Oil quantity change trend prediction method and system for automatic oil injection device

Abstract

The invention discloses an oil quantity change trend prediction method and system for an automatic oil injection device, and relates to the related field of data processing, and the method comprises the steps: extracting an oil quantity state data set of an oil injection execution mechanism, carrying out the three-stage decomposition, and obtaining the three-stage structured state features; acquiring a currently executed equipment working condition mode of the oil injection executing mechanism; a self-adaptive structured weight vector is introduced according to an equipment working condition mode, the three-stage structured state features are processed based on the self-adaptive structured weight vector, and an oil quantity change form fingerprint vector is extracted; and performing similarity matching on the oil mass change form fingerprint vector and an oil mass change trend template library to obtain a matched oil mass change trend template, and performing oil mass change trend prediction of the oil mass state data set according to the matched oil mass change trend template. The technical problem that an existing automatic oiling device is low in oil quantity change trend prediction precision is solved, and the technical effect of improving the oil quantity change trend prediction precision is achieved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a method and system for predicting the oil quantity change trend of an automatic oil injection device. Background Technology

[0002] Precise oil level control in automatic oiling systems is crucial for ensuring stable operation and extending the lifespan of industrial equipment. Accurate prediction of oil level trends directly impacts production efficiency and maintenance cost optimization. Currently, most industry methods rely on traditional time series fitting to predict oil level trends. This involves collecting historical oil level data and constructing a fitting function using methods such as moving averages or ARIMA models to extrapolate trends. However, these methods only focus on the surface-level changes in oil level values, neglecting the multifactorial nature of oil level variations and the dynamic differences in equipment operating conditions. This makes the predictions susceptible to interference from actual industrial factors such as operating condition fluctuations and oil penetration losses, resulting in low prediction accuracy and insufficient robustness.

[0003] Currently, the prediction of oil quantity change trends in automatic oil injection devices suffers from low accuracy. Summary of the Invention

[0004] This application provides a method and system for predicting oil volume change trends in automatic oil injection devices. It employs methods such as extracting oil volume status datasets from the oil injection actuator, decomposing them into structured state features based on three categories of oil volume change contributions (oil injection input, structural penetration, and operational consumption) through a three-stage process, obtaining the current operating mode of the equipment, and introducing an adaptive structured weight vector to process the structured state features. This extracts an oil volume change pattern fingerprint vector, matches this fingerprint vector with an oil volume change trend template library, and predicts the oil volume change trend based on the matched trend template. These techniques solve the technical problem of low prediction accuracy in existing automatic oil injection device oil volume change trend prediction methods, achieving a significant improvement in the accuracy of oil volume change trend prediction.

[0005] This application provides a method for predicting the oil volume change trend of an automatic oil injection device, comprising: extracting an oil volume status dataset of an oil injection actuator; performing a three-stage decomposition on the oil volume status dataset to obtain three-stage structured status features, wherein the three-stage structured status features include a first type of oil volume change contribution based on oil injection input, a second type of oil volume change contribution based on structural penetration, and a third type of oil volume change contribution based on operating consumption; obtaining the current operating mode of the equipment of the oil injection actuator; introducing an adaptive structured weight vector according to the operating mode; processing the three-stage structured status features based on the adaptive structured weight vector to extract an oil volume change morphology fingerprint vector; performing similarity matching between the oil volume change morphology fingerprint vector and an oil volume change trend template library to obtain a matching oil volume change trend template; and performing oil volume change trend prediction of the oil volume status dataset according to the matching oil volume change trend template.

[0006] In a possible implementation, the oil quantity status dataset is decomposed into three stages to obtain three-stage structured state features, and the following processing is performed: the oil quantity status dataset is processed into time series data to obtain oil quantity status time series data; the first-order difference sequence data of the oil quantity status time series data is calculated; a threshold for the rate of change of the oil injection event is set, and the corresponding time series nodes in the first-order difference sequence data that are greater than the rate of change of the oil injection event are marked as the time of occurrence of the oil injection event, and the oil injection event time point set is recorded; each oil injection event in the oil injection event time point set is decomposed into three stages to obtain three-stage structured state features.

[0007] In a possible implementation, the following processing is performed: Each oil injection event in the set of oil injection event time points is decomposed into three stages to obtain three-stage structured state features. The method for obtaining the first type of oil quantity change contribution includes: performing pulse integration on the oil injection input amount of each oil injection event with respect to the oil quantity state data to obtain oil injection input integral data; calculating the instantaneous effective coefficient of oil injection on the oil injection input integral data using a monotonically bounded mapping function to obtain the instantaneous effective coefficient of oil injection; and outputting the first type of oil quantity change contribution based on the product of the instantaneous effective coefficient of oil injection and the oil injection input integral data.

[0008] In a possible implementation, the method for obtaining the second type of oil volume change contribution by performing the following processing includes: determining the penetration start time and penetration end time for each injection event; establishing a double exponential decay model, wherein the double exponential decay model calculates the decay of the fast penetration component and the decay of the slow penetration component, and the double exponential decay model obtains integral data based on the penetration start time and penetration end time to obtain injection penetration integral data; setting a viscosity influence factor according to the viscosity of the injection medium; and normalizing the injection penetration integral data according to the viscosity influence factor to obtain the second type of oil volume change contribution.

[0009] In a possible implementation, the following processing is performed: the method for obtaining the third type of oil quantity change contribution includes: calculating the oil quantity consumption per unit time for each oil injection event based on the oil quantity status data; optimizing the oil quantity consumption by collecting the temperature status of the oil injection actuator; and normalizing the optimized oil quantity consumption as a negative oil quantity change output to obtain the third type of oil quantity change contribution.

[0010] In a possible implementation, an adaptive structured weight vector is introduced based on the equipment operating condition mode, and the following processing is performed: an initial structured weight vector is obtained based on the equipment operating condition mode; wherein, the equipment operating condition mode includes obtaining the current operating condition feature vector, and obtaining it by calculating fuzzy membership degrees based on multiple sets of operating condition feature vector samples from multiple pre-built operating condition modes and the current operating condition feature vector; wherein, the multiple pre-built operating condition modes include at least a shutdown operating condition mode, a low-load operating condition mode, a normal operating condition mode, a high-load operating condition mode, and an overload operating condition mode; the operating condition mode deviation is calculated based on the fuzzy membership degrees, and the initial structured weight vector is adaptively corrected according to the operating condition mode deviation to obtain an adaptive structured weight vector.

[0011] In a possible implementation, the three-stage structured state features are processed based on the adaptive structured weight vector to extract the oil volume change morphology fingerprint vector. The following processing is performed: high-dimensional feature convolution is performed on the three-stage structured state features to obtain high-dimensional features of the three-stage structured state in multiple dimensions; the adaptive structured weight vector is expanded into an adaptive structured weight high-dimensional vector in the multiple dimensions; feature weighted fusion is performed on the high-dimensional features of the three-stage structured state according to the adaptive structured weight high-dimensional vector to obtain state fusion features; the first-order statistical feature vector, trend feature vector, frequency domain feature vector, and nonlinear feature vector of the state fusion features are extracted to form the oil volume change morphology fingerprint vector.

