deviation sequence with a total length of 3.6 million time steps. The original deviation sequence includes simulation deviations caused by various working conditions such as oil temperature fluctuation, component wear, pipeline scaling, and flow disturbance. The data type is uniformly floating-point value, without any classification labels, and is only a pure time series value sequence.

[0161] (2) Preprocessing operation is performed on the original deviation sequence dataset. The preprocessing only includes two steps: normalization and sequence partitioning. The normalization process adopts the min-max normalization method, which maps all values ​​in the original deviation sequence to the range of 0 to 1. When calculating, the maximum and minimum values ​​in the original sequence are determined first, and then the minimum value is subtracted from each original deviation value and divided by the difference between the maximum and minimum values. After normalization, the difference in the dimensions of the deviation values ​​is eliminated, and the large range of values ​​is avoided from affecting the stability of model training. The sequence partitioning is carried out in a ratio of 8:1:1, dividing the normalized complete deviation sequence into training set, validation set and test set. The partitioning process is strictly carried out in the temporal order, without disrupting the temporal correlation of the data, to ensure that it meets the data characteristics of the temporal prediction task. Among them, min-max normalization is the existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0162] (3) Construct a long short-term memory network structure. The bias correction sub-model adopts a single-layer long short-term memory network superimposed with a fully connected layer as its basic architecture. The overall network hierarchy is as follows: input layer, long short-term memory layer, fully connected layer, and output layer. There is a unidirectional fully connected relationship between each layer. Data enters the network from the input layer, extracts temporal features through the long short-term memory layer, and then outputs the prediction results through the fully connected layer. The number of nodes in the network input layer is consistent with the time step length of the input bias sequence. Each node corresponds to the bias value of one time step. The length of the input sequence is set to 200 time steps, that is, 200 consecutive bias values ​​are input to capture the short-term and medium-term temporal change patterns of the bias sequence.

[0163] (4) Set the core structural parameters of the long short-term memory layer. The long short-term memory layer is the core feature extraction layer of the model. It contains four key modules: input gate, forget gate, output gate and cell state. It is used to effectively remember the long-term dependencies of time series data. The number of hidden units in the long short-term memory layer is set to 128. This value is the standard empirical value in the hydraulic time series deviation prediction scenario, which balances the model's feature extraction capability and computational complexity. The activation function is the hyperbolic tangent function, which is used to introduce nonlinear transformation and improve the model's ability to fit complex time series patterns. The dropout regularization parameter inside the long short-term memory layer is set to 0.2, which is used to randomly discard some hidden unit outputs to prevent overfitting during model training and enhance the model's generalization ability. The hyperbolic tangent function is the existing technology in this field and is not an inventive solution of this application. It will not be described in detail here.

[0164] (5) Set the structural parameters of the fully connected layer and the output layer. The temporal feature vector output by the long short-term memory layer is passed to the fully connected layer. The number of nodes in the fully connected layer is set to 64. The activation function is a linear activation function, which is used to perform nonlinear mapping and feature fusion on the temporal features without introducing additional nonlinear transformations to avoid destroying the linear change trend of the deviation value. The feature vector output by the fully connected layer is passed to the output layer. The number of nodes in the output layer is consistent with the preset time domain length of the model output, that is, it is set to 1000 nodes. Each node corresponds to the deviation prediction value of one time step within 1 second. The activation function of the output layer is a linear activation function.

[0165] (6) Determine the core parameters and optimization configuration for model training, including: the model training adopts supervised learning, the loss function is the mean squared error loss function, which is used to calculate the error between the model prediction deviation value and the true deviation value. This loss function is suitable for continuous numerical prediction tasks and can effectively measure the degree of deviation between the predicted value and the true value; the optimization algorithm is the Adam adaptive moment estimation optimization algorithm, which has fast convergence speed and high training stability, and is suitable for the training scenario of time series prediction models; the training batch size is set to 32, that is, 32 sample sequences are input in each iteration to balance training efficiency and gradient update stability; the initial learning rate is set to 0.001 to control the magnitude of model parameter updates, and to avoid the model parameters from oscillating due to an excessively large learning rate or the training from being too small; the upper limit of the number of training iterations is set to 200 rounds to prevent the model from being overtrained, while reserving sufficient iterations to ensure model convergence; the patience value of the early stopping mechanism is set to 20 rounds. When the validation set loss value no longer decreases for 20 consecutive rounds, the training is automatically terminated to avoid model overfitting. The adaptive moment estimation optimization algorithm is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0166] (7) Input the preprocessed training set bias sequence samples into the long short-term memory network in batches. The model calculates the loss function value batch by batch according to the rules of the moment estimation optimization algorithm. The gradient of each network layer parameter is calculated layer by layer through the backpropagation algorithm. The network weight parameters and bias parameters are updated along the gradient descent direction. After each round of iterative training of all training set samples is completed, the performance of the current model is evaluated using the validation set data. The validation set loss value is calculated. The model is monitored in real time to see if overfitting occurs. The model parameters with the best validation set loss value are automatically saved during the training process as the final optimal model. The backpropagation algorithm is the existing technology in this field and is not an inventive solution of this application. It will not be described in detail here.

[0167] (8) After training is terminated, load the model parameters with the optimal loss value on the validation set, evaluate the generalization ability of the model using the bias sequence data of the test set, calculate the mean square error of the test set, and require the mean square error of the test set to not exceed 0.0005 to ensure that the model has good prediction accuracy and generalization ability. If the evaluation result is not up to standard, check the dataset preprocessing process, network structure parameters or training configuration, correct and re-execute the training process; after the evaluation is qualified, solidify the network structure, weight parameters, bias parameters and normalization parameters of the optimal model to form a bias correction sub-model file that can be directly called. The model file contains complete network configuration information and training completion parameters and can be deployed and run independently.

[0168] (9) Deploy the solidified deviation correction sub-model, receive the real-time deviation sequence calculated from the output value of the hydraulic system flow mechanism model and the measured value of the physical entity, and extract the latest 200 consecutive deviation values ​​as the input sequence according to the model input requirements and input them into the deviation correction sub-model; after receiving the input sequence, the model automatically performs forward calculation and outputs the deviation prediction value sequence of 1000 time steps within 1 second, which is the deviation prediction value in the preset time domain; transmit the output deviation prediction value sequence to the digital twin mirror to correct the simulation output value of the flow mechanism model and complete the real-time correction of the simulation deviation.

[0169] S2.6: Add the predicted deviation value to the output value of the joint mechanism model to obtain the twin output value;

[0170] S2.7: The hydraulic system geometric model, the oil flow mechanism sub-model, the contaminant transport mechanism sub-model, and the deviation correction sub-model are coupled together through a data interface, and the twin output value is used as the output end of the digital twin mirror to form a digital twin mirror of the hydraulic system.

[0171] The method of predicting pollution diffusion path sequences by combining pollution types includes:

[0172] S2.8: Query the corresponding physical property parameters from the preset pollutant attribute database according to the pollution type identifier; the physical property parameters include average particle size, particle density, particle shape coefficient, oil miscibility index and bubble surface tension coefficient;

[0173] Furthermore, the specific steps in S2.8 include:

[0174] (1) Establish and initialize the pollutant attribute database, which is used to store the physical attribute parameters corresponding to the four types of external pollutants in the hydraulic system. The database structure adopts a classified storage method and is divided into four independent data tables according to the type of pollution, which correspond to water mist intrusion pollutants, dust particle intrusion pollutants, fuel mixed pollutants, and bubble entrainment pollutants respectively. Each data table is uniformly set with 5 standard fields, namely, average particle size, particle density, particle shape coefficient, oil miscibility index, and bubble surface tension coefficient. The database storage format adopts text structured storage, the field order is fixed, and the data type is uniformly floating point type.

[0175] (2) Enter the physical property parameters of water mist intrusion pollutants. Water mist intrusion pollutants are fine liquid water droplets. The parameters are set as follows: average particle size is set to 5 micrometers, and particle density is set to 1010 kg / m³. 3The particle shape coefficient is set to 0.95, the oil miscibility index is set to 0.85, and the bubble surface tension coefficient is set to 0. This type of pollutant does not have bubble characteristics, so the corresponding parameter is set to 0. All parameters are entered into the corresponding data table fields, and the numerical format is uniformly in decimal form.

[0176] (3) Enter the physical property parameters of intrusive dust particles. Intrusive dust particles are solid impurity particles. The parameters are set as follows: average particle size is set to 12 micrometers, and particle density is set to 2800 kg / m³. 3 The particle shape coefficient is set to 0.65, the oil miscibility index is set to 0.1, the bubble surface tension coefficient is set to 0, and the pollutants are solid particles that are insoluble in oil and have no bubble characteristics. The corresponding parameters are set to 0. All parameters are entered into the corresponding data table fields.

[0177] (4) Enter the physical property parameters of pollutants mixed in with fuel oil. The pollutants mixed in with fuel oil are light liquid fuel oil. The parameters are set as follows: average particle size is set to 0, and particle density is set to 830 kg / m³. 3 The particle shape coefficient is set to 0, the oil miscibility index is set to 0.95, and the bubble surface tension coefficient is set to 0. This type of pollutant is mixed in at the liquid molecular level and has no solid particle or bubble characteristics. The corresponding particle-related parameters are set to 0, and all parameters are entered into the corresponding data table fields.

[0178] (5) Enter the physical property parameters of air bubble entrainment pollutants. Air bubble entrainment pollutants are tiny air bubbles. The parameters are set as follows: average particle size is set to 0, and particle density is set to 1.225 kg / m³. 3 The particle shape coefficient is set to 0, the oil miscibility index is set to 0.05, the bubble surface tension coefficient is set to 0.025 N / m, the pollutant is a gas bubble, and the corresponding particle-related parameters are set to 0. All parameters are entered into the corresponding data table fields.

[0179] (6) After entering all parameters, perform integrity and consistency checks on the pollutant attribute database. Check each of the five fields in each data table to see if they are complete, if the numerical format is correct, and if the parameter range conforms to the physical characteristics of pollutants in the hydraulic system. For example, the particle density is not negative, the shape coefficient is between zero and one, and the surface tension coefficient conforms to the gas-liquid interface characteristics. Correct entry errors and fill in missing fields during the verification process to ensure that the database data is accurate. After the verification is passed, solidify the database structure and data, set the database to read-only permission to prevent the data from being modified or deleted by mistake. After solidification, the database can be used stably for a long time without frequent parameter modifications.

