Method and system for controlling the opening of pressure line filter bypass

Abstract

This invention belongs to the field of bypass control, and particularly relates to a method and system for opening control of bypasses in pressure pipeline filters. This method constructs a multi-parameter correlation mapping space of temperature, cleanliness, flow rate, and pressure, and generates precise bypass opening and closing time points and opening / closing commands by combining a pressure prediction model. It uses fluid viscosity correlation curves, blockage degree correlation mapping, and a triangular pressure difference model to dynamically calculate three levels of reaction delay, achieving pre-time compensation control. Based on an enhanced prediction model, it triggers multi-level execution actions: when the predicted pressure difference exceeds the limit, a pressure relief rule is activated, and the bypass opening / closing degree is adjusted through real-time monitoring feedback; when the pressure difference still rises even when the bypass is fully open, a fault warning is triggered. It innovatively couples pressure-level control with cleanliness optimization objectives, solving for the optimal pressure relief efficiency through a linear mapping algorithm, ensuring pressure difference reduction while maintaining fluid cleanliness. This method significantly improves the accuracy and reliability of bypass control.

Description

Technical Field

[0001] This invention belongs to the field of bypass control, and particularly relates to a method and system for opening control of pressure line filter bypass. Background Technology

[0002] In pressure piping systems, filters are used to remove impurities from fluids, ensuring the normal operation of equipment within the system. When a filter becomes clogged, causing the pressure difference across it to exceed a certain range, a bypass needs to be opened to prevent excessive system pressure from damaging the equipment. This allows the fluid to bypass the filter and flow directly through. Traditional bypass opening control methods primarily use simple differential pressure switches, which directly open the bypass when the pressure difference reaches a set value. This method has low control precision and is prone to opening the bypass too early or too late. Opening too early results in insufficient filtration of the fluid, affecting equipment operation; opening too late may damage the filter or other equipment.

[0003] The technical problem to be solved: Existing pressure line filter bypass opening control methods have low precision and cannot reasonably control the bypass opening timing according to the actual system operating conditions, which can easily damage system equipment or affect fluid filtration efficiency. Simple differential pressure switch control methods cannot adapt to changes in filter clogging under different operating conditions and lack flexibility and intelligence; therefore, this application proposes an opening control method and system for pressure line filter bypasses. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a method and system for controlling the opening of bypasses in pressure pipeline filters. This method constructs a multi-parameter correlation mapping space for the target device's temperature, fluid cleanliness, flow rate, and pressure. Combined with a pressure change prediction sub-model pre-trained using an LSTM algorithm, it generates bypass opening and closing time points and graded control commands for the opening degree. Utilizing fluid viscosity correlation curves, blockage degree correlation mappings, and a triangular pressure difference correlation model, it dynamically calculates the delays of the first, second, and third reaction mappings, achieving pre-time point compensation control. It employs an enhanced prediction model to trigger multi-level execution actions, including triggering pressure relief rules based on predicted pressure difference exceeding limits, adjusting the bypass opening degree through real-time monitoring data feedback, and initiating a fault warning when pressure difference still rises after the bypass is fully opened. Innovatively, it couples pressure graded control rules with cleanliness optimization objectives, using a linear mapping algorithm to solve for the optimal pressure relief efficiency, ensuring that fluid cleanliness is maintained while reducing pressure difference. This method significantly improves the accuracy of bypass opening degree and timing.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The method for controlling the opening of a pressure line filter bypass includes the following steps:

[0007] An enhanced predictive control model is constructed based on a preset target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space combined with a preset pressure graded control rule, generating the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass.

[0008] Real-time bypass control is performed based on the opening and closing time points and opening and closing degree control commands corresponding to the generated target device bypass, and the controlled fluid cleanliness and target device pressure change values ​​are fed back to the associated mapping space to adjust the opening and closing time points and opening and closing degree in real time.

[0009] Specifically, the process of constructing the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space includes:

[0010] The fluid temperature, filtration flow rate, particulate matter content per unit volume of fluid, pressure difference before and after the target device, and bypass pressure difference are obtained and preprocessed to obtain the pressure relief control data sequence of the target device;

[0011] A fluid cleanliness index is established based on the particulate matter content per unit volume of fluid in the pressure relief control data sequence of the target device.

[0012] Based on fluid cleanliness index, fluid temperature and fluid viscosity, a fluid viscosity co-variance curve and viscosity coefficient scaling function are established.

[0013] The first reaction mapping delay is obtained by monitoring the time point of change in fluid temperature or fluid cleanliness index and the time point of change in the fluid viscosity-coordinated change curve.

[0014] Specifically, the process of constructing the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space also includes:

[0015] Based on the cross-correlation coefficients between viscosity coefficient and filtration flow rate, a target device clogging degree index is established, and a clogging degree correlation mapping curve is obtained.