[0012] In a possible implementation, the oil volume change pattern fingerprint vector is matched with the oil volume change trend template library to obtain a matching oil volume change trend template. The following processing is then performed: A fuel volume change trend template library is constructed, where each template includes a template pattern fingerprint vector, a trend prediction parameter set, a trend type, a template confidence level, and an applicable operating condition mode label. Based on the fuel volume change trend template library, feature space similarity, trend pattern similarity, and operating condition matching degree are calculated on the oil volume change pattern fingerprint vector to obtain a comprehensive score. Matching oil volume change trend templates with a comprehensive score greater than a preset score threshold are then selected.

[0013] In a possible implementation, in addition to filtering matching oil volume change trend templates that are greater than a preset score threshold according to the comprehensive score, the following processing is also performed: if the number of matching oil volume change trend templates returned is multiple, the multiple returned matching oil volume change trend templates are fused to obtain a fused oil volume change trend template; and oil volume change trend prediction of the oil volume status dataset is performed based on the fused oil volume change trend template.

[0014] This application also provides a system for predicting the oil volume change trend of an automatic oil injection device, comprising: an oil volume state decomposition module, used to extract an oil volume state dataset of an oil injection actuator, perform a three-stage decomposition on the oil volume state dataset to obtain three-stage structured state features, the three-stage structured state features including a first type of oil volume change contribution based on oil injection input, a second type of oil volume change contribution based on structural penetration, and a third type of oil volume change contribution based on operating consumption; an equipment operating mode acquisition module, used to acquire the currently executed equipment operating mode of the oil injection actuator; an oil volume change morphology fingerprint vector extraction module, used to introduce an adaptive structured weight vector according to the equipment operating mode, process the three-stage structured state features based on the adaptive structured weight vector, and extract an oil volume change morphology fingerprint vector; and an oil volume change trend prediction module, used to perform similarity matching between the oil volume change morphology fingerprint vector and an oil volume change trend template library to obtain a matching oil volume change trend template, and perform oil volume change trend prediction of the oil volume state dataset according to the matching oil volume change trend template.

[0015] This application proposes a method and system for predicting oil quantity change trends in automatic oil injection devices. First, it extracts an oil quantity status dataset from the oil injection actuator. This dataset is then decomposed into three stages to obtain three-stage structured state features. These features include a first type of oil quantity change contribution based on oil injection input, a second type based on structural penetration, and a third type based on operational consumption. Next, it obtains the current operating mode of the oil injection actuator. Then, it introduces an adaptive structured weight vector based on this operating mode. The three-stage structured state features are processed using this adaptive weight vector to extract an oil quantity change pattern fingerprint vector. Finally, the fingerprint vector is matched with a similarity database of oil quantity change trend templates to obtain a matching oil quantity change trend template. The oil quantity change trend is then predicted based on this template. Through this process, the proposed method and system achieve the technical effect of improving the accuracy of oil quantity change trend prediction. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating a method for predicting the oil quantity change trend of an automatic oiling device, provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of a system for predicting the oil quantity change trend of an automatic oiling device, provided in an embodiment of this application.

[0019] Figure labeling: Oil quantity status decomposition module 10, equipment operating condition mode acquisition module 20, oil quantity change pattern fingerprint vector extraction module 30, oil quantity change trend prediction module 40. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structure, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] This application provides a method for predicting the oil quantity change trend of an automatic oiling device, such as... Figure 1As shown, the method includes: Step S100: Extract the oil quantity status dataset of the oil injection actuator, perform a three-stage decomposition on the oil quantity status dataset, and obtain three-stage structured status features. The three-stage structured status features include a first type of oil quantity change contribution based on oil injection input, a second type of oil quantity change contribution based on structural penetration, and a third type of oil quantity change contribution based on operating consumption.

[0022] Specifically, the oil injection actuator refers to the core mechanical component in an automatic oil injection device that completes the oil injection action and carries the lubricating oil, such as the bearing oil injection chamber, gearbox oil injection module, and guide rail oil injection chamber. Continuous raw data on the oil quantity status is collected by the oil quantity sensor, flow sensor, and pressure sensor of the oil injection actuator, forming an oil quantity status dataset containing real-time oil quantity values, acquisition time, oil injection trigger signal, and mechanism operating speed. The acquisition frequency of the dataset is set to a fixed value, for example, once per second. Data preprocessing is performed on the oil quantity status dataset, including using the quartile method to remove abnormal oil quantity data collected by the sensors that are less than the lower quartile minus 1.5 times the interquartile range or greater than the upper quartile plus 1.5 times the interquartile range; missing oil quantity data is completed using linear interpolation; and the oil quantity data is normalized to a normalization range of 0 to 1. Based on the physical causes of oil volume changes, a three-stage decomposition process is executed. This process breaks down the oil volume changes in the oil volume state dataset into three independent and non-overlapping components according to their core causes. First, the time point of the oil injection event is located. Then, for each oil injection event and its corresponding oil volume change cycle, the quantified values ​​of the three types of oil volume change contributions are calculated. Finally, a three-stage structured state feature consisting of these three types of contributions is output. Specifically: the first type of oil volume change contribution is the positive change in oil volume caused solely by the active oil injection action; the second type is the oil loss change resulting from oil seeping into the gaps, cavities, and lubrication contact surfaces of the actuator after injection; and the third type is the oil consumption change resulting from oil depletion due to lubrication friction, high-temperature evaporation, and normal leakage during normal operation of the actuator. The three-stage structured state feature is represented as a three-dimensional feature vector, where the values ​​within the vector are normalized quantified values ​​of the oil volume changes.

[0023] In one possible implementation, the oil quantity status dataset is decomposed into three stages to obtain three-stage structured status features. Step S100 further includes step S110, which involves performing time series processing on the oil quantity status dataset to obtain oil quantity status time series data. Specifically, two core fields are extracted from the oil quantity status dataset: the oil quantity monitoring value and the corresponding data collection timestamp. Sample data containing null or invalid values ​​in these two fields are removed. The remaining valid sample data are sorted in ascending order according to the collection timestamp values, based on the absolute value of the timestamp. The sorted oil quantity values ​​are then arranged sequentially in chronological order to form oil quantity status time series data. This data is a one-dimensional array, where the array index represents a time node, and the array elements are the oil quantity values ​​corresponding to the time nodes. The interval between time nodes is a fixed collection frequency.

[0024] Step S120: Calculate the first-order difference sequence data of the oil quantity state time series data. Specifically, the length of the oil quantity state time series data is determined to be N data points, which are labeled in chronological order as the 1st to the Nth data points, corresponding to the first oil quantity value, the second oil quantity value, and so on up to the Nth oil quantity value. Perform first-order difference calculation, the calculation rule being: the Mth difference value equals the (M+1)th oil quantity value minus the Mth oil quantity value, where M ranges from 1 to N-1. Arrange all the calculated difference values ​​in chronological order to form the first-order difference sequence data, the length of which is N-1. The values ​​represent the change in oil quantity per unit time, with positive values ​​representing an increase in oil quantity, negative values ​​representing a decrease in oil quantity, and zero representing no change in oil quantity.

[0025] Step S130: Set a threshold for the rate of change of the oil injection event detection. Mark the corresponding time series nodes in the first-order difference sequence data that are greater than the threshold as the time of occurrence of the oil injection event, and record the oil injection event time points. Specifically, the threshold for the rate of change of the oil injection event detection is a pre-set critical value for the change in oil volume used to determine whether an oil injection event has occurred. The value of this threshold is positive and much larger than the natural change in oil volume under normal operating conditions. This threshold is obtained based on the historical oil injection data of the oil injection actuator. Iterate through each value in the first-order difference sequence data and compare it with the threshold one by one. If a difference value is greater than the threshold, the time node corresponding to the difference value is determined to be the time of occurrence of the oil injection event. This time node is the (M+1)th node in the crude oil volume status time series, corresponding to the Mth value of the first-order difference. Record the timestamps and time node numbers of all determined oil injection event occurrence times. After removing duplicate marked time nodes, arrange them in chronological order to form an oil injection event time point set, which is an ordered list.