[0180] (7) Obtain the pollution type identifier from the output of the association identification model. The identifier is in text format and contains only four fixed contents: water mist intrusion, dust particle intrusion, fuel mixing, and bubble entrainment. After receiving, first perform format verification to check whether the identifier content is the above four standard texts, exclude empty values, garbled characters or non-standard identifiers. If the verification fails, return an identifier error message and terminate the query process. After the verification passes, retain the standard pollution type identifier and proceed to the next query operation.

[0181] (8) The received standard pollution type identifier is compared with the names of four data tables in the pollutant attribute database using a precise string matching method. The matching principle is that the identifier text and the data table name must be completely consistent. For example, if the identifier is water mist intrusion, the water mist intrusion pollutant data table is directly matched. If the match is successful, the data table is located and used as the target data table for this query. If the match fails, the system returns a message that there is no corresponding pollutant data and terminates the query process.

[0182] (9) After locating the target data table that has been successfully matched, read the five physical property parameters stored in the target data table in the order of the preset fields. The reading order strictly follows the average particle size, particle density, particle shape coefficient, oil miscibility index, and bubble surface tension coefficient to ensure that the parameter reading order is consistent with the subsequent model input order. Do not modify the original data during the reading process, directly extract the values, and form a complete set of pollutant physical property parameters.

[0183] (10) The five physical property parameters extracted are sorted and output in a fixed order. The output format is structured text. Each parameter is labeled with the corresponding field name to ensure that the parameter meaning is clear and unambiguous. After output, the parameter value range is automatically checked to confirm that all parameters conform to the physical characteristics of pollutants in the hydraulic system. After the check is correct, the parameters are transferred to the pollutant transport mechanism sub-model as the physical property input conditions for simulation calculation.

[0184] S2.9: Assign the physical property parameters to the pollutant transport mechanism sub-model, and use the oil velocity distribution and pressure distribution output by the current oil flow mechanism sub-model as initial boundary conditions to input the pollutant transport mechanism sub-model;

[0185] Furthermore, the specific steps in S2.9 include:

[0186] (1) Confirm the parameter interface and variable definition of the pollutant transport mechanism sub-model. The pollutant transport mechanism sub-model has a pre-set variable storage unit that corresponds one-to-one with the physical property parameters. This unit is used to receive and store five physical property parameters. The variable storage units correspond to the average particle size, particle density, particle shape coefficient, oil miscibility index, and bubble surface tension coefficient, respectively. Each variable storage unit is a floating-point variable. An independent parameter assignment entry is reserved. The assignment entry has a data type verification function and only accepts floating-point values ​​to ensure that the data format matches during parameter assignment and avoids model operation abnormalities caused by type errors.

[0187] (2) The following five parameters were obtained sequentially: average particle size, particle density, particle shape factor, oil miscibility index, and bubble surface tension coefficient. First, the completeness of the parameters was verified to ensure that all five parameters were present and there were no missing items. Second, the range of the parameter values ​​was verified. The verification range for average particle size was set to 0 to 25 micrometers, and the verification range for particle density was set to 1 to 3000 kg / m³. 3 The particle shape coefficient verification range is set to 0~1, the oil miscibility index verification range is set to 0~1, and the bubble surface tension coefficient verification range is set to 0 to 0.03 N / m. All parameter values ​​must fall within the corresponding range. If the verification fails, a parameter abnormality prompt will be returned, and the assignment process will be terminated. After the verification passes, the original values ​​of the 5 parameters will be retained, and the assignment operation will be prepared.

[0188] (3) In a fixed order of average particle size, particle density, particle shape coefficient, oil miscibility index, and bubble surface tension coefficient, the qualified parameter values ​​are assigned to the corresponding variable storage units inside the pollutant transport mechanism sub-model. The assignment process is a direct numerical transfer to ensure that the parameters received by the sub-model are completely consistent with the database query results. After the assignment is completed, the current value of the variable storage unit inside the pollutant transport mechanism sub-model is read and compared with the original parameter values ​​for the second time to confirm that the assignment is successful and the values ​​are without deviation. The assignment of physical property parameters is completed.

[0189] (4) Establish a data interaction channel between the oil flow mechanism sub-model and the pollutant transport mechanism sub-model. The channel adopts a real-time data transmission protocol and the data transmission frequency is set to 100Hz to ensure that the data synchronization delay between the two sub-models does not exceed 10ms, which meets the real-time requirements of the dynamic simulation of the hydraulic system. The channel includes independent velocity distribution data interface and pressure distribution data interface. The velocity distribution data interface is used to transmit the oil velocity vector data of each grid node, and the pressure distribution data interface is used to transmit the oil pressure scalar data of each grid node. The interface has the function of data format standardization.

[0190] (5) Obtain the flow field data output by the oil flow mechanism sub-model at the current moment. The flow field data includes two core data: oil velocity distribution and oil pressure distribution. The current moment is defined as the initial moment of the start of the contamination intrusion simulation. The flow mechanism sub-model completes the steady-state calculation of the flow field at this moment and outputs complete flow field data covering the entire hydraulic system fluid domain grid nodes. The velocity distribution data includes the three-directional velocity components of each grid node, and the pressure distribution data includes the absolute pressure value of each grid node.

[0191] (6) Verify the completeness and validity of the flow field data. First, verify the data coverage to confirm that the velocity and pressure distribution data completely cover all fluid domain grid nodes of the hydraulic system without missing node data. Second, verify the rationality of the data values. The verification range for oil velocity is set to 0 to 5 m / s, and the verification range for oil pressure is set to 0.1 to 35 MPa. The velocity and pressure values ​​of all grid nodes must fall within the corresponding range. If the verification fails, return a flow field data anomaly prompt and terminate the boundary condition input process. After the verification passes, retain the complete velocity and pressure distribution data and prepare to input the pollutant transport mechanism sub-model.

[0192] (7) Input the oil flow velocity distribution data as the initial velocity boundary condition into the pollutant transport mechanism sub-model. Through the flow velocity distribution data interface of the data interaction channel, the oil flow velocity vector data of each grid node that has passed the verification is completely transmitted to the pollutant transport mechanism sub-model. After receiving the data, the sub-model automatically maps the flow velocity distribution data to its own fluid domain grid nodes. The mapping process strictly follows the principle of one-to-one correspondence between grid node numbers to ensure that the flow velocity data of each node is accurately matched without misalignment or omission. The initial velocity boundary condition is used to define the flow state of the oil at the initial moment of the simulation, which directly determines the initial convective transport direction and velocity of the pollutants. After input, the sub-model completes the initial flow field velocity field initialization settings.

[0193] (8) Input the oil pressure distribution data as the initial pressure boundary condition into the pollutant transport mechanism sub-model. Through the pressure distribution data interface of the data interaction channel, transmit the oil pressure scalar data of each grid node that has passed the verification to the pollutant transport mechanism sub-model. After receiving the data, the pollutant transport mechanism sub-model automatically maps the pressure distribution data to its own fluid domain grid nodes. The mapping process strictly follows the principle of one-to-one correspondence between grid node numbers to ensure that the pressure data of each node is accurately matched without misalignment or omission. The initial pressure boundary condition is used to define the pressure field distribution of the oil at the initial moment of the simulation, which directly affects the compression state and transport path of pollutants in the high-pressure region. After input, the sub-model completes the initial flow field pressure field initialization setting.

[0194] (9) After receiving the initial boundary conditions of velocity distribution and pressure distribution, the pollutant transport mechanism sub-model automatically verifies the physical consistency between the initial velocity field and the initial pressure field. For example, if the velocity in the high-pressure area is reasonable and the pressure gradient matches the velocity direction, the verification fails and a boundary condition conflict prompt is returned, and the simulation is terminated. After the verification passes, the pollutant transport mechanism sub-model completes the assignment of physical property parameters and the input of initial boundary conditions. All variables inside the model are initialized and enter the ready state.

[0195] S2.10: After receiving the physical property parameters and the initial boundary conditions, the pollutant transport mechanism sub-model calculates the mass concentration of pollutants at each node of the hydraulic system over time based on the convection-diffusion equation, and obtains the concentration time series of each node.

[0196] Furthermore, the specific steps of S2.10 include:

[0197] (1) Confirm that the pollutant transport mechanism sub-model has been initialized;

[0198] (2) Clarify the application form of the convection-diffusion equation in the hydraulic system scenario. The convection-diffusion equation is used to describe the mass concentration change law with time caused by the convection effect generated by the oil flow, the diffusion effect caused by the concentration difference, and the source and sink effect caused by deposition or reaction during the oil flow process. The equation contains five core parts: time change term, convection term, diffusion term, source term, and sink term. The time change term reflects the rate of concentration change with time, the convection term reflects the transport effect of oil flow carrying pollutants, the diffusion term reflects the diffusion effect of pollutants caused by the concentration gradient, the source term reflects the concentration replenishment of continuous pollution invasion, and the sink term reflects the deposition loss of pollutants on the wall or filter element. It is fully adapted to the physical process of pollutant transport in high-pressure, high-speed, and high-viscosity oil in hydraulic systems.

[0199] (3) Set the basic parameters for solving the convection-diffusion equation. First, determine the physical time range of the solution, set it to two seconds, to cover the complete process of pollutants from initial intrusion to the completion of initial diffusion. Second, set the time step, fixed at 0.001 seconds, to ensure that the transient changes in pollutant concentration can be captured and the flow field and pollutant transport process can be synchronized in time. Then, set the spatial discretization parameters, reuse the hydraulic system fluid domain grid, keep the total number of grid nodes between 1.5 million and 2 million, the grid size of straight pipe sections is 2 mm, and the grid size of complex areas is refined to 0.5 mm to ensure that the spatial resolution of concentration calculation meets the accuracy requirements. Finally, set the initial concentration of pollutants, set according to the type of pollution. The initial concentration of water mist intrusion is set to 10 mg / L, the initial concentration of dust particle intrusion is set to 15 mg / L, the initial concentration of fuel oil mixing is set to 8 mg / L, and the initial concentration of bubble entrainment is set to 5 mg / L. The initial concentration is only applied to the pollution intrusion source node, and the initial concentration of all other nodes is set to 0.