[0016] The second reaction mapping delay is obtained by mapping the change in viscosity coefficient value or filtration flow rate to the change in clogging degree reaction time point of the correlation curve.

[0017] Specifically, the process of constructing the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space also includes:

[0018] Based on the target device blockage degree index, the target device front and rear pressure difference value and bypass pressure difference value, and the sequence of target device front and rear pressure difference changes under different historical bypass opening degrees, the triangular correlation mapping and corresponding curves of blockage degree, target device front and rear pressure difference and bypass pressure difference are obtained.

[0019] The triangular correlation mapping includes the first correlation mapping curve between the degree of blockage and the pressure difference across the target device, the second correlation mapping curve between the degree of blockage and the bypass pressure difference, and the third correlation mapping curve between the bypass opening degree and the change in the pressure difference across the target device.

[0020] The third reaction mapping delay is obtained based on the time point from the change in congestion level to the time point of bypass opening.

[0021] Specifically, the process of constructing the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space also includes:

[0022] Based on the fluid viscosity co-variance curve, the blockage degree correlation mapping curve and the triangular correlation mapping and the corresponding curve, combined with the first reaction mapping delay, the second reaction mapping delay or the third reaction mapping delay, the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space is constructed through a multi-parameter coupling algorithm.

[0023] Specifically, the process of building a enhanced predictive control model includes:

[0024] The input state of the enhanced predictive control model is constructed based on the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space, the pressure difference before and after the target device, the pressure difference threshold before and after the target device, the maximum upper limit of bypass shunting, and the preset pressure classification control rules.

[0025] Based on the constructed input state, the preset bypass opening and closing action trigger information is as follows:

[0026] If the pressure change prediction sub-model in the current enhanced predictive control model predicts that the change in the pressure difference across the target device at a certain future moment is greater than the preset threshold for the pressure difference across the target device, the first execution action information is triggered.

[0027] That is, based on the preset target device front and rear pressure difference threshold, pressure increase change value, and bypass pressure relief efficiency per unit time, the pressure relief rule of the corresponding level in the pressure classification control rule is triggered, and the bypass pressure relief opening degree is obtained by inversion.

[0028] Specifically, the process of building enhanced predictive control models also includes:

[0029] When the first execution action information is triggered, the second execution action information is also triggered simultaneously;

[0030] That is, the time point when the pressure difference change value before and after the target device is greater than the preset pressure difference threshold before and after the target device is predicted by the pressure change prediction sub-model, combined with the first reaction mapping delay, the second reaction mapping delay or the third reaction mapping delay, is used to calculate the corresponding pre-time point for issuing the bypass opening and closing command.

[0031] The pressure change prediction sub-model is obtained by pre-training with the LSTM algorithm using fluid temperature, filtration velocity, particulate matter content per unit volume of fluid, and pressure difference before and after the target device.

[0032] When the second execution action information is triggered, the third execution action information is triggered simultaneously, that is, real-time monitoring of the pressure difference before and after the target device, the bypass pressure difference, and the fluid cleanliness after the bypass opening command under the corresponding pressure relief level is issued at the corresponding pre-set time point.

[0033] Specifically, the process of building enhanced predictive control models also includes:

[0034] When the third execution action information is triggered, if the pressure difference before and after the target device is still increasing while the bypass pressure difference is decreasing, the fourth execution action information is triggered. That is, based on the pressure difference increase obtained from the second monitoring of the target device, the first and second execution action information are triggered for the second time. The pressure relief opening degree of the bypass is gradually adjusted through the third correlation mapping curve so that the pressure difference before and after the target device is decreasing while the total cleanliness of the target device and the bypass filtered fluid is lower than the preset fluid cleanliness standard value.

[0035] If the bypass opening level reaches its maximum and the voltage difference across the target device is still increasing, the bypass is determined to be in an abnormal or faulty state, and a feedback warning is issued.

[0036] Specifically, the process of constructing the third correlation mapping curve includes:

[0037] Based on the monitored pressure difference changes before and after the target device and the pressure relief efficiency and pressure classification control rules corresponding to different opening and closing degrees of the bypass, the optimal solution is obtained through a linear coupling mapping algorithm. This maximizes the cleanliness of the target device and the fluid filtered by the bypass under the premise of satisfying the pressure relief of the target device, and obtains the classification control opening and closing level-pressure relief efficiency mapping function and the third correlation mapping curve.

[0038] The pressure grading control rule is constructed by mapping the pressure level obtained through evaluation to the preset bypass opening degree.

[0039] The control system for opening the bypass of the pressure line filter includes: an enhanced control module and a control feedback module;

[0040] The enhanced control module, based on the preset target device temperature-fluid cleanliness-flow-pressure correlation mapping space and the preset pressure graded control rules, constructs an enhanced predictive control model to generate the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass.