[0026] Step S140: Perform a three-stage decomposition on each oil injection event in the set of oil injection event time points to obtain three-stage structured state features. Specifically, each oil injection event in the set of oil injection event time points is independently divided, and each oil injection event corresponds to an independent oil volume change cycle. The start point of the cycle is the time when the oil injection event occurs, and the end point of the cycle is the time when the oil volume change rate recovers to the normal operating consumption rate. For the oil volume state data in each independent cycle, calculate the contributions of three types of oil volume changes caused by oil injection input, structural penetration, and operating consumption. The calculation of the three types of contributions is independent of each other, and the calculation order is not required. Normalize the calculated values ​​of the three types of contributions. The normalization range is 0 to 1, and the normalization rule is to divide the value of a single type of contribution by the sum of the absolute values ​​of the three types of contribution values. Combine the normalized values ​​of the three types of contributions in a fixed order to form a three-dimensional feature vector, which is the three-stage structured state feature corresponding to the oil injection event. After summing the feature vectors of all oil injection events, a complete three-stage structured state feature set is formed.

[0027] In one possible implementation, each oil injection event in the set of oil injection event time points is decomposed into three stages to obtain three-stage structured state features and the contribution of the first type of oil quantity change. Step S140 further includes step S141, performing pulse integration on the oil injection input amount of each oil injection event with respect to the oil quantity state data to obtain oil injection input integral data. Specifically, the oil injection input amount is the total volume of oil actively injected into the oil injection actuator when the oil injection event occurs, and this value is the instantaneous increment of oil quantity at the moment the oil injection event occurs. Pulse integration is a time-dimensional integral calculation of the oil injection input amount, and the integration interval is from the moment the oil injection event occurs to the moment the oil injection action completely stops. The oil injection input integral data is a quantized value after integral calculation, representing the comprehensive contribution value of the total input oil quantity and input duration of the oil injection action. Specifically, the duration of the oil injection action for each oil injection event is determined, which is the time interval from the moment the oil injection event occurs to the moment the oil quantity value reaches its peak value, and the oil quantity input rate of the oil injection action is a uniform input. The pulse integral calculation is performed, and the integral formula is: Injection input integral data = Injection input rate × Injection action duration. The integral result is the total input oil volume of the injection action. If the input rate of the injection action is non-uniform, a piecewise integration method is used, dividing the input process into multiple uniform speed segments, integrating each segment separately, and then summing the results to obtain the total integral data.

[0028] Step S142: The immediate effective coefficient of the fuel injection is calculated by applying a monotonically bounded mapping function to the fuel injection input integral data. Specifically, the monotonically bounded mapping function is a pre-defined mathematical function that possesses monotonically increasing and numerically bounded characteristics. The input to the function is the fuel injection input integral data, and the output is the coefficient value within the interval 0 to 1. The specific form of the monotonically bounded mapping function is determined to be a logarithmic mapping function, and the function's expression is: The maximum oil injection input is a fixed value, representing the effective lubrication chamber volume of the oil injection actuator. The core characteristic of this function is that the larger the integral data of the oil injection input, the closer the coefficient value is to 1, and the coefficient value is always greater than 0 and less than or equal to 1, satisfying the requirement of monotonicity and boundedness. The integral data of the oil injection input for each oil injection event is substituted into this function to complete the coefficient calculation. That is, the instantaneous effective oil injection coefficient characterizes the proportion of the oil input that can actually remain in the effective lubrication area of ​​the oil injection actuator. The closer the coefficient value is to 1, the higher the efficiency of the oil injection input; the closer the coefficient value is to 0, the more oil is lost, because not all of the oil input can participate in effective lubrication; some oil will overflow due to excessive instantaneous pressure. This coefficient is the correction value for overflow loss.

[0029] Step S143: Based on the product of the instantaneous effective oil injection coefficient and the oil injection input integral data, output the first type of oil quantity change contribution. Specifically, retrieve the oil injection input integral data and the instantaneous effective oil injection coefficient for the corresponding oil injection event; both values ​​are pre-calculated and determined. Perform a multiplication calculation, with the calculation rule being: original value of the first type of oil quantity change contribution = oil injection input integral data × instantaneous effective oil injection coefficient. That is, the first type of oil quantity change contribution is the effective oil quantity increment brought about by the oil injection input behavior. This value eliminates overflow losses during the oil injection process and is a real, positive change in oil quantity that can participate in lubrication.

[0030] In one possible implementation, to obtain the contribution of the second type of oil volume change, step S140 further includes step S144, determining the penetration start time and penetration end time of each oil injection event. Specifically, the penetration start time is the moment when the oil begins to penetrate into the structural gaps, lubrication contact surfaces, and cavity gaps of the oil injection actuator after the oil injection event occurs. The determination rule is the moment when the oil injection action completely stops and the oil volume reaches its peak value. The penetration end time is the moment when the oil penetration process completely terminates and the oil volume no longer changes due to the penetration behavior. The determination rule is that the absolute value of the oil volume change over three consecutive unit times is less than a preset oil volume change threshold. The timestamps and time node numbers of the penetration start time and end time are recorded, and the time interval between the two time points is calculated, which is the penetration duration. The penetration process is one of the links in the oil volume loss after oil injection. The oil volume change in this process is passive, determined by structural characteristics, and is unrelated to equipment operation.

[0031] Step S145: Establish a double exponential decay model. This model calculates the decay of the rapid penetration component and the slow penetration component. It also obtains integral data based on the penetration start and end times to obtain the oil injection penetration integral data. Specifically, the double exponential decay model is a mathematical model constructed for the structural penetration behavior of oil. The core logic of this model is that the structural penetration of oil is divided into two stages: the first stage is rapid penetration, where the oil quickly fills the surface gaps of the actuator, with a fast penetration rate and large oil loss; the second stage is slow penetration, where the oil slowly penetrates to the deep gaps of the actuator, with a slow penetration rate and small oil loss. The decay patterns in both stages conform to the characteristics of an exponential function. The decay of the rapid penetration component and the decay of the slow penetration component are the two core calculation terms of the model, representing the oil loss in the two stages, respectively. The expression for the double exponential decay model is constructed, and the output of the model is a function of the oil loss during the penetration process over time, as shown in the formula: , where V P (t) represents the amount of oil lost through permeation. It is a fast-penetrating component. It is the slow osmotic component, A f It is the initial amplitude of the rapid penetration component, A s It is the initial amplitude of the slow permeation component, τ f τ is the decay time constant of the rapid permeation process. s It is the decay time constant of the slow permeation process, where t is the current time point. This is the end time of the i-th oil injection event. Substituting the start and end times of the permeation into the model, the integration interval is determined as the permeation duration. A definite integral is calculated on the model function, and the result of the integral calculation is the oil injection permeation integral data. This data is positive and represents the total oil loss during the permeation process.