[0200] (4) Calculate the key coefficients of the convection-diffusion equation, including the convection velocity coefficient, molecular diffusion coefficient, turbulent diffusion coefficient, deposition loss coefficient, and source term intensity coefficient. The convection velocity coefficient is directly adopted from the oil flow velocity distribution data in the initial boundary conditions, and the convection velocity of each grid node is consistent with the oil flow velocity. The molecular diffusion coefficient is fixed according to the pollutant type, and is set to 0.00001m for water mist intrusion. 2 / s, dust particle intrusion class set to 0.000005m 2 / s, fuel mixture type set to 0.00002m 2 / s, bubble entrainment class set to 0.000008m 2 / s; The turbulent diffusion coefficient is calculated based on the flow field parameters, with the turbulence intensity fixed at 5% and the turbulence viscosity ratio fixed at 10. It is calculated by combining the oil flow velocity and the grid size. The turbulent diffusion coefficient is larger in the high-pressure pipeline area and smaller in low-speed areas such as the oil tank; The deposition loss coefficient is calculated based on the particle shape coefficient and particle density in the physical property parameters. The smaller the particle shape coefficient and the larger the density, the larger the deposition loss coefficient. The deposition loss coefficient for dust particle intrusion is set to 0.001 / s, and for other pollution types, it is set to 0.0001 / s; The source term intensity coefficient is calculated based on the initial concentration and the continuous release rate of pollution. All pollution types are uniformly set to 2 mg / (L·s) to simulate continuous pollution intrusion;

[0201] (5) The convection-diffusion equation is discretized. The spatial discretization of the equation is achieved by the finite volume method. During the discretization process, the convection term is discretized using the second-order upwind scheme, the diffusion term is discretized using the central difference scheme, and the time-varying term is discretized using the first-order implicit scheme to ensure numerical stability and calculation accuracy. The second-order upwind scheme uses data from two upstream nodes to calculate the convection contribution, avoiding numerical oscillations during high-speed flow. The central difference scheme uses data from adjacent nodes to calculate the diffusion contribution, ensuring that the diffusion term calculation is smooth and accurate. The first-order implicit scheme is unconditionally stable and adaptable to long-term transient calculations. The finite volume method and the first-order implicit scheme discretization are existing technologies in this field and are not the inventive solutions of this application. They will not be elaborated here.

[0202] (6) The pollutant concentration of each grid node is taken as an unknown quantity. Based on the discretized convection-diffusion equation, a corresponding linear equation is established for each grid node. The equation coefficients are determined by the discretization results in (5) and the key coefficients calculated in (4). The number of unknown quantities in the equation set is consistent with the total number of grid nodes. The right-hand side of the equation includes the initial concentration, source term, sink term and the concentration contribution of adjacent nodes, ensuring that the equation set fully reflects the coupling relationship between the concentrations of each node.

[0203] (7) Select an iterative solver, set the upper limit of the number of iterations to 300, set the convergence residual threshold to 0.001, update the concentration value of each node in turn until the concentration residual of all nodes is less than the convergence threshold, determine that the current time step has converged. If the upper limit of the number of iterations has been reached but convergence has not been achieved, the relaxation factor is automatically reduced to 0.5 and the calculation is re-iterated to ensure the stability of the solution.

[0204] (8) Starting from the initial time, i.e., zero seconds, the process is advanced step by step according to a time step of 0.001 seconds. In each time step, the flow field data is updated first, and the velocity distribution and pressure distribution output by the oil flow mechanism sub-model at the current time are obtained synchronously. The convection velocity coefficient and turbulent diffusion coefficient are updated. Then, the source term and sink term are updated. The source term intensity is updated according to the continuous release rate of pollutants, and the sink term loss is updated according to the wall deposition conditions. Then, the discrete linear equation system is solved to obtain the pollutant mass concentration values ​​of all grid nodes at the current time step. Finally, the concentration values ​​of all nodes are recorded to form a snapshot of the node concentration at the current time step.

[0205] (9) From the initial moment to the preset two-second physical time range, record the pollutant mass concentration value corresponding to each grid node step by step. Each node corresponds to an independent concentration time series. The length of the concentration time series is consistent with the total number of time steps, i.e. 2000 time points. Each data point in the concentration time series contains a timestamp and the corresponding node concentration value. The timestamps are arranged sequentially at 0.001-second intervals. The unit of the concentration value is uniformly mg / L. The sequence is strictly arranged in chronological order. The pollutant concentration of each node is fully recorded from the initial zero value or the initial set value.

[0206] S2.11: For each node, mark the time when the mass concentration first reaches the preset detection threshold in the corresponding concentration time series as the pollution arrival time of the node. Sort the corresponding nodes in order of pollution arrival time from early to late, and use the path of pollutant propagation between adjacent nodes as edges to generate a tree-structured pollution diffusion path sequence.

[0207] Furthermore, the specific steps of S2.11 include:

[0208] (1) Unify the node numbering and concentration time series format. All grid nodes in the hydraulic system fluid domain have a unique fixed number. The numbering is arranged continuously in the order of pipeline direction, component position, and flow channel area, without repetition or skipping. Each node corresponds to an independent concentration time series. The sequence length is uniformly two thousand time points. Each time point corresponds to a physical duration of 0.001 seconds. The concentration value unit is uniformly mg / L. The timestamps are arranged continuously increasing from the initial time of zero seconds to ensure that the time axes of all nodes are completely aligned.

[0209] (2) Set the detection threshold to 0.5 mg / L, which is the lowest concentration standard at which contamination can be detected in industrial hydraulic systems;

[0210] (3) Process each node in the order of node number. For the concentration time series corresponding to each node, start from the first time point, i.e. the initial time zero second, and read the concentration value from front to back along the timestamp. Compare it with the preset detection threshold of 0.5 mg / L in turn. Strictly implement the first-to-meet principle. Once the first concentration value greater than or equal to the detection threshold is read, immediately record the physical time corresponding to that time point. This time is the pollution arrival time of that node. If the concentration value is always lower than the detection threshold in the entire time series, mark the node as an uncontaminated node, do not assign a pollution arrival time, and do not include it in the sorting and path generation range.

[0211] (4) After traversing, summarize all nodes marked as polluted and generate a polluted node set. Each record in the polluted node set contains three pieces of information: the node's unique number, the node's physical location, and the time when the pollution arrived.

[0212] (5) Using the arrival time of pollution as the sole sorting criterion, all nodes in the pollution node set are sorted in ascending order, with the earlier the arrival time being sorted first and the later the arrival time being sorted last. If multiple nodes have the same arrival time of pollution, they are sorted in ascending order according to their node numbers. After sorting, an ordered list of nodes is formed, and the order of the list strictly corresponds to the order of pollution propagation from first to last.

[0213] (6) Based on the actual pipeline connections, component interfaces, and flow channel connections of the hydraulic system geometric model, establish a physical adjacency table between nodes. The adjacency table clearly records the upstream, downstream, and lateral connected nodes directly connected to each node. The adjacency relationship is strictly based on the actual physical connectivity and only includes nodes that are directly adjacent and can be directly reached by oil flow. It does not include indirect connected nodes that cross pipelines or components.

[0214] (7) Start from the first node in the ordered node list, which is the earliest contaminated node. This node is the root node of the contamination diffusion tree. Iterate through each subsequent node in turn, find all its upstream neighboring nodes, select the node with the earliest contamination arrival time as the parent node, and the current node as the child node. The physical connection path from the parent node to the child node is the diffusion edge. Each child node is only associated with one parent node to ensure that the path is unique and the structure is tree-like, and to avoid loops or cross connections.

[0215] (8) Starting from the root node, all polluted nodes are connected layer by layer according to the edge connection relationship from the parent node to the child node to form a tree structure. Each layer of the tree corresponds to the order of arrival time of the pollution. Nodes in the same layer arrive at similar times, and nodes in the lower layer arrive at later times than nodes in the upper layer. Each node retains a unique number, physical location, and pollution arrival time in the tree. Each edge is labeled with the corresponding physical connection path. Finally, a complete, unique, loop-free, and hierarchical tree-like pollution diffusion path sequence is generated, which fully reflects the order and propagation path of pollution from the source to each node.

[0216] (9) Verify the validity of the pollution diffusion path sequence. The verification includes three items: first, check whether the root node is the node with the earliest arrival time of pollution; second, check whether the arrival time of all child nodes is later than the corresponding parent node; and third, check whether the path structure is strictly tree-like, without loops or intersections. After all verifications pass, the path sequence is solidified for damage deviation calculation. If the verification fails, the node sorting or adjacency relationship construction steps are backtracked and the sequence is regenerated.

[0217] The damage priority deviation coefficient for calculating the actual pollutant diffusion state and pollution diffusion path sequence includes:

[0218] S2.12: Extract the actual pollutant concentration value, actual pollutant particle size distribution and actual pollutant accumulation rate of each hydraulic component in real time from the twin output value output by the digital twin mirror of the hydraulic system, and combine them into an actual pollutant diffusion state vector;

[0219] S2.13: Extract the predicted pollutant concentration value, predicted pollutant particle size distribution and predicted arrival time carried by each node from the pollution diffusion path sequence, and combine them into a predicted diffusion state vector;

[0220] S2.14: Calculate the state deviation between the actual pollutant diffusion state vector and the predicted diffusion state vector based on Mahalanobis distance; each component of the state deviation corresponds to the deviation value of a hydraulic component. The Mahalanobis distance calculation formula is prior art in this field and is not an inventive solution of this application, so it will not be elaborated here.

[0221] S2.15: Query the damage sensitivity coefficient of each hydraulic component from the preset component damage sensitivity coefficient table; the damage sensitivity coefficient is obtained by weighted summation of the hydraulic component's fit clearance, surface hardness, and working pressure level;

[0222] Furthermore, the damage sensitivity coefficient is used to quantify the sensitivity of hydraulic components to wear, blockage, and corrosion caused by contaminants. The larger the value, the more easily the component is damaged by contamination, and the smaller the value, the stronger the component's resistance to contamination. This coefficient is determined by three core indicators: fit clearance, surface hardness, and working pressure level. It is calculated using a weighted summation algorithm. The algorithm weights are determined through statistical analysis of damage test data of industrial hydraulic components. Specifically, the weight of the fit clearance indicator is set to 0.4, the weight of the surface hardness indicator is set to 0.3, and the weight of the working pressure level indicator is set to 0.3.

[0223] Furthermore, the clearance refers to the gap size between moving parts of hydraulic components. The classification standard is divided into five levels: Level 1 is the smallest clearance, with a clearance value of less than or equal to 5 micrometers and a corresponding index value of 0.1; Level 2 is the smallest clearance, with a clearance value of more than 5 micrometers and less than or equal to 10 micrometers and a corresponding index value of 0.3; Level 3 is the medium clearance, with a clearance value of more than 10 micrometers and less than or equal to 20 micrometers and a corresponding index value of 0.5; Level 4 is the largest clearance, with a clearance value of more than 20 micrometers and less than or equal to 50 micrometers and a corresponding index value of 0.7; and Level 5 is the largest clearance, with a clearance value of more than 50 micrometers and a corresponding index value of 0.9, which is suitable for the clearance characteristics of various hydraulic components.