[0041] The control feedback module performs real-time bypass control based on the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass, and feeds back the controlled fluid cleanliness and target device pressure change values ​​to the associated mapping space to adjust the opening and closing time points and opening and closing degree in real time.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] This invention addresses the shortcomings of existing technologies by constructing a multi-parameter correlation mapping space and pressure-level control rules, combined with LSTM prediction models for time-series modeling of differential pressure changes, significantly improving the dynamic adaptability of bypass control. A dynamic delay compensation mechanism based on the correlation mapping between fluid viscosity variation curves and blockage levels effectively solves the control lag problem caused by temperature drift and differences in particulate matter accumulation rates in traditional differential pressure switches, improving the accuracy of matching bypass opening timing with actual system operating conditions. Through the coordination of a triangular differential pressure correlation model and multi-level execution actions, coupled control of bypass opening degree, differential pressure fluctuations, and fluid cleanliness is achieved, avoiding filtration efficiency loss caused by single-parameter control. When abnormal differential pressure still exists even with the bypass fully open, a progressive adjustment and fault warning mechanism based on graded pressure relief efficiency mapping can accurately distinguish between bypass blockage and main line failure risks, preventing system overpressure damage due to insufficient bypass response. This method can adapt to complex operating conditions with different flow rates, temperatures, and contamination levels, maintaining fluid filtration quality while suppressing differential pressure overload, solving the problems of equipment damage and decreased filtration effect caused by rigid control in traditional methods. Attached Figure Description

[0044] Figure 1 This is a flowchart of the opening control method for the bypass of the pressure pipeline filter in Embodiment 1 of the present invention;

[0045] Figure 2 This is a diagram illustrating the construction process of the temperature-fluid cleanliness-flow rate-pressure correlation mapping space for the target device in Embodiment 1 of the present invention.

[0046] Figure 3 This is a block diagram of the control system module for the bypass of the pressure pipeline filter in Embodiment 2 of the present invention. Detailed Implementation

[0047] Example 1

[0048] Please see Figure 1 The present invention provides an embodiment of a method for controlling the opening of a pressure line filter bypass, comprising the following steps:

[0049] S1. Based on the preset target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space combined with the preset pressure classification control rules, the enhanced predictive control model is constructed to generate the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass.

[0050] Furthermore, the target device in this embodiment is a filter;

[0051] Furthermore, the pressure grading control rule in this embodiment is constructed by mapping the pressure level obtained through evaluation to the preset bypass opening degree.

[0052] Furthermore, the steps for constructing the pressure grading control rules in this embodiment include:

[0053] Based on filter design parameters and equipment safety specifications, and combined with the operating requirements of the fluid system, a critical pressure threshold is defined as a classification benchmark; this threshold must meet the maximum allowable differential pressure limit to ensure that the system operates within a safety margin.

[0054] The pressure difference range from zero to the critical threshold is divided into multiple continuous intervals, each interval corresponding to a specific pressure level and associated with a differentiated operating mode, such as normal operation, early warning, and emergency response. The interval division follows the principle of increasing risk, with higher levels indicating that the system is closer to the safety boundary.

[0055] A linear mapping relationship is established between pressure level and bypass opening. Lower pressure level corresponds to lower opening to maintain main filtration efficiency, while higher pressure level corresponds to higher opening to achieve rapid pressure relief and diversion. The opening ratio gradually increases with the pressure level, forming a gradient control strategy. Furthermore, the opening ratio here is specifically set according to the cleanliness requirements.

[0056] The opening mapping rule is dynamically adjusted according to the fluid cleanliness requirements. In high cleanliness scenarios, the diversion ratio is increased in advance to reduce the filtration load. At the same time, the pressure level threshold is corrected in real time based on fluid temperature changes to avoid measurement errors caused by viscosity changes and ensure the accuracy of classification.

[0057] When the pressure difference exceeds the safety threshold redundancy limit, the maximum bypass opening is forcibly executed, triggering the system protection mechanism. At the same time, the manual operation interface is retained, allowing the automatic control logic to be overridden and the bypass state to be set directly in an emergency.

[0058] Continuously collect real-time pressure difference data, obtain the ratio of real-time pressure difference to critical threshold, match the corresponding pressure level and obtain the preset opening degree, dynamically adjust the opening degree mapping rule based on real-time operating conditions, and optimize the control response.

[0059] S2. Real-time bypass control is performed based on the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass, and the controlled fluid cleanliness and target device pressure change values ​​are fed back to the associated mapping space to adjust the opening and closing time points and opening and closing degree in real time, so as to meet the target device pressure threshold while satisfying the fluid cleanliness.