[0032] Step S146: Set the viscosity influence factor based on the viscosity of the injection medium. Specifically, a table corresponding to viscosity values ​​and viscosity influence factors is pre-established, based on statistical data from lubricating oil penetration experiments. The viscosity of the injection medium is a core physical characteristic of lubricating oil. Higher viscosity lubricating oil has poorer fluidity, weaker penetration ability, and less oil loss due to structural penetration; conversely, lower viscosity lubricating oil has better fluidity, stronger penetration ability, and greater oil loss due to structural penetration. The viscosity influence factor is a quantitative correction coefficient set based on the actual viscosity value of the lubricating oil. The coefficient value is negatively correlated with the viscosity value; that is, the higher the viscosity, the smaller the coefficient value. The actual viscosity value of the lubricating oil used in the current injection actuator is collected, and the corresponding viscosity influence factor is obtained by looking up the table.

[0033] Step S147: Normalize the oil injection penetration integral data according to the viscosity influence factor to obtain the second type of oil volume change contribution. Specifically, retrieve the oil injection penetration integral data and viscosity influence factor for the corresponding oil injection event; both values ​​are pre-calculated and determined. Perform a multiplication calculation, the calculation rule being: original value of the second type of oil volume change contribution = oil injection penetration integral data × viscosity influence factor. Normalize this original value, the normalization rule being that the original value is divided by the sum of the absolute values ​​of the original values ​​of the three types of oil volume change contributions, which is the final value of the second type of oil volume change contribution. The second type of oil volume change contribution is the effective oil volume loss caused by structural penetration of the oil. This value, after correction for viscosity characteristics, is the oil volume change determined by structural penetration behavior.

[0034] In one possible implementation, to obtain the contribution of the third type of oil quantity change, step S140 further includes step S148, which calculates the oil consumption per unit time for each oil injection event based on the oil quantity status data. Specifically, the oil consumption per unit time is the rate of oil loss caused by lubrication friction, high-temperature evaporation, and normal leakage during the normal operation of the oil injection actuator. The calculation rule is to divide the oil quantity change in the stable operation phase after the penetration process ends by the corresponding time interval. This value is a fixed positive number, representing the uniform oil loss caused by equipment operation. This loss is independent of oil injection input and structural penetration; it is an inherent characteristic of equipment operation and the core long-term influencing factor of oil quantity change. For each oil injection event, the penetration end time point is located as the starting time point of operation consumption. Starting from this time point, the oil quantity status data of the subsequent continuous stable operation phase without interference from other oil injection events are selected. The total oil quantity loss in this phase is calculated, i.e., the initial oil quantity value minus the ending oil quantity value. Then, the total time interval of this phase is calculated by performing a division calculation.

[0035] Step S149: The temperature status of the lubrication actuator is collected to optimize the oil consumption. The optimized oil consumption is normalized as a negative oil consumption change, resulting in the third type of oil consumption change contribution. Specifically, the temperature status of the lubrication actuator is a core environmental factor affecting the oil consumption rate. Higher temperatures lead to faster lubricant evaporation, greater lubrication friction losses, and a higher oil consumption rate; lower temperatures result in a lower oil consumption rate. A table is established to correspond to temperature values ​​and temperature correction coefficients, based on experimental data statistics. The real-time operating temperature of the lubrication actuator is collected, and the corresponding temperature correction coefficient is obtained by looking up the table. Optimization calculations are performed, and the optimized oil consumption per unit time equals the original consumption multiplied by the temperature correction coefficient. The total oil consumption corresponding to the lubrication event is calculated, i.e., the optimized consumption multiplied by the operating duration. This value is used as a negative oil consumption change and normalized according to the same rules as the first two types of contributions. The normalized value is the final value of the third type of oil consumption change contribution.

[0036] Step S200: Obtain the current operating mode of the oil injection actuator.

[0037] Specifically, the equipment operating condition mode refers to the operational status type of the main equipment to which the lubrication actuator belongs. It is a comprehensive classification of the core operating parameters such as equipment load, speed, running time, and temperature. Operating condition parameter acquisition sensors are deployed on the main equipment of the lubrication actuator. The core operating condition parameters collected include equipment load rate, spindle speed, equipment operating temperature, equipment continuous running time, and equipment start-stop frequency. The acquisition frequency is consistent with the oil quantity status data acquisition frequency. The collected operating condition parameters are standardized to eliminate dimensional differences between parameters. The standardization formula is the actual value of the parameter minus the parameter mean, then divided by the parameter standard deviation. The standardized operating condition parameters are combined into an operating condition feature vector. Through preset operating condition mode matching rules, the current equipment operating condition mode is determined and output. It is noteworthy that the contribution of the three types of oil quantity changes of the lubrication actuator to the overall oil quantity change varies significantly under different operating conditions.

[0038] Step S300: Based on the equipment operating mode, an adaptive structured weight vector is introduced, and the three-stage structured state features are processed based on the adaptive structured weight vector to extract the oil quantity change pattern fingerprint vector.

[0039] Specifically, based on the current operating mode of the equipment, a preset initial structured weight vector is retrieved. Then, the initial weight vector is dynamically corrected using the calculation results of the operating mode deviation, resulting in an adaptive structured weight vector. This adaptive structured weight vector is a three-dimensional weight value set dynamically adjusted according to the current operating mode of the equipment. The three weight values ​​in this vector correspond to the three types of oil quantity change contributions in the three-stage structured state features. The magnitude of the weight value represents the influence weight of the corresponding oil quantity change contribution on the overall oil quantity change, and the sum of the weight values ​​is always equal to 1. High-dimensional feature convolution processing is performed on the three-stage structured state features to expand the three-dimensional basic features into multi-dimensional high-dimensional features. The number of expanded dimensions is determined according to the type of oil injection actuator, and the expansion rule is the basic feature dimension multiplied by the number of operating parameters. The adaptive structured weight vector is synchronously expanded into a high-dimensional weight vector of the same dimension. The high-dimensional three-stage structured state features are weighted and fused to obtain the fused state features. Full-dimensional feature extraction is performed on the fused features. The extracted feature types include statistical laws, trend laws, frequency laws, and nonlinear laws of oil volume changes. After combining the four types of features in a fixed order, the oil volume change morphological fingerprint vector is output.

[0040] In one possible implementation, an adaptive structured weight vector is introduced based on the equipment operating condition mode. Step S300 further includes step S310, obtaining an initial structured weight vector based on the equipment operating condition mode. The equipment operating condition mode includes obtaining the current operating condition feature vector and calculating fuzzy membership degrees based on multiple sets of operating condition feature vector samples from multiple pre-built operating condition modes and the current operating condition feature vector. The multiple pre-built operating condition modes include at least a shutdown operating condition mode, a low-load operating condition mode, a normal operating condition mode, a high-load operating condition mode, and an overload operating condition mode. Specifically, five basic operating condition modes are pre-built, and a corresponding operating condition feature vector sample library is established for each mode. Each sample library contains multiple sets of operating condition feature vector samples, and the feature dimensions of the samples are consistent with the current operating condition feature vector. Fuzzy membership degrees are calculated between the current operating condition feature vector and the sample vectors of each mode. The calculation method used is the cosine similarity method, and the formula is: the membership degree value is equal to the dot product of the two vectors divided by the product of the magnitudes of the two vectors. The membership values ​​of the five modes are normalized, and the operating mode corresponding to the maximum normalized value is the current operating mode of the equipment. A fixed initial structured weight vector is preset for each operating mode. The weight values ​​correspond to the three types of contributions of the three-stage structured state features. The weight allocation is based on the degree of influence of the three types of contributions on oil quantity changes under different operating conditions. For example, the contribution of operating consumption under shutdown conditions is extremely low, so the weight value of the third type of contribution is small. The vector corresponding to the current operating condition is the initial structured weight vector.