[0224] Furthermore, surface hardness refers to the material hardness of the key friction surfaces of hydraulic components. The grading standard is divided into five levels: Level 1 is ultra-high hardness, with a hardness value greater than or equal to Rockwell hardness 60, corresponding to a value of 0.1; Level 2 is high hardness, with a hardness value greater than or equal to Rockwell hardness 50 and less than 60, corresponding to a value of 0.3; Level 3 is medium hardness, with a hardness value greater than or equal to Rockwell hardness 40 and less than 50, corresponding to a value of 0.5; Level 4 is low hardness, with a hardness value greater than or equal to Rockwell hardness 30 and less than 40, corresponding to a value of 0.7; and Level 5 is extremely low hardness, with a hardness value less than Rockwell hardness 30, corresponding to a value of 0.9. The grading standard and values ​​are fixed throughout, adapting to the hardness range of commonly used metal materials in hydraulic components.

[0225] Furthermore, the working pressure rating refers to the rated working pressure of the hydraulic component. The classification standard is divided into five levels: Level 1 is low pressure, with a pressure value of less than or equal to 10 MPa and a corresponding index value of 0.1; Level 2 is medium-low pressure, with a pressure value of greater than 10 MPa and less than or equal to 20 MPa and a corresponding index value of 0.3; Level 3 is medium pressure, with a pressure value of greater than 20 MPa and less than or equal to 30 MPa and a corresponding index value of 0.5; Level 4 is medium-high pressure, with a pressure value of greater than 30 MPa and less than or equal to 40 MPa and a corresponding index value of 0.7; and Level 5 is high pressure, with a pressure value of greater than 40 MPa and a corresponding index value of 0.9. The classification standard and values ​​are fixed throughout, covering the common pressure rating range of industrial hydraulic systems.

[0226] Furthermore, the component damage sensitivity coefficient table uses hydraulic components as the basic unit and includes six fixed fields: component name, component number, mating clearance index value, surface hardness index value, working pressure level index value, and damage sensitivity coefficient. The table data is based on the parameters of commonly used hydraulic components in engineering machinery. Before entering the data, the measured values ​​of the mating clearance, surface hardness, and rated working pressure of each component are checked one by one, and the corresponding index values ​​are determined according to the grading standards to ensure that the three index values ​​of each component are accurate.

[0227] Furthermore, when calculating the damage sensitivity coefficients of various hydraulic components, a weighted summation algorithm is used. Specifically, the mating clearance index is multiplied by 0.4, the surface hardness index is multiplied by 0.3, and the working pressure level index is multiplied by 0.3. The three products are then added together to obtain the damage sensitivity coefficient of the corresponding component. The coefficient value ranges from 0.1 to 0.9. The entire calculation process is manually checked to ensure there are no calculation errors.

[0228] Furthermore, damage sensitivity coefficients for common hydraulic components were entered and solidified. For the hydraulic pump component, with a clearance of 8 micrometers, a surface hardness of Rockwell hardness 55, and a working pressure of 31 MPa, the corresponding damage sensitivity coefficient was calculated as 0.4 multiplied by 0.3 plus 0.3 multiplied by 0.3 plus 0.3 multiplied by 0.7, resulting in 0.42. For the servo valve component, with a clearance of 3 micrometers, a surface hardness of Rockwell hardness 58, and a working pressure of 35 MPa, the corresponding damage sensitivity coefficient was calculated as 0.4 multiplied by 0.7. 0.1 plus 0.3 multiplied by 0.3 plus 0.3 multiplied by 0.7 equals 0.34; the multi-way valve component has a mating clearance of 15 micrometers, a surface hardness of Rockwell hardness 48 degrees, and a working pressure of 25 MPa, corresponding to a damage sensitivity coefficient calculated as 0.4 multiplied by 0.5 plus 0.3 multiplied by 0.5 plus 0.3 multiplied by 0.5, resulting in 0.5; the hydraulic cylinder component has a mating clearance of 30 micrometers, a surface hardness of Rockwell hardness 42 degrees, and a working pressure of 20 MPa, corresponding to a damage sensitivity coefficient calculated as... The calculation is 0.4 multiplied by 0.7 plus 0.3 multiplied by 0.5 plus 0.3 multiplied by 0.3, resulting in 0.52; the hydraulic motor component has a mating clearance of 25 micrometers, a surface hardness of Rockwell hardness of 45 degrees, and a working pressure of 28 MPa, corresponding to a damage sensitivity coefficient calculated as 0.4 multiplied by 0.7 plus 0.3 multiplied by 0.5 plus 0.3 multiplied by 0.5, resulting in 0.58; the filter element has a mating clearance of 100 micrometers, a surface hardness of Rockwell hardness of 50 degrees, and a working pressure of 10 MPa. The damage sensitivity coefficient for component a is calculated as 0.4 multiplied by 0.9 plus 0.3 multiplied by 0.3 plus 0.3 multiplied by 0.1, resulting in 0.48. The fit clearance of the fuel tank components is 200 micrometers, the surface hardness is Rockwell hardness 35 degrees, and the working pressure is zero MPa. The corresponding damage sensitivity coefficient is calculated as 0.4 multiplied by 0.9 plus 0.3 multiplied by 0.7 plus 0.3 multiplied by 0.1, resulting in 0.6. After all component coefficients are entered into the table, the table data is fixed and read-only access is set to prevent accidental modification.

[0229] S2.16: Multiply each component of the state deviation by the damage sensitivity coefficient of the corresponding hydraulic component to obtain a weighted deviation component. Then, accumulate all weighted deviation components according to the node order in the pollution diffusion path sequence and perform min-max normalization to output the damage priority deviation coefficient. The damage priority deviation coefficient is a quantitative index in the normalized range of 0 to 1, which is used to comprehensively characterize the degree of deviation between the actual pollutant diffusion conditions of the hydraulic system and the digital twin predicted diffusion path. The larger the value, the higher the deviation between the actual pollution causing component damage and the simulation prediction result, and the more prominent the potential damage risk.

[0230] S3: Compare the damage priority deviation coefficient with the preset grading threshold to determine the risk level of external pollution intrusion into the hydraulic system. For hydraulic systems with external pollution intrusion, match the corresponding intrusion path according to the particle size of the pollutants to trace the source of pollution.

[0231] By comparing the damage priority deviation coefficient with a preset grading threshold, the risk level of external contamination intrusion into the hydraulic system is determined, including:

[0232] S3.1: Obtain the preset first grading threshold and second grading threshold, and at the same time obtain the damage priority deviation coefficient; the first grading threshold is less than the second grading threshold. In this embodiment, the first grading threshold and the second grading threshold are set to 0.3 and 0.7 respectively, and the damage priority deviation coefficient is 0~1.

[0233] S3.2: When the damage priority deviation coefficient is less than the first classification threshold, it is determined to be a low-risk level and a corresponding low-risk level code is generated. In this embodiment, the low-risk level code is set to 01.

[0234] S3.3: When the damage priority deviation coefficient is greater than or equal to the first grading threshold and less than the second grading threshold, it is determined to be of medium risk level, and a corresponding medium risk level code is generated. In this embodiment, the medium risk level code is set to 10.

[0235] S3.4: When the damage priority deviation coefficient is greater than or equal to the second classification threshold, it is determined to be a high-risk level and a corresponding high-risk level code is generated. In this embodiment, the high-risk level code is set to 11.

[0236] S3.5: Output the generated low-risk level code, medium-risk level code, and high-risk level code as the results of the determination of the risk level of external pollution intrusion.

[0237] For hydraulic systems susceptible to external contamination, the method of matching corresponding intrusion paths according to contaminant particle size levels to trace the source of contamination includes:

[0238] S3.6: Obtain the risk level of external pollution intrusion. When the risk level of external pollution intrusion is medium risk or high risk, execute:

[0239] S3.7: Acquire the time-series signal of oil particle size, and perform hierarchical analysis of the time-series signal of oil particle size according to the preset particle size interval boundaries to obtain sub-concentration sequences corresponding to multiple particle size levels; the particle size interval boundaries include 4 micrometers, 6 micrometers, 14 micrometers and 21 micrometers;

[0240] Furthermore, the time-series signal of oil particle size is subjected to hierarchical analysis to obtain sub-concentration sequences corresponding to multiple particle size levels, including:

[0241] (1) The particle size channel definition of the time-series signal of oil particle size is analyzed. The original particle size time-series signal contains multiple particle size detection channels. Each channel corresponds to a particle size range. The channel particle size division accuracy is 1 micrometer, covering the particle size range from 0 micrometer to 50 micrometer. The channel number is arranged in order of particle size from small to large. The signal data at each sampling time contains the concentration value of all particle size channels.

[0242] (2) The four preset boundary values ​​of 4 micrometers, 6 micrometers, 14 micrometers and 21 micrometers are mapped to the corresponding particle size channels of the original particle size time series signal respectively. The 4 micrometer boundary corresponds to the channel with a particle size of 4 micrometers in the signal, the 6 micrometer boundary corresponds to the channel with a particle size of 6 micrometers, the 14 micrometer boundary corresponds to the channel with a particle size of 14 micrometers, and the 21 micrometer boundary corresponds to the channel with a particle size of 21 micrometers. The mapping process is one-to-one.

[0243] (3) Process the data in five levels according to the boundaries of the four particle size ranges: The first level is less than 4 micrometers. Extract the concentration data of all channels with a particle size of less than 4 micrometers from the original signal, sum them according to the sampling time, and obtain the total concentration at each time of the level to form a sub-concentration sequence of less than 4 micrometers; The second level is from 4 micrometers to 6 micrometers. Extract the concentration data of all channels with a particle size greater than or equal to 4 micrometers and less than 6 micrometers from the original signal, sum them according to the sampling time, and form a sub-concentration sequence of 4 micrometers to 6 micrometers; The third level is from 6 micrometers to 14 micrometers. Extract the concentration data of all channels with a particle size of greater than or equal to 4 micrometers and less than 6 micrometers from the original signal, sum them according to the sampling time, and form a sub-concentration sequence of 4 micrometers to 6 micrometers; The concentration data of all channels with a particle size greater than or equal to 6 micrometers and less than 14 micrometers are summed according to the sampling time to form a sub-concentration sequence from 6 micrometers to 14 micrometers; the fourth level is from 14 micrometers to 21 micrometers, extracting the concentration data of all channels with a particle size greater than or equal to 14 micrometers and less than 21 micrometers from the original signal, summing according to the sampling time to form a sub-concentration sequence from 14 micrometers to 21 micrometers; the fifth level is for particles larger than 21 micrometers, extracting the concentration data of all channels with a particle size larger than 21 micrometers from the original signal, summing according to the sampling time to form a sub-concentration sequence for particles larger than 21 micrometers.