[0041] Step S320: Calculate the deviation of the working condition mode based on the fuzzy membership degree, and adaptively correct the initial structured weight vector according to the working condition mode deviation to obtain an adaptive structured weight vector. Specifically, calculate the working condition mode deviation, which is a quantitative value representing the degree of difference between the current working condition feature vector and the sample vector of the corresponding pre-constructed working condition mode. The formula is: Deviation = 1 - Fuzzy membership value of the current working condition mode, with a value range of 0 to 1. The closer the value is to 0, the more closely the current working condition fits the pre-constructed mode; the closer the value is to 1, the greater the difference. Set the basic rules for weight correction: the correction coefficient for each weight value = 1 - deviation × correction magnitude coefficient. The correction rule is: the corrected weight value = the initial weight value × the correction coefficient, and the sum of the three corrected weight values ​​needs to be normalized back to 1. The core logic of the correction is: if the deviation is small, the weight value is finely adjusted to approach the initial value; if the deviation is large, the weight value is significantly adjusted to adapt to the actual working condition. After completing the correction and normalization, output the adaptive structured weight vector.

[0042] In one possible implementation, the three-stage structured state features are processed based on the adaptive structured weight vector to extract the oil volume change morphological fingerprint vector. Step S300 further includes step S330, performing high-dimensional feature convolution on the three-stage structured state features to obtain high-dimensional features of the three-stage structured state in multiple dimensions. Specifically, high-dimensional feature convolution refers to the operation of expanding the feature dimension of the three-dimensional three-stage structured state features using a one-dimensional convolution kernel. The purpose of convolution is to map low-dimensional basic features to high-dimensional features containing more detailed information. The expanded feature dimension is consistent with the dimension of the equipment's operating parameters, enabling dimensional matching between oil volume change features and operating condition features. The three-dimensional three-stage structured state features are used as convolution input, and convolution operation is performed. The operation rule is: each dimension value of the input feature is multiplied pairwise with each weight value of the convolution kernel, and then all multiplication results are arranged in order and combined into a new high-dimensional feature vector.

[0043] Step S340: Expand the adaptive structured weight vector into an adaptive structured weight high-dimensional vector across the multiple dimensions. Specifically, the three-dimensional adaptive structured weight vector is synchronously expanded into a high-dimensional weight vector of the same dimension according to the dimensionality expansion rules of high-dimensional features. The core rule of expansion is: copy each weight value of the original vector to a dimension with the same number of dimensions as the convolution kernel size. The dimension of the expanded weight vector is completely consistent with the dimension of the high-dimensional features of the three-stage structured state.

[0044] Step S350: Perform feature weighted fusion on the three-stage structured state high-dimensional features according to the adaptive structured weight high-dimensional vector to obtain state fusion features. Specifically, feature weighted fusion refers to multiplying each dimension value of the high-dimensional feature with the corresponding high-dimensional weight value to obtain a weighted feature value. All weighted feature values ​​arranged in order constitute the state fusion features. The core logic of weighting is that the larger the weight value of a dimension, the greater the influence of the corresponding feature value on the overall oil volume change; the smaller the weight value of a dimension, the lower the influence. The state fusion features are a high-dimensional feature set that integrates the operating condition weights and oil volume change features, eliminating the interference of operating condition differences.

[0045] Step S360: Extract the first-order statistical feature vector, trend feature vector, frequency domain feature vector, and nonlinear feature vector from the state fusion features to form the oil volume change morphology fingerprint vector. Specifically, the first-order statistical feature vector is a feature set obtained by performing basic statistical calculations on the state fusion features, including core statistical quantities such as mean, variance, maximum, minimum, and median, representing the overall distribution law of the features. The trend feature vector is a feature set obtained by linear fitting the state fusion features using the least squares method, including fitting slope, intercept, and goodness of fit, representing the changing trend law of the features. The frequency domain feature vector is a feature set obtained by performing a fast Fourier transform on the state fusion features, including dominant frequency, secondary dominant frequency, and spectral energy, representing the frequency distribution law of the features. The nonlinear feature vector is a feature set obtained by performing nonlinear analysis on the state fusion features, including fractal dimension, chaotic eigenvalues, and entropy, representing the nonlinear change law of the features. The oil volume change morphology fingerprint vector is a high-dimensional feature vector obtained by combining and normalizing the above four types of feature vectors in a fixed order. This vector is a unique representation of the oil volume change law.

[0046] Step S400: The oil volume change pattern fingerprint vector is matched with the oil volume change trend template library to obtain a matching oil volume change trend template. Based on the matching oil volume change trend template, the oil volume change trend prediction of the oil volume status dataset is performed.

[0047] Specifically, a pre-built template library of oil volume change trends is constructed. This library is based on historical oil volume status data and corresponding historical operating condition data, with each template representing a specific oil volume change trend with a fixed pattern. For the predicted oil volume change pattern fingerprint vector, a multi-dimensional similarity calculation method is used to calculate feature space similarity, trend pattern similarity, and operating condition matching degree. The three similarity scores are then weighted and summed to obtain a comprehensive matching score. A fixed preset score threshold is set, and templates with comprehensive scores higher than the threshold are selected as matching templates. Depending on the number of matching templates, either direct invocation or template fusion is chosen. Based on the trend prediction parameter set within the template, key indicators such as subsequent oil volume change values, oil volume change rate, oil depletion time, and the optimal time for the next oil injection are quantitatively calculated for the oil volume status dataset. The specific prediction results are output, including quantified oil volume change values ​​and corresponding confidence levels.

[0048] In one possible implementation, the oil volume change pattern fingerprint vector is matched with an oil volume change trend template library to obtain a matching oil volume change trend template. Step S400 further includes step S410, constructing an oil volume change trend template library. Each oil volume change trend template in the library includes a template pattern fingerprint vector, a trend prediction parameter set, a trend type, a template confidence level, and an applicable operating condition mode label. Specifically, historical operating data of similar oil injection actuators are collected, covering all operating conditions including shutdown, low load, normal, high load, and overload. All historical data undergo the same processing flow as S100 to S360, sequentially completing the three-stage decomposition of the oil volume status dataset, operating condition mode matching, adaptive weighted fusion, and oil volume change pattern fingerprint vector extraction to obtain historical oil volume pattern fingerprint vectors. K-means clustering algorithm is applied to the historical oil volume pattern fingerprint vector to perform feature clustering. The number of clusters is set according to the complexity of the working condition type and oil volume change pattern. Each cluster result corresponds to a fixed oil volume change trend, which is an oil volume change trend template. For each template type, five core components are standardized and configured: ① Template morphological fingerprint vector: The mean vector of all fingerprint vectors in the clustering results of this type is taken, and its dimension is completely consistent with the fingerprint vector of the oil volume change morphology to be predicted; ② Trend prediction parameter set: A fixed structured parameter set, which includes at least the benchmark value of oil consumption per unit time, oil volume decay coefficient, oil volume critical warning value, calculation coefficient of the optimal time for the next oil injection, and fitting coefficient of oil volume depletion time. All parameters are quantitative values ​​based on historical data statistics; ③ Trend type: A qualitative classification label, divided into five categories: uniform consumption type, slow decay type, fluctuating consumption type, rapid loss type, and extremely rapid depletion type, corresponding to the rate and pattern of oil volume change; ④ Template confidence: Based on the historical prediction accuracy statistics of this type of template, the calculation rule is the average degree of consistency between the historical prediction results of this template and the actual oil volume change; ⑤ Applicable operating condition mode label: A single or multiple operating condition modes matched by this template. The label corresponds one-to-one with the pre-built shutdown, low load, normal, high load, and overload operating condition modes. After all templates are configured, they are categorized and stored according to operating condition tags to form a structured oil volume change trend template library.