[0244] (4) All five sub-concentration sequences maintain the same sampling frequency as the original particle size time series signal, i.e., one data point per second, the timestamp is completely aligned with the original signal, the concentration unit is uniformly mg / L, the sequence length is consistent with the original signal, each sequence contains only the total concentration data of the corresponding particle size level, and the data order is strictly arranged according to the sampling time.

[0245] (5) Check each of the five sub-concentration sequences to confirm that the sequences are complete and without missing data, the timestamps are aligned, the concentration values ​​are non-negative and within a reasonable range. The reasonable range is set to 0~100mg / L. If there is abnormal data, backtrack the stratification analysis steps to correct it. After the verification is passed, the sub-concentration sequences of the five particle size levels are the stratification analysis results.

[0246] S3.8: Perform mutation point detection on the sub-concentration sequences corresponding to each particle size level, calculate the cumulative sum and statistics of each sub-concentration sequence, mark the time point when the cumulative sum and statistics exceed the preset mutation judgment threshold as the mutation start time of the corresponding particle size level, and arrange the mutation start times of all particle size levels in ascending order of particle size to construct a particle size-time mutation matrix.

[0247] Furthermore, the specific steps in S3.8 include:

[0248] (1) Obtain the sub-concentration sequences of each particle size level;

[0249] (2) Clarify the algorithm principle and parameter settings of the cumulative sum statistic. The cumulative sum statistic is used to capture the deviation of the mean in the time series. By accumulating the deviation value point by point, the small mutation is amplified. The algorithm parameters are set as follows: the benchmark mean of the calculation window is the average concentration of the first 100 time points of the sequence. This value is used as the benchmark for normal fluctuation of the sequence. The deviation is calculated by subtracting the benchmark mean from the concentration value at the current time. The initial value of the cumulative sum statistic is set to zero. The current deviation value is accumulated in each iteration.

[0250] (3) Calculate the average concentration of the first 100 time points for each of the five sub-concentration sequences in turn, and use it as the normal fluctuation baseline average for each sequence. The baseline average for the less than 4 micrometer level is set to 0.2 mg / L, the baseline average for the 4 to 6 micrometer level is set to 0.3 mg / L, the baseline average for the 6 to 14 micrometer level is set to 0.4 mg / L, the baseline average for the 14 to 21 micrometer level is set to 0.25 mg / L, and the baseline average for the greater than 21 micrometer level is set to 0.1 mg / L. The baseline average is set independently according to the normal pollution level of each particle size.

[0251] (4) The mutation determination threshold is set to 1.2 mg / L, which is applicable to all particle size level sub-concentration sequences;

[0252] (5) Process the five sub-concentration sequences in order of increasing particle size. For each sequence, starting from the first time point, calculate the deviation value one by one. The deviation value is equal to the current concentration value minus the baseline mean of the sequence. Then, add the deviation value to the cumulative sum statistic. The initial cumulative sum statistic is zero. Update the cumulative sum statistic once for each time point processed. Generate the cumulative sum statistic value corresponding to each time point in sequence. The calculation is performed in the forward order of time throughout the process without backtracking.

[0253] (6) For the cumulative sum and statistics sequence of each particle size level, starting from the first time point, the cumulative sum and statistics value of each time point is compared with the preset mutation judgment threshold of 1.2 mg / L. The first exceedance principle is strictly implemented. Once the first time point with a cumulative sum and statistics value greater than 1.2 mg / L is found, the time point is immediately marked as the mutation start time of the particle size level. Subsequent time points will no longer participate in the mutation start time judgment of the level. If the cumulative sum and statistics of the entire sequence never exceed the threshold, the level is marked as having no mutation start time.

[0254] (7) After completing the mutation detection at five levels, summarize the mutation initiation time information at each level, including the particle size level name, mutation initiation time or no mutation marker;

[0255] (8) Arrange the mutation start time of each level in the order of less than 4 micrometers, 4 micrometers to 6 micrometers, 6 micrometers to 14 micrometers, 14 micrometers to 21 micrometers, and greater than 21 micrometers. Mark the level without mutation as null. The arrangement order strictly follows the order of particle size from small to large, without disrupting the level order, and ensuring that the matrix row order is fixed.

[0256] (9) Construct a particle size time mutation matrix. The matrix is ​​arranged with particle size level as the row and timestamp as the column. The matrix element is a marker indicating whether a mutation has occurred at the corresponding particle size level at that time point. If there is a mutation start time, the corresponding timestamp is marked. If there is no mutation, an empty value is marked. The number of rows in the matrix is ​​fixed at 5, corresponding to the five particle size levels. The number of columns is consistent with the number of time points of the sub-concentration sequence. The matrix format is standardized, and the rows and columns are aligned to fully present the temporal distribution characteristics of mutation occurrence at each particle size level.

[0257] S3.9: Perform cosine similarity matching between the particle size-time mutation matrix and each invasion path feature template in the preset invasion path feature library, and select the invasion path corresponding to the invasion path feature template with the largest cosine similarity as the pollution source output; the invasion path feature library stores standard matrices corresponding to the respirator filter failure path, piston rod seal wear path, pipeline joint loosening path, and new oil filling pollution path, wherein the cosine similarity calculation process is the prior art in this field and is not an inventive solution of this application, and will not be described in detail here.

[0258] Furthermore, the intrusion path feature library is a structured repository that stores standard matrices corresponding to four common contamination intrusion paths: the standard matrix for respirator filter failure path, the standard matrix for piston rod seal wear path, the standard matrix for pipeline joint loosening path, and the standard matrix for new oil filling contamination path. Each standard matrix has the same format as the particle size time mutation matrix, which is 5 rows and 5 columns. The 5 rows correspond to the five particle size levels, and the number of columns is consistent with the number of time points in the sub-concentration sequence. The matrix elements are the standard mutation start times of the corresponding particle size level. The standard times are determined statistically based on a large amount of historical contamination intrusion test data.

[0259] Furthermore, the specific parameter settings for the four standard matrices are as follows: For the respirator filter failure path standard matrix, the start times of abrupt changes at each particle size level are, in order: less than 4 micrometers: 10 seconds; 4 micrometers to 6 micrometers: 12 seconds; 6 micrometers to 14 micrometers: 15 seconds; 14 micrometers to 21 micrometers: 18 seconds; greater than 21 micrometers: 20 seconds. For the piston rod seal wear path standard matrix, the start times of abrupt changes at each particle size level are, in order: less than 4 micrometers: 8 seconds; 4 micrometers to 6 micrometers: 10 seconds; 6 micrometers to 14 micrometers: 13 seconds; 14 micrometers to 21 micrometers: 16 seconds; greater than 21 micrometers: 19 seconds. For loose pipe joints… The standard matrix for the dynamic path shows the start times of abrupt changes at each particle size level as follows: less than 4 micrometers: 15 seconds; 4 to 6 micrometers: 17 seconds; 6 to 14 micrometers: 20 seconds; 14 to 21 micrometers: 23 seconds; and greater than 21 micrometers: 25 seconds. The standard matrix for the contamination path of new oil refueling shows the start times of abrupt changes at each particle size level as follows: less than 4 micrometers: 5 seconds; 4 to 6 micrometers: 7 seconds; 6 to 14 micrometers: 9 seconds; 14 to 21 micrometers: 11 seconds; and greater than 21 micrometers: 13 seconds. All standard abrupt change start times are accurate to the second, consistent with the timestamp accuracy of the particle size time abrupt change matrix.

[0260] Furthermore, the execution process of cosine similarity template-by-template matching includes: following the order of the respirator filter failure path, piston rod seal wear path, pipeline joint loosening path, and new oil filling contamination path, the particle size time mutation matrix is ​​matched with four standard matrices in sequence. Each matching performs the same calculation process: first, the two matrices are converted into one-dimensional vectors, then the dot product of the two vectors is calculated, then the magnitudes of the two vectors are calculated separately, and finally the dot product is divided by the product of the two magnitudes to obtain the corresponding cosine similarity value. Each matching process is performed independently. After the calculation is completed, the similarity values ​​corresponding to the four standard templates are recorded respectively.

[0261] Furthermore, when calculating the dot product, the elements at corresponding positions of the two vectors are multiplied one by one, and then all the product results are summed to obtain the dot product value. When calculating the modulus, the squares of each element of the vectors are summed, and then the square root of the sum is taken to obtain the modulus value. When calculating the cosine similarity, the dot product value is divided by the product of the two modulus values ​​to obtain the final similarity. For example, if the dot product of the particle size time mutation matrix vector and the standard matrix vector of the respirator filter failure path is 80, and their moduli are 10 and 8.5 respectively, then the similarity is 80 divided by 80.5, and the result is approximately 0.9938.

[0262] Furthermore, the cosine similarity values ​​corresponding to the four standard templates are compared pairwise to find the standard template with the highest similarity value. If two or more similarity values ​​are the same and both are the maximum, the intrusion path template with the highest ranking is selected by default, i.e., the breather filter failure path is prioritized, followed by the piston rod seal wear path, the pipeline joint loosening path, and the new oil filling contamination path. Based on the standard template corresponding to the maximum similarity, the corresponding intrusion path is determined, which is the source of hydraulic system contamination. The output includes the name of the contamination source, the corresponding intrusion path feature template, and the cosine similarity value. The output format is a standardized text record to ensure that the information is complete, clear, and unambiguous. For example, if the template corresponding to the maximum similarity is the standard matrix of piston rod seal wear path, then the output contamination source is piston rod seal wear path, along with the corresponding similarity value.

[0263] S4: Based on the pollution source, perform oil pipeline zone isolation verification, generate pollution disposal instruction summary, embed the verification timestamp into the disposal instruction summary after binary encoding, generate early warning disposal data package and upload it to the industrial blockchain for evidence storage.