[0049] Step S420: Based on the oil volume change trend template library, calculate the feature space similarity, trend pattern similarity, and operating condition matching degree of the oil volume change pattern fingerprint vector to obtain a comprehensive score. Specifically, set the calculation rules and weighting weights for the three types of similarity. The weighting weights are fixed empirical values, and the sum of the weights is 1. Perform quantitative calculations of the three types of similarity in sequence. The first is feature space similarity, calculated using the cosine similarity algorithm. The calculation formula is the dot product of the fingerprint vector to be predicted and the template pattern fingerprint vector, divided by the product of the magnitudes of the two vectors. The larger the value, the higher the feature matching degree. The second is trend pattern similarity. Extract the trend feature vector dimension from the two fingerprint vectors, including slope, intercept, goodness of fit, etc., calculate the Euclidean distance of the sub-vector, and then perform inverse normalization on the distance value to obtain the similarity value. The smaller the Euclidean distance, the larger the similarity value, which represents a more consistent trend of oil volume change. Third is the operating condition matching degree, which is a qualitative-to-quantitative calculation value. If the current operating condition mode of the equipment to be predicted is completely consistent with the applicable operating condition mode label of the template, the value is 1; if it is an adjacent operating condition mode, such as low load and normal, or normal and high load, the value is 0.8; if it is a cross-level operating condition mode, such as low load and high load, or shutdown and overload, the value is 0.5; if it is completely unrelated, the value is 0. Finally, the comprehensive score is calculated. The calculation formula is: Comprehensive Score = Feature Space Similarity × Feature Space Similarity Weight + Trend Pattern Similarity × Trend Pattern Similarity Weight + Operating Condition Matching Degree × Operating Condition Matching Degree Weight. The higher the value, the higher the matching degree between the oil volume change pattern to be predicted and the template.

[0050] Step S430: Filter matching oil volume change trend templates with a comprehensive score greater than a preset score threshold. Specifically, the preset score threshold is a pre-set quantitative critical value used to determine whether a template is a valid match. The value of this threshold is determined based on the overall matching accuracy statistics of the template library. Iterate through all templates in the oil volume change trend template library, comparing the comprehensive score of each template with the preset score threshold. Filter out all templates with a comprehensive score greater than the preset score threshold to form a matching oil volume change trend template set. Then, sort the templates in this set in descending order of comprehensive score, with the priority of the sorted templates decreasing sequentially. If the filtered matching template set is empty, trigger the template library addition process, entering the current oil volume change pattern fingerprint vector and the corresponding oil volume change rule as a new template into the library. If a matching template is found after filtering, proceed directly to the prediction stage.

[0051] In one possible implementation, the matching oil volume change trend templates that are greater than a preset score threshold are selected according to the comprehensive score. Step S400 further includes step S440: if multiple matching oil volume change trend templates are returned, template fusion is performed on the returned multiple matching oil volume change trend templates to obtain a fused oil volume change trend template. Specifically, a weighting rule for template fusion is determined, and the weight value is positively correlated with the comprehensive score of the matching template. The calculation rule is: fusion weight of a single template = comprehensive score of that template ÷ sum of comprehensive scores of all matching templates, and the sum of fusion weights of all templates is 1. The five core components of the matching template are sequentially weighted and fused. The fusion rules for each component are as follows: ① Template morphological fingerprint vector: The fingerprint vectors of each template are weighted and summed according to the fusion weight to obtain the fused mean fingerprint vector, with the same dimension as the original vector; ② Trend prediction parameter set: The parameters of the same type for each template are weighted and summed. For example, the fused unit time fuel consumption baseline value = consumption value of template 1 × weight of template 1 + consumption value of template 2 × weight of template 2. All prediction parameters are fused according to this rule to obtain a complete fused trend prediction parameter set; ③ Trend type: The trend type of the template with the highest comprehensive score is taken as the fused trend type; ④ Template confidence: The confidence of each template is weighted and summed according to the fusion weight to obtain the fused confidence value; ⑤ Applicable operating condition mode label: The operating condition labels of all matching templates are integrated to form the fused operating condition coverage label. The five fused components are integrated according to a standardized structure to obtain a complete fused fuel change trend template. All parameters of this template are the optimal values ​​after weighting, which are adapted to the current compound fuel change pattern.

[0052] Step S450: Predict the oil volume change trend of the oil volume status dataset based on the fused oil volume change trend template. Specifically, if the matching template obtained in S430 is a single template, prediction is performed directly based on that single template; if there are multiple templates, prediction is performed based on the fused oil volume change trend template from S440. The core logic is to use the core configuration items of the template as the calculation basis to quantify the subsequent oil volume change pattern of the oil volume status dataset and output the results. The matching single template or the fused oil volume change trend template is retrieved, and the core calculation basis within the template is extracted: the trend prediction parameter set and the template confidence level. The indicators for predicting oil quantity change trends are set. All indicators are quantitatively calculated based on the trend prediction parameter set. The calculation rules are as follows: Remaining oil quantity for a future preset time period = Current oil quantity - Baseline oil consumption per unit time × Preset time × Oil quantity decay coefficient. The preset time can be set to a fixed duration such as 1h, 2h, 8h, 24h, etc. The oil consumption rate change curve is a continuous change curve obtained by fitting the oil quantity decay coefficient with time as the horizontal axis and oil consumption rate as the vertical axis, representing the evolution law of oil consumption rate. Critical oil quantity warning trigger time = (Current oil quantity - Critical oil quantity warning value) ÷ (Baseline oil consumption per unit time × Oil quantity decay coefficient). Next optimal refueling time = Critical oil quantity warning trigger time × Next optimal refueling time calculation coefficient. The coefficient is an empirical value greater than 1 to reserve sufficient refueling buffer time and avoid oil depletion. Complete oil depletion time = Current oil quantity ÷ (Baseline oil consumption per unit time × Oil quantity decay coefficient) × Oil depletion time fitting coefficient. The fitting coefficient is used to correct the deviation between the theoretical calculation value and the actual value. For all calculated prediction indicators, the confidence level of the template is attached as a reference for the accuracy of the prediction results, forming a structured prediction result of oil volume change trend, which is used for intelligent control of automatic oil injection device.

[0053] This application embodiment extracts the oil quantity status dataset of the oil injection actuator, and obtains structured status features composed of three types of oil quantity change contributions: oil injection input, structural penetration, and operational consumption through a three-stage decomposition. It then obtains the current operating mode of the equipment and introduces an adaptive structured weight vector to process the structured status features, extracting an oil quantity change pattern fingerprint vector. This fingerprint vector is then matched with an oil quantity change trend template library for similarity. Based on the matched trend template, oil quantity change trend prediction is performed. These techniques solve the technical problem of low prediction accuracy in existing automatic oil injection devices, achieving the technical effect of improving the prediction accuracy of oil quantity change trends.

[0054] In the above text, refer to Figure 1 A method for predicting the oil quantity change trend of an automatic oiling device according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2This invention describes a system for predicting oil quantity change trends in an automatic oiling device according to an embodiment of the present invention.

[0055] An oil quantity change trend prediction system for an automatic oil filling device, according to an embodiment of the present invention, addresses the technical problem of low prediction accuracy in existing automatic oil filling device oil quantity change trend prediction, thereby improving the accuracy of oil quantity change trend prediction. The oil quantity change trend prediction system for an automatic oil filling device includes: an oil quantity state decomposition module 10, an equipment operating condition mode acquisition module 20, an oil quantity change pattern fingerprint vector extraction module 30, and an oil quantity change trend prediction module 40.

[0056] The oil quantity status decomposition module 10 is used to extract the oil quantity status dataset of the oil injection actuator, perform a three-stage decomposition on the oil quantity status dataset, and obtain three-stage structured status features. The three-stage structured status features include a first type of oil quantity change contribution based on oil injection input, a second type of oil quantity change contribution based on structural penetration, and a third type of oil quantity change contribution based on operating consumption. The equipment operating condition mode acquisition module 20 is used to acquire the current operating condition mode of the oil injection actuator. The oil quantity change morphology fingerprint vector extraction module 30 is used to introduce an adaptive structured weight vector according to the equipment operating condition mode, process the three-stage structured status features based on the adaptive structured weight vector, and extract the oil quantity change morphology fingerprint vector. The oil quantity change trend prediction module 40 is used to perform similarity matching between the oil quantity change morphology fingerprint vector and the oil quantity change trend template library to obtain a matching oil quantity change trend template, and perform oil quantity change trend prediction of the oil quantity status dataset according to the matching oil quantity change trend template.

[0057] The detailed configuration of the oil quantity state decomposition module 10 is explained as follows: As described above, the oil quantity state dataset is decomposed into three stages to obtain three-stage structured state features. The oil quantity state decomposition module 10 may further include: a time series processing unit for performing time series processing on the oil quantity state dataset to obtain oil quantity state time series data; a first-order difference sequence data calculation unit for calculating the first-order difference sequence data of the oil quantity state time series data; an oil injection event time point set acquisition unit for setting an oil injection event detection change rate threshold, marking the corresponding time series nodes in the first-order difference sequence data that are greater than the oil injection event detection change rate threshold as the oil injection event occurrence time, and recording them to obtain an oil injection event time point set; and a three-stage decomposition unit for performing three-stage decomposition on each oil injection event in the oil injection event time point set to obtain three-stage structured state features.

[0058] Specifically, each oil injection event in the set of oil injection event time points is decomposed into three stages to obtain three-stage structured state features and to obtain the first type of oil quantity change contribution. The three-stage decomposition unit may further include: a pulse integration subunit for performing pulse integration on the oil injection input amount of each oil injection event with respect to the oil quantity state data to obtain oil injection input integral data; an oil injection instantaneous effective coefficient calculation subunit for calculating the oil injection instantaneous effective coefficient on the oil injection input integral data through a monotonically bounded mapping function to obtain the oil injection instantaneous effective coefficient; and a first type of oil quantity change contribution output subunit for outputting the first type of oil quantity change contribution based on the product of the oil injection instantaneous effective coefficient and the oil injection input integral data.

[0059] The three-stage decomposition unit for obtaining the second type of oil volume change contribution can further include: a penetration time point determination subunit for determining the penetration start time point and penetration end time point for each injection event; a double exponential decay model establishment subunit for establishing a double exponential decay model, wherein the double exponential decay model calculates the decay of the fast penetration component and the decay of the slow penetration component, and the double exponential decay model obtains integral data based on the penetration start time point and penetration end time point to obtain injection penetration integral data; a viscosity influence factor setting subunit for setting a viscosity influence factor according to the viscosity of the injection medium; and a second type of oil volume change contribution acquisition subunit for normalizing the injection penetration integral data according to the viscosity influence factor to obtain the second type of oil volume change contribution.

[0060] The three-stage decomposition unit for obtaining the third type of oil quantity change contribution can further include: a unit for calculating oil quantity consumption per unit time for each oil injection event based on oil quantity status data; and a unit for obtaining the third type of oil quantity change contribution by collecting the temperature status of the oil injection actuator to optimize the oil quantity consumption, and normalizing the optimized oil quantity consumption as a negative oil quantity change output to obtain the third type of oil quantity change contribution.

[0061] The detailed description of the specific configuration of the oil quantity change pattern fingerprint vector extraction module 30 is as follows: As mentioned above, based on the equipment operating condition mode, an adaptive structured weight vector is introduced. The oil quantity change pattern fingerprint vector extraction module 30 may further include: an initial structured weight vector acquisition unit for acquiring an initial structured weight vector based on the equipment operating condition mode; wherein, the equipment operating condition mode includes acquiring the current operating condition feature vector, and obtaining the current operating condition feature vector by performing fuzzy membership calculation based on multiple sets of operating condition feature vector samples of multiple pre-built operating condition modes and the current operating condition feature vector; wherein, the multiple pre-built operating condition modes include at least a shutdown operating condition mode, a low load operating condition mode, a normal operating condition mode, a high load operating condition mode, and an overload operating condition mode; an adaptive correction unit is used to calculate the operating condition mode deviation degree according to the fuzzy membership degree, and adaptively correct the initial structured weight vector according to the operating condition mode deviation degree to obtain an adaptive structured weight vector.

[0062] Specifically, the oil volume change morphology fingerprint vector extraction module 30 further includes: a high-dimensional feature convolution unit for performing high-dimensional feature convolution on the three-stage structured state features to obtain high-dimensional features of the three-stage structured state in multiple dimensions; an adaptive structured weight high-dimensional vector generation unit for expanding the adaptive structured weight vector into an adaptive structured weight high-dimensional vector in the multiple dimensions; a feature weighted fusion unit for performing feature weighted fusion on the high-dimensional features of the three-stage structured state according to the adaptive structured weight high-dimensional vector to obtain state fusion features; and an oil volume change morphology fingerprint vector generation unit for extracting the first-order statistical feature vector, trend feature vector, frequency domain feature vector, and nonlinear feature vector of the state fusion features to form the oil volume change morphology fingerprint vector.

[0063] The detailed description of the specific configuration of the oil volume change trend prediction module 40 is explained as follows: As mentioned above, the oil volume change pattern fingerprint vector is matched with the oil volume change trend template library to obtain a matching oil volume change trend template. The oil volume change trend prediction module 40 may further include: an oil volume change trend template library construction unit for constructing an oil volume change trend template library, wherein each oil volume change trend template in the oil volume change trend template library includes a template pattern fingerprint vector, a trend prediction parameter set, a trend type, a template confidence level, and an applicable operating condition mode label; a comprehensive score calculation unit for calculating the feature space similarity, trend pattern similarity, and operating condition matching degree of the oil volume change pattern fingerprint vector according to the oil volume change trend template library to obtain a comprehensive score; and a matching oil volume change trend template filtering unit for filtering matching oil volume change trend templates that are greater than a preset score threshold according to the comprehensive score.

[0064] The oil volume change trend prediction module 40, which filters matching oil volume change trend templates that are greater than a preset score threshold according to the comprehensive score, may further include: a template fusion unit, which, if the number of matching oil volume change trend templates returned is multiple, performs template fusion on the returned multiple matching oil volume change trend templates to obtain a fused oil volume change trend template; and an oil volume change trend prediction unit, which performs oil volume change trend prediction on the oil volume status dataset based on the fused oil volume change trend template.