[0264] The process of verifying the isolation of oil pipelines based on the source of pollution and generating a pollution disposal instruction summary includes:

[0265] S4.1: Based on the pollution source, query the corresponding oil circuit partition boundary to be isolated from the preset isolation strategy library; the oil circuit partition boundary to be isolated includes the isolation valve tag number, the isolation pipeline section identifier, and the list of associated components;

[0266] Furthermore, the isolation strategy library is a pre-defined structured repository specifically storing isolation strategies corresponding to four types of contamination sources and the boundary information of the oil circuit zones to be isolated. Each contamination source corresponds to a unique isolation scheme, ensuring accurate matching during queries and avoiding confusion. The structure of the isolation strategy library is categorized by contamination source, with each category containing three fixed information modules: isolation valve tag number, isolation pipeline section identifier, and a list of associated components. The information in these three modules is fully correlated and together constitutes the boundary of the oil circuit zones to be isolated. All information in the library is based on the hydraulic system oil circuit layout and contamination isolation test data.

[0267] Furthermore, the process of constructing and solidifying the isolation strategy library is as follows: Isolation information is entered one by one according to the four types of pollution sources, ensuring that the isolation valve tag number, isolation pipeline section identifier, and related component list for each pollution source are complete and accurate. Specific entry content is as follows:

[0268] For the isolation information corresponding to the failure path of the respirator filter, the isolation valve tag numbers are set to 1001 and 1002. The two isolation valves are installed at both ends of the connection pipeline between the respirator and the fuel tank, respectively, to completely cut off the connection between the respirator and the fuel tank. The isolation pipeline section is marked as pipeline 101 to pipeline 103, corresponding to the three connection pipelines between the respirator and the fuel tank, which is the pipeline range that needs to be isolated. The list of associated components includes the respirator filter, fuel tank filling port, and pipeline joints, all of which are directly related to the source of pollution and need to be isolated and monitored simultaneously.

[0269] For the isolation information corresponding to the piston rod seal wear path, the isolation valve tag numbers are set to 2001 and 2002, which are installed on the hydraulic cylinder inlet and return oil lines respectively, to cut off the connection between the hydraulic cylinder and the main oil circuit of the system; the isolation pipeline section identifiers are set to pipelines 201 to 204, corresponding to the four connecting pipelines of the hydraulic cylinder inlet and return oil; the list of associated components includes the hydraulic cylinder, piston rod seal, inlet oil filter, and return oil filter to ensure that no key associated components are missed during the isolation process.

[0270] For the isolation information corresponding to the loose pipe joint path, the isolation valve tag numbers are set to 3001, 3002 and 3003. The three isolation valves are installed on the pipes and branch pipes on both sides of the loose joint to form a closed-loop isolation. The isolation pipe section identifier is set to pipes 301 to pipes 305, corresponding to 5 associated pipes in the loose joint area. The associated component list includes pipe joints, branch pipe filters and pipe supports, clearly identifying all components related to the loose joint.

[0271] For the isolation information corresponding to the contamination path of new oil filling, the isolation valve tag numbers are set to 4001 and 4002, which are installed at both ends of the connection pipeline between the new oil filling port and the oil tank, respectively; the isolation pipeline section identifier is set to pipeline 401 to pipeline 402, corresponding to the two connection pipelines of new oil filling; the associated component list includes the new oil filling port, filling filter, oil tank inlet, and pipeline joints, ensuring that while cutting off the new oil filling channel, all associated key components are covered.

[0272] Furthermore, the name of the pollution source to be queried is compared one by one with the names of the four categories in the isolation strategy library. The matching principle is that the names must be completely consistent. For example, if the pollution source is the piston rod seal wear path, only the category corresponding to the piston rod seal wear path in the isolation strategy library will be matched, and the other three paths will not be matched to avoid matching deviations.

[0273] Furthermore, the isolation strategy query operation is performed, including: locating the isolation strategy category corresponding to the pollution source according to the exact matching rules; upon entering the category, reading the information from the three modules one by one; first, reading the isolation valve tag number to clarify the specific number of the isolation valve that needs to be closed, ensuring that each valve tag number is clearly traceable; then, reading the isolation pipeline section identifier to clarify the specific range of the pipeline that needs to be isolated, ensuring that no pipeline sections are omitted or redundant; finally, reading the list of associated components to clarify the names of the components that need to be isolated and monitored simultaneously, ensuring that the component list is complete and that no key associated components are missing. During the query process, no original information in the database is modified; only the reading operation is performed.

[0274] Furthermore, the boundary information of the oil circuit to be isolated is organized, including: organizing the queried isolation valve tag numbers, isolation pipeline section identifiers, and related component lists in a fixed order to form a complete record of the boundary of the oil circuit to be isolated. The record format is uniform, first marking the name of the pollution source, and then listing the isolation valve tag numbers, isolation pipeline section identifiers, and related component lists in sequence. The information of each module is organized separately to ensure that it is clear and easy to understand, and to facilitate the execution of isolation operations.

[0275] S4.2: In the digital twin mirror of the hydraulic system, set the opening parameter of the valve element corresponding to the isolation valve position number to zero, simulate the valve closing operation, and based on the simulated closing conditions, rerun the oil flow mechanism sub-model and the contaminant transport mechanism sub-model to calculate the predicted decay curve of oil contaminant concentration in the non-isolation zone.

[0276] Furthermore, the specific steps in S4.2 include:

[0277] (1) Start the hydraulic system digital twin mirror and confirm the initial state. First, start the hydraulic system digital twin mirror and ensure that the mirror has fully loaded the hydraulic system geometric model, oil flow mechanism sub-model, contaminant transport mechanism sub-model and all preset parameters. The mirror is running normally, with no errors or missing data. Confirm that the current working condition of the mirror is completely consistent with the actual hydraulic system working condition. Set the oil temperature to 40℃, the system working pressure to 31MPa, and the oil kinematic viscosity to 30mm. 2 / s, all sub-models are in a ready state;

[0278] (2) Based on the isolation valve position number obtained in the previous query, the corresponding valve element is accurately found in the hydraulic system geometric model of the digital twin mirror by the component number. Each isolation valve position number corresponds to a unique valve element. For example, isolation valve position number 1001 corresponds to the shut-off valve at one end of the pipeline connecting the breather and the oil tank, and isolation valve position number 2001 corresponds to the shut-off valve of the hydraulic cylinder inlet pipeline. During the positioning process, ensure that the position and model of the valve element are consistent with the actual system. At the same time, record the initial opening parameters of the valve element. The initial opening is set to 100%, that is, the valve is fully open.

[0279] (3) Simulate the closing operation. In the component parameter setting interface of the digital twin mirror, find the opening parameter adjustment inlet corresponding to each isolation valve after positioning. Set the valve opening parameter corresponding to all isolation valve positions to zero. The value range of the opening parameter is 0 to 100%. 0 means the valve is completely closed and 100% means the valve is completely open. During the setting process, check the correspondence between the valve position number and the valve component one by one to avoid misoperation of other non-isolation valves. After the setting is completed, save the parameter modification. The mirror will automatically update the state of the valve component to simulate the physical state after the valve is completely closed and cut off the oil flow channel of the corresponding isolation pipeline.

[0280] (4) After the valve opening parameters are set, check the flow status of the pipeline corresponding to the isolation valve in the digital twin mirror. Confirm that after the valve is completely closed, the flow path of the corresponding isolation pipeline is blocked and the oil cannot flow through the pipeline. At the same time, check the pipeline connectivity in the non-isolation area to ensure that the oil flow path in the non-isolation area is not affected and there is no situation of accidentally closing the non-isolation valve or accidentally disconnecting the non-isolation pipeline. At the same time, check the valve opening parameter setting record to confirm that the opening of all isolation valves has been accurately set to zero, without omissions or errors, and ensure that the simulated closing operation meets the preset requirements.

[0281] (5) Based on the simulated operating conditions after the valve is closed, reset the initial parameters of the oil flow mechanism sub-model to ensure that the parameters match the operating conditions after closure. The specific parameter settings are as follows: the oil density is maintained at 870 kg / m³. 3The dynamic viscosity and compressibility coefficient remain unchanged at 0.028 and 0.0001, respectively. The inlet boundary condition is adjusted to the hydraulic pump outlet in the non-isolated zone, with the velocity inlet condition set to 2 m / s and the temperature set to 40℃. The outlet boundary condition is adjusted to the return oil pipeline inlet in the non-isolated zone, with the pressure outlet condition set to 0.2 MPa. The wall boundary and internal coupling boundary parameters remain unchanged, consistent with the previous flow mechanism sub-model settings, to ensure that the sub-model operating conditions are consistent with the actual state after the valve is closed.

[0282] (6) The initial parameters of the pollutant transport mechanism sub-model are adjusted synchronously to adapt to the working conditions after the valve is closed. Among them, the average particle size, particle density and other parameters are used according to the pollutant type corresponding to the pollution source. The initial boundary conditions are updated to the oil flow velocity distribution and pressure distribution in the non-isolated zone after the valve is closed. This distribution is provided by the initial calculation of the oil flow mechanism sub-model. The initial concentration is set to the actual pollutant concentration of each node in the non-isolated zone at the time of valve closure to ensure that the initial concentration is connected with the closing working condition without any disconnection. The deposition loss coefficient and source term intensity coefficient remain unchanged.

[0283] (7) Clarify the calculation parameters for rerunning the oil flow mechanism sub-model and the pollutant transport mechanism sub-model to ensure calculation accuracy and efficiency. The time step is still set to 0.001 seconds, and the total calculation time is set to 30 seconds to cover the complete process of pollutant concentration in the non-isolated area decaying from the current value to the stable value. The solution convergence residual threshold of the oil flow mechanism sub-model is set to 0.0001, the upper limit of the number of iterations is set to 500, the second-order upwind scheme is adopted for the convection term, the central difference scheme is adopted for the diffusion term, the semi-implicit pressure correlation equation algorithm is adopted for the pressure-velocity coupling algorithm, the standard two-equation turbulence model is selected for the turbulence model, the turbulence intensity is set to 5%, and the turbulence viscosity ratio is set to 10. The solution convergence residual threshold of the pollutant transport mechanism sub-model is set to 0.0001, the upper limit of the number of iterations is set to 300, the number of discrete phase particles is set to 100,000, and the tracking time is set to 30 seconds to ensure that the calculation parameters of the sub-model are reasonable and stable.

[0284] (8) First, start the oil flow mechanism sub-model. Based on the reset initial working parameters and calculation parameters, start iterative calculation. After each time step is completed, output the oil velocity distribution and pressure distribution data of each grid node in the non-isolated area at the current time step, and transmit them to the pollutant transport mechanism sub-model in real time. After the oil flow mechanism sub-model converges, start the pollutant transport mechanism sub-model, couple the real-time received flow field data, and start iterative calculation. After each time step is completed, output the pollutant concentration values ​​of each grid node in the non-isolated area at the current time step.