[0065] The oil quantity change trend prediction system for automatic oil filling devices provided in this embodiment of the invention can execute the oil quantity change trend prediction method for automatic oil filling devices provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0066] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0067] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for predicting the oil quantity change trend of an automatic oiling device, characterized in that, The method includes: Extract the oil quantity status dataset of the oil injection actuator, perform a three-stage decomposition on the oil quantity status dataset, and obtain three-stage structured status features. The three-stage structured status features include a first type of oil quantity change contribution based on oil injection input, a second type of oil quantity change contribution based on structural penetration, and a third type of oil quantity change contribution based on operating consumption. Obtain the current operating mode of the equipment being operated by the oil injection actuator; An adaptive structured weight vector is introduced based on the equipment operating mode. The three-stage structured state features are processed based on the adaptive structured weight vector to extract the oil quantity change pattern fingerprint vector. The oil volume change pattern fingerprint vector is matched with the oil volume change trend template library to obtain a matching oil volume change trend template. Based on the matching oil volume change trend template, the oil volume change trend prediction of the oil volume status dataset is performed.

2. The method for predicting oil quantity change trends in an automatic oiling device as described in claim 1, characterized in that, The oil quantity state dataset is decomposed into three stages to obtain three-stage structured state features. The method includes: The oil quantity status dataset is subjected to time series processing to obtain oil quantity status time series data; Calculate the first-order difference sequence data of the oil quantity state time series data; Set a threshold for the rate of change of oil injection event detection, mark the corresponding time series nodes in the first-order difference sequence data that are greater than the threshold for the rate of change of oil injection event detection as the time when the oil injection event occurs, and record the set of time points of oil injection event. Each oil injection event in the set of oil injection event time points is decomposed into three stages to obtain three-stage structured state features.

3. The method for predicting oil quantity change trends in an automatic oiling device as described in claim 2, characterized in that, The method for obtaining the contribution of the first type of oil volume change to the three-stage decomposition of each oil injection event in the set of oil injection event time points and obtaining the three-stage structured state features includes: For each oil injection event, the oil injection input quantity related to the oil quantity status data is integrated by pulse integration to obtain the oil injection input integral data. The immediate effective coefficient of oil injection is obtained by calculating the oil injection effective coefficient on the oil injection input integral data through a monotonically bounded mapping function. Based on the product of the instantaneous effective coefficient of oil injection and the integral data of oil injection input, the contribution of the first type of oil quantity change is output.

4. The method for predicting oil quantity change trends in an automatic oiling device as described in claim 3, characterized in that, Methods for obtaining the contribution of the second type of oil volume change include: Determine the initiation and termination time of penetration for each oil injection event; A double exponential decay model is established, which calculates the decay of the fast permeation component and the slow permeation component. The double exponential decay model obtains integral data based on the permeation start time point and the permeation end time point to obtain oil injection permeation integral data. Set the viscosity influence factor according to the viscosity of the injection medium; The oil injection penetration integral data is normalized according to the viscosity influence factor to obtain the contribution of the second type of oil volume change.

5. The method for predicting the oil quantity change trend of an automatic oiling device as described in claim 3, characterized in that, Methods for obtaining the contribution of the third type of oil volume change include: For each oil injection event, calculate the oil consumption per unit time based on the oil quantity status data. The temperature status of the oil injection actuator is collected to optimize the oil consumption. The optimized oil consumption is normalized as a negative oil change and output to obtain the third type of oil change contribution.

6. The method for predicting oil quantity change trends for an automatic oiling device as described in claim 1, characterized in that, An adaptive structured weight vector is introduced based on the equipment operating mode, the method of which includes: The initial structured weight vector is obtained based on the equipment operating mode; The equipment operating condition mode includes obtaining the current operating condition feature vector and calculating the fuzzy membership degree based on multiple sets of operating condition feature vector samples from multiple pre-built operating condition modes and the current operating condition feature vector. The pre-built operating mode includes at least a shutdown operating mode, a low load operating mode, a normal operating mode, a high load operating mode, and an overload operating mode. The deviation of the working condition mode is calculated based on the fuzzy membership degree, and the initial structured weight vector is adaptively corrected according to the deviation of the working condition mode to obtain an adaptive structured weight vector.

7. The method for predicting oil quantity change trends in an automatic oiling device as described in claim 6, characterized in that, The three-stage structured state features are processed based on the adaptive structured weight vector to extract the oil volume change morphological fingerprint vector. The method includes: High-dimensional feature convolution is performed on the three-stage structured state features to obtain high-dimensional features of the three-stage structured state in multiple dimensions; The adaptive structured weight vector is extended into an adaptive structured weight high-dimensional vector under the multiple dimensions; The three-stage structured state high-dimensional features are weighted and fused according to the adaptive structured weight high-dimensional vector to obtain state fusion features; The first-order statistical feature vector, trend feature vector, frequency domain feature vector, and nonlinear feature vector of the state fusion feature are extracted to form the oil volume change morphological fingerprint vector.

8. The method for predicting oil quantity change trends for an automatic oiling device as described in claim 1, characterized in that, The method involves performing similarity matching between the oil volume change pattern fingerprint vector and the oil volume change trend template library to obtain a matching oil volume change trend template. Construct an oil volume change trend template library. Each oil volume change trend template in the oil volume change trend template library includes a template morphological fingerprint vector, a trend prediction parameter set, a trend type, a template confidence level, and an applicable operating condition mode label. Based on the oil volume change trend template library, the feature space similarity, trend pattern similarity, and operating condition matching degree of the oil volume change pattern fingerprint vector are calculated to obtain a comprehensive score; Based on the comprehensive score, templates matching the oil volume change trend that are greater than the preset score threshold are selected.

9. The method for predicting the oil quantity change trend of an automatic oiling device as described in claim 8, characterized in that, The method further includes filtering matching oil volume change trend templates that are greater than a preset score threshold based on the comprehensive score. If the number of matching oil volume change trend templates returned is multiple, the multiple returned matching oil volume change trend templates are merged to obtain a merged oil volume change trend template. Based on the fused oil volume change trend template, the oil volume change trend of the oil volume status dataset is predicted.

10. A system for predicting the oil quantity change trend of an automatic oiling device, characterized in that, The system is used to implement the oil quantity change trend prediction method for an automatic oil injection device according to any one of claims 1-9, the system comprising: The oil quantity status decomposition module is used to extract the oil quantity status dataset of the oil injection actuator, perform a three-stage decomposition on the oil quantity status dataset, and obtain three-stage structured status features. The three-stage structured status features include a first type of oil quantity change contribution based on oil injection input, a second type of oil quantity change contribution based on structural penetration, and a third type of oil quantity change contribution based on operating consumption. The equipment operating mode acquisition module is used to acquire the current operating mode of the oil injection actuator. The oil quantity change pattern fingerprint vector extraction module is used to introduce an adaptive structured weight vector according to the equipment operating mode, process the three-stage structured state features based on the adaptive structured weight vector, and extract the oil quantity change pattern fingerprint vector. The oil volume change trend prediction module is used to perform similarity matching between the oil volume change pattern fingerprint vector and the oil volume change trend template library to obtain a matching oil volume change trend template, and to perform oil volume change trend prediction of the oil volume status dataset based on the matching oil volume change trend template.

Images