[0285] (9) During the rerun of the two sub-models, the pollutant concentration values ​​of all grid nodes in the non-isolated area are collected at time points with a time step of 0.001 seconds. A complete concentration data is collected once at each time point, and the concentration value and corresponding timestamp of each node are recorded. The collection range is strictly limited to the non-isolated area and does not include node data in the isolated area to avoid interference of isolated area data with the calculation of concentration decay curve. After the collection is completed, the non-isolated area pollutant concentration time series dataset is formed. The dataset is arranged in the order of timestamps, and each time point corresponds to a set of non-isolated area node concentration data.

[0286] (10) Based on the collected time series dataset of pollutant concentration in the non-isolated area, a predicted decay curve is constructed with time as the horizontal axis and the average pollutant concentration in the non-isolated area as the vertical axis. Specifically, the average pollutant concentration of all nodes in the non-isolated area at each time step is calculated. The calculation method is to add the concentration values ​​of all nodes in the non-isolated area and divide by the total number of nodes to obtain the average concentration at the current time step. Then, the average concentration of each time step is matched with the corresponding timestamp, and all data points are connected in chronological order to form a continuous concentration prediction decay curve. The starting point of the curve is the average concentration at the moment the valve is closed, and the ending point is the average concentration at the end of the 30-second calculation period, which fully reflects the decay process of pollutant concentration in the non-isolated area over time after the valve is closed.

[0287] S4.3: Obtain the safety threshold. In this embodiment, the safety threshold is set to 0.3 mg / L. When all concentration values ​​of the predicted decay curve within the preset time window are lower than the safety threshold, an oil circuit partition isolation verification pass mark is generated.

[0288] Furthermore, all concentration values ​​within a preset time window are extracted from the predicted decay curve. First, the previously calculated predicted decay curve of oil contaminant concentration in the non-isolation zone is obtained. Then, according to the start and end times of the preset time window, the concentration values ​​corresponding to all time points within that time period are extracted. The extraction range is strictly limited to the 5th to 30th second, excluding concentration data before the 5th second, to avoid initial fluctuations interfering with the verification results. After extraction, a concentration dataset within the time window is formed, arranged in chronological order by timestamp, with each time point corresponding to a concentration value. Then, each concentration value within the time window is directly compared with the safety threshold of 0.3 mg / L according to the chronological order of timestamps. Yes, the comparison method is a simple size judgment, only checking whether the current concentration value is lower than the safety threshold. The oil circuit zoning isolation verification is deemed to have passed only if all concentration values ​​within the preset time window are strictly lower than the safety threshold of 0.3 mg / L, without any exceptions. If the concentration value at any time point is greater than or equal to the safety threshold, the verification is deemed to have failed. An oil circuit zoning isolation verification pass mark is generated. When the isolation verification is deemed to have passed, the system automatically generates an isolation verification pass mark. The generation process is a direct assignment. After generation, the verification record is saved. The record content includes information such as the mark, verification time, safety threshold, time window, and concentration comparison result, ensuring that the record is complete and clear.

[0289] S4.4: Combine the pollution source, the boundary of the oil passage to be isolated, and the oil passage isolation verification mark according to the preset instruction template to generate a pollution disposal instruction summary.

[0290] Furthermore, the preset instruction template is a standardized text template. The template structure is set according to the logical order of "pollution source - boundary of the oil pipeline to be isolated - isolation verification result". Three fixed information fill positions are reserved in the template, corresponding to three key information items respectively. The information type of each fill position is clearly marked to avoid misfilling. For example, if the source of contamination is the piston rod seal wear path, the boundary of the oil circuit to be isolated is the isolation valve position number 2001, 2002, the isolation pipeline section is identified as pipeline 201 to pipeline 204, the list of associated components is hydraulic cylinder, piston rod seal, inlet filter, and return filter, and the isolation verification pass indicator is isolation verification pass, then the filled template content is: Contamination Disposal Instruction Summary, the source of contamination is the piston rod seal wear path, the specific boundary of the oil circuit to be isolated is as follows, the isolation valve position number is 2001, 2002, the isolation pipeline section is identified as pipeline 201 to pipeline 204, the list of associated components is hydraulic cylinder, piston rod seal, inlet filter, and return filter, the oil circuit zoning isolation verification result is qualified, and the verification pass indicator is isolation verification pass.

[0291] Furthermore, the step of embedding the binary encoding of the verification timestamp into the processing instruction digest to generate an early warning processing data packet includes:

[0292] (1) Obtain the system time when the oil circuit partition isolation verification passed the identification, and convert the system time into a timestamp string in Coordinated Universal Time format;

[0293] (2) Based on the binary encoding table, each character in the timestamp string is converted into the corresponding eight-bit binary code, and all the eight-bit binary codes are concatenated in order to obtain the binary timestamp sequence;

[0294] (3) Obtain the pollution disposal instruction summary and locate the last reserved field in the pollution disposal instruction summary;

[0295] (4) Replace the least significant bit of each byte in the last reserved field with the binary timestamp sequence in turn according to the least significant bit embedding method to obtain the processing instruction digest after embedding the timestamp;

[0296] (5) Perform cyclic redundancy check on the processing instruction digest after the embedded timestamp, append the calculated check code to the end of it, and generate an early warning processing data packet.

[0297] Furthermore, uploading the early warning and handling data packet to the industrial blockchain for evidence storage includes:

[0298] (1) Obtain the early warning and handling data packet, and extract the payload and metadata from it; the metadata includes the device identifier, generation timestamp and data packet sequence number;

[0299] (2) The payload is calculated based on the hash algorithm to obtain the payload hash value, and the payload hash value is combined with the metadata to form a blockchain transaction data body. The hash algorithm is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0300] (3) The blockchain transaction data body is sent to the consensus node in the industrial blockchain network via the remote procedure call protocol;

[0301] (4) The consensus node extracts the sender's signature from the received blockchain transaction data body, performs a signature verification operation on the sender's signature using the sender's public key to restore the original digest, and recalculates the local hash digest of the payload based on the hash algorithm. The original digest obtained from the signature verification is compared with the locally recalculated hash digest to complete the signature verification.

[0302] (5) The consensus node extracts the timestamp field from the blockchain transaction data body, calculates the absolute value of the difference between the timestamp field and the current system time, and completes the timestamp verification when the absolute value of the difference is less than the preset time window threshold.

[0303] (6) After both the signature verification and the timestamp verification are passed, the consensus node packages the blockchain transaction data into the latest block and broadcasts it to all nodes in the industrial blockchain network, and then returns the transaction hash as a certificate of evidence.

[0304] Example 2

[0305] Please see Figure 3 Another embodiment of the present invention provides: an early warning and response system for contamination intrusion into hydraulic systems, comprising:

[0306] The feature recognition module 10 is used to collect the time-series signal of oil particle size in the return oil pipeline and main oil pipeline of the hydraulic system, the dielectric constant of oil in the return oil zone in the middle of the oil tank, the air content in the high-pressure pipeline, the moisture content at the bottom of the oil tank and the pressure pulsation at the pump outlet, extract the time-series abrupt change feature factors of external pollution intrusion, and input them into the radial basis kernel function support vector machine classifier after dimensionality reduction by principal component analysis, and output the pollution type identifiers of water mist intrusion, dust particle intrusion, fuel mixing or bubble entrainment.

[0307] The digital twin modeling module 20 is used to collect topological parameters of the hydraulic system to construct a geometric model, couple the oil flow mechanism sub-model, the pollutant transport mechanism sub-model and the long short-term memory network deviation correction sub-model to form a digital twin image of the hydraulic system, query the physical property parameters of pollutants by pollution type, predict the pollution diffusion path sequence, calculate the damage priority deviation coefficient between the actual and predicted diffusion states, and compare the classification threshold to determine the low, medium and high levels of external pollution intrusion risk.

[0308] The pollution source tracing module 30 is used to analyze the time-series signal of oil particle size in medium and high risk levels by particle size interval, construct a particle size-time mutation matrix, and perform cosine similarity matching between the matrix and the intrusion path feature library to trace the pollution source such as respirator filter failure, piston rod seal wear, pipeline joint loosening or new oil filling contamination.

[0309] The isolation and disposal module 40 is used to query the isolation strategy library based on the pollution source, simulate the oil circuit zoning isolation operation in the digital twin mirror and verify the isolation effectiveness, generate a pollution disposal instruction summary containing the pollution source, isolation boundary and verification mark, embed the verification timestamp binary code into the instruction summary and verify it, generate an early warning disposal data package and upload it to the industrial blockchain, complete transaction verification and evidence storage, and ensure that the disposal process is trustworthy and traceable.

[0310] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of the present invention without departing from the spirit and scope of the present invention. All of these variations are within the protection scope of the present invention.

Claims

1. A method for early warning and handling of contamination intrusion in hydraulic systems, characterized in that, include: Collect time-series signals of oil particle size, oil dielectric constant, air content, moisture content and pressure pulsation of the hydraulic system, extract time-series abrupt change characteristic factors of external pollution intrusion, input the time-series abrupt change characteristic factors into a pre-constructed association identification model, and output the pollution type; Construct a digital twin image of the hydraulic system, predict the pollution diffusion path sequence by combining the pollution type, and calculate the damage priority deviation coefficient between the actual pollutant diffusion state and the pollution diffusion path sequence. By comparing the damage priority deviation coefficient with the preset grading threshold, the risk level of external pollution intrusion into the hydraulic system is determined. For hydraulic systems with external pollution intrusion, the corresponding intrusion path is matched according to the particle size level of the pollutants to trace the source of pollution. Based on the pollution source, the oil pipeline is isolated and verified, a pollution disposal instruction summary is generated, and the verification timestamp is binary encoded and embedded into the disposal instruction summary to generate an early warning disposal data package and upload it to the industrial blockchain for evidence storage.

2. The early warning and handling method for contamination intrusion in hydraulic systems as described in claim 1, characterized in that, The extraction of temporal mutation characteristic factors of exogenous pollution invasion includes: Based on the cubic spline interpolation method, the timestamps of the oil particle size time series signal, oil dielectric constant, air content, moisture content and pressure pulsation are unified to the same time reference sequence to obtain a time-aligned multidimensional physical quantity sequence. A sliding window differential operation is performed on the time-aligned oil particle size time series signal to obtain the particle concentration change rate sequence. The 3σ criterion is used to detect outliers in the particle concentration change rate sequence, and the time position corresponding to the particle concentration change rate exceeding three standard deviations is marked as the particle size mutation start point. Based on the particle size mutation start point, the oil dielectric constant segment, air content segment, moisture content segment, and pressure pulsation segment within a preset time window are extracted from the time-aligned multidimensional physical quantity sequence, and first-order difference operations are performed on each segment to obtain the corresponding dielectric constant change rate sequence, air content change rate sequence, moisture content change rate sequence, and pressure pulsation change rate sequence. These sequences are then combined with the particle concentration change rate sequence to form a multidimensional mutation feature vector, which is used as the temporal mutation feature factor.

3. The early warning and handling method for contamination intrusion in hydraulic systems as described in claim 2, characterized in that, The step of inputting time-series mutation feature factors into a pre-constructed association identification model and outputting contamination types includes: The multidimensional mutation feature vector is subjected to dimensionality reduction processing based on principal component analysis. Principal components with a cumulative variance contribution rate greater than a preset contribution rate threshold are selected to obtain the dimensionality-reduced time-series mutation feature factors. Load a pre-built association recognition model; the association recognition model is a support vector machine classifier based on radial basis function kernel function; The dimensionality-reduced temporal mutation feature factors are used as input vectors and arranged in chronological order to form a feature input sequence. Each trained support vector is used as a reference vector, and the radial basis function kernel is used to calculate the kernel function value between the input vector and each reference vector. The support vectors are obtained by training on historical pollution intrusion samples using a sequence minimum optimization algorithm. A one-to-many classification strategy is used to calculate the weighted sum values ​​corresponding to the four types of pollution. The weighted sum values ​​of the four types are compared, and the pollution type identifier corresponding to the maximum value is output. The pollution type identifier is one of the following: water mist intrusion, dust particle intrusion, fuel oil mixing, or bubble entrainment.

4. The early warning and handling method for contamination intrusion in hydraulic systems as described in claim 3, characterized in that, The construction of the digital twin mirror image of the hydraulic system includes: Collect the topological parameters of the physical entities of the hydraulic system and construct a geometric model of the hydraulic system; A sub-model of oil flow mechanism was constructed based on the Navier-Stokes equations, and a sub-model of pollutant transport mechanism was constructed based on the principles of mass conservation and momentum conservation. The oil flow mechanism sub-model and the pollutant transport mechanism sub-model are coupled through a data interface to obtain the output value of the joint mechanism model; The measured values ​​of various flow field physical quantities corresponding to key measuring points of the hydraulic system physical entity are collected. Simultaneously, the simulated physical quantities output by the joint mechanism model at the corresponding measuring points and at the same time are obtained. For the same measuring point and the same physical quantity, the difference between the simulated physical quantity and the measured value is calculated to obtain the deviation of a single physical quantity at a single measuring point. After performing min-max normalization on the deviations of a single physical quantity at all measuring points and all physical quantities, the deviations are spliced ​​according to the spatial order and temporal order of the measuring points to form a deviation sequence. A deviation correction sub-model is constructed based on a long short-term memory network. The deviation sequence is used as input to train the deviation correction sub-model to output the deviation prediction value in a preset time domain. The predicted deviation value is added to the output value of the joint mechanism model to obtain the twin output value; The hydraulic system geometric model, the oil flow mechanism sub-model, the contaminant transport mechanism sub-model, and the deviation correction sub-model are coupled together through a data interface, and the twin output value is used as the output end of the digital twin mirror to form a digital twin mirror of the hydraulic system.

5. The early warning and handling method for contamination intrusion in hydraulic systems as described in claim 4, characterized in that, The method of predicting pollution diffusion path sequences by combining pollution types includes: The corresponding physical property parameters are retrieved from a preset pollutant attribute database based on the pollution type identifier; the physical property parameters include average particle size, particle density, particle shape coefficient, oil miscibility index, and bubble surface tension coefficient. The physical property parameters are assigned to the pollutant transport mechanism sub-model, and the oil velocity distribution and pressure distribution output by the current oil flow mechanism sub-model are used as initial boundary conditions to input the pollutant transport mechanism model. After receiving the physical property parameters and the initial boundary conditions, the pollutant transport mechanism sub-model calculates the mass concentration of pollutants at each node of the hydraulic system over time based on the convection-diffusion equation, thus obtaining the concentration time series of each node. For each node, the time when the mass concentration first reaches the preset detection threshold in the corresponding concentration time series is marked as the pollution arrival time of that node. The corresponding nodes are sorted in order from early to late according to the pollution arrival time, and the path of pollutant propagation between adjacent nodes is used as the edge to generate a tree-structured pollution diffusion path sequence.

6. The early warning and handling method for contamination intrusion in hydraulic systems as described in claim 5, characterized in that, The damage priority deviation coefficient for calculating the actual pollutant diffusion state and pollution diffusion path sequence includes: The actual pollutant concentration, actual pollutant particle size distribution, and actual pollutant accumulation rate of each hydraulic component are extracted in real time from the twin output value output by the digital twin mirror of the hydraulic system and combined into an actual pollutant diffusion state vector. The predicted pollutant concentration, predicted pollutant particle size distribution, and predicted arrival time carried by each node are extracted from the pollution diffusion path sequence and combined into a predicted diffusion state vector. The state deviation between the actual pollutant diffusion state vector and the predicted diffusion state vector is calculated based on Mahalanobis distance; each component of the state deviation corresponds to the deviation value of a hydraulic component. Look up the damage sensitivity coefficient of each hydraulic component from the preset component damage sensitivity coefficient table; Each component of the state deviation is multiplied by the damage sensitivity coefficient of the corresponding hydraulic component to obtain a weighted deviation component. All weighted deviation components are then accumulated in the order of nodes in the pollution diffusion path sequence and normalized by min-max to output the damage priority deviation coefficient.

7. The early warning and handling method for contamination intrusion in hydraulic systems as described in claim 6, characterized in that, By comparing the damage priority deviation coefficient with a preset grading threshold, the risk level of external contamination intrusion into the hydraulic system is determined, including: Obtain a preset first grading threshold and a second grading threshold, and simultaneously obtain a damage priority deviation coefficient; the first grading threshold is less than the second grading threshold; When the damage priority deviation coefficient is less than the first classification threshold, it is determined to be a low-risk level, and a corresponding low-risk level code is generated. When the damage priority deviation coefficient is greater than or equal to the first grading threshold and less than the second grading threshold, it is determined to be of medium risk level, and a corresponding medium risk level code is generated. When the damage priority deviation coefficient is greater than or equal to the second classification threshold, it is determined to be a high-risk level, and a corresponding high-risk level code is generated. The generated low-risk, medium-risk, and high-risk level codes are output as the results of the determination of the risk level of external pollution intrusion.

8. The early warning and handling method for contamination intrusion in hydraulic systems as described in claim 7, characterized in that, For hydraulic systems susceptible to external contamination, the method of matching corresponding intrusion paths according to contaminant particle size levels to trace the source of contamination includes: Obtain the risk level of external pollution intrusion. When the risk level of external pollution intrusion is medium risk or high risk, execute: The time-series signal of oil particle size is acquired, and the time-series signal of oil particle size is analyzed in layers according to the preset particle size interval boundary to obtain sub-concentration sequences corresponding to multiple particle size levels. Mutation point detection is performed on the sub-concentration sequences corresponding to each particle size level. The cumulative sum and statistics of each sub-concentration sequence are calculated. The time point when the cumulative sum and statistics exceed the preset mutation judgment threshold is marked as the mutation start time of the corresponding particle size level. The mutation start times of all particle size levels are arranged in ascending order of particle size to construct a particle size-time mutation matrix. The particle size-time mutation matrix is ​​matched with each invasion path feature template in the preset invasion path feature library by cosine similarity, and the invasion path corresponding to the invasion path feature template with the largest cosine similarity is selected as the pollution source output; the invasion path feature library stores the standard matrices corresponding to the respirator filter failure path, piston rod seal wear path, pipeline joint loosening path, and new oil filling pollution path, respectively.

9. A method for early warning and handling of contamination intrusion in hydraulic systems as described in claim 8, characterized in that, The process of verifying the isolation of oil pipelines based on the source of pollution and generating a pollution disposal instruction summary includes: Based on the pollution source, the corresponding oil circuit partition boundary to be isolated is queried from the preset isolation strategy library. The oil circuit partition boundary to be isolated includes the isolation valve tag number, the isolation pipeline section identifier, and the list of associated components. In the digital twin mirror of the hydraulic system, the opening parameter of the valve element corresponding to the isolation valve position number is set to zero, the valve closing operation is simulated, and based on the simulated closing conditions, the oil flow mechanism sub-model and the contaminant transport mechanism sub-model are rerun to calculate the predicted decay curve of the oil contaminant concentration in the non-isolation zone. Obtain a safety threshold; when all concentration values ​​of the predicted decay curve within a preset time window are lower than the safety threshold, generate an oil circuit zoning isolation verification pass flag. The pollution source, the boundary of the oil pipeline to be isolated, and the oil pipeline isolation verification mark are combined according to the preset instruction template to generate a pollution disposal instruction summary.

10. A warning and response system for contamination intrusion in hydraulic systems, used to implement the warning and response method for contamination intrusion in hydraulic systems as described in any one of claims 1-9, characterized in that, include: The feature recognition module is used to collect time-series signals of oil particle size, oil dielectric constant, air content, moisture content and pressure pulsation, extract time-series abrupt change feature factors of external pollution intrusion, and input them into a radial basis kernel function support vector machine classifier after dimensionality reduction by principal component analysis, and output pollution type identifier. The digital twin modeling module is used to construct a digital twin image of the hydraulic system, query the physical property parameters of pollutants by combining the pollution type, predict the pollution diffusion path sequence, calculate the damage priority deviation coefficient between the actual and predicted diffusion states, and compare the classification threshold to determine the risk level of external pollution intrusion. The pollution source tracing module is used to analyze the time-series signal of oil particle size according to particle size intervals under medium and high risk levels, construct a particle size-time mutation matrix, and perform cosine similarity matching with the intrusion path feature library to trace the pollution source. The isolation and disposal module is used to query the isolation strategy library based on the pollution source, simulate the oil circuit zoning isolation operation in the digital twin mirror and verify the isolation effectiveness, generate a pollution disposal instruction summary, embed the verification timestamp binary code into the pollution disposal instruction summary and verify it, generate an early warning disposal data package and upload it to the industrial blockchain to complete transaction verification and evidence storage.

Images