[0060] When developing a filter bypass switch monitoring system, we faced the problem of state characterization failure under the coupling effect of multiple parameters. Specifically, the synergistic effect of fluid temperature and cleanliness on viscosity makes it impossible for single parameter monitoring to accurately reflect changes in media properties. The nonlinear mapping relationship between various physical quantities makes it difficult for traditional linear models to capture dynamic coupling effects. The reaction delay in the system leads to signal distortion in real-time monitoring, and if not calibrated, it will cause a lag in fault warning. For example, there is a time lag from temperature change to viscosity response and a transmission delay from blockage to pressure difference change. In addition, there is a conflict between multiple objectives in pressure relief control, including bypass opening degree, device pressure difference and fluid cleanliness. Excessive pressure relief can reduce pressure difference but may sacrifice filtration accuracy, while ensuring cleanliness may lead to pressure over-limit. Traditional single-variable control cannot take into account multiple constraints.

[0061] Further, please refer to Figure 2 The process of constructing the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space in this embodiment includes:

[0062] The fluid temperature, filtration flow rate, particulate matter content per unit volume of fluid, pressure difference before and after the target device, and bypass pressure difference are obtained and preprocessed to obtain the pressure relief control data sequence of the target device;

[0063] A fluid cleanliness index is established based on the particulate matter content per unit volume of fluid in the pressure relief control data sequence of the target device.

[0064] Based on fluid cleanliness index, fluid temperature and fluid viscosity, a fluid viscosity co-variance curve and viscosity coefficient scaling function are established.

[0065] The first reaction mapping delay is obtained by monitoring the time point of change in fluid temperature or fluid cleanliness index and the time point of change in the fluid viscosity-coordinated change curve.

[0066] Furthermore, the process of constructing the first reaction mapping delay in this embodiment includes:

[0067] The system collects raw data such as fluid temperature, filtration flow rate, particulate matter content, pressure difference before and after the target device and bypass pressure difference in real time through sensors. After noise reduction and normalization, a time-aligned pressure relief control data sequence is constructed.

[0068] By statistically analyzing the distribution characteristics of particulate matter and combining the cleanliness level defined by industry standards, the particulate matter content is mapped to a normalized index of 0-1.

[0069] By fitting the relationship between temperature, cleanliness and viscosity through multivariate nonlinear regression, a viscosity coefficient scaling function is generated to describe the dynamic equilibrium relationship among the three.

[0070] Analyze the time difference between the abrupt change in temperature or cleanliness and the response time of the viscosity curve, calculate the maximum hysteresis value using the cross-correlation function, and determine the first reaction mapping delay.

[0071] Based on the cross-correlation coefficients between viscosity coefficient and filtration flow rate, a target device clogging degree index is established, and a clogging degree correlation mapping curve is obtained.

[0072] The second reaction mapping delay is obtained by monitoring the time point of change in viscosity coefficient or filtration flow rate and the time point of change in the correlation curve with the degree of clogging.

[0073] Furthermore, the construction process of the second reaction mapping delay in this embodiment includes:

[0074] Based on the cross-correlation coefficient between viscosity coefficient and flow velocity, a clogging degree-flow velocity relationship model is constructed, and the correlation mapping curve between clogging degree and viscosity coefficient is fitted using historical data to characterize the influence of clogging on flow velocity.

[0075] Based on the correlation mapping curve, monitor the time point of sudden change in viscosity coefficient or filtration flow rate, track the change in clogging degree index, and determine the second reaction mapping delay by time series alignment;

[0076] Based on the target device blockage degree index, the target device front and rear pressure difference value and bypass pressure difference value, and the sequence of target device front and rear pressure difference changes under different historical bypass opening degrees, the triangular correlation mapping and corresponding curves of blockage degree, target device front and rear pressure difference and bypass pressure difference are obtained.

[0077] Furthermore, in this embodiment, the process of constructing the triangular correlation mapping and the corresponding curve includes:

[0078] A three-dimensional dynamic relationship model was established between the degree of blockage, the target device differential pressure, and the bypass differential pressure to reveal the multi-parameter coupling effect of the system. Specifically, an exponential growth curve (the first correlation mapping curve) was fitted based on the positive correlation between the degree of blockage and the target device differential pressure; a linear decay curve (the second correlation mapping curve) was fitted based on the change in bypass differential pressure with the degree of blockage; and the optimal pressure relief efficiency curve (the third correlation mapping curve) was calibrated using experimental data based on the relationship between bypass opening / closing degree and differential pressure change. Furthermore, all these mapping curves were obtained by fitting the corresponding input parameters using a mirror kernel function.

[0079] The triangular correlation mapping includes the first correlation mapping curve between the degree of blockage and the pressure difference across the target device, the second correlation mapping curve between the degree of blockage and the bypass pressure difference, and the third correlation mapping curve between the bypass opening degree and the change in the pressure difference across the target device.

[0080] Furthermore, the third correlation mapping curve construction process in this embodiment includes:

[0081] Based on the monitored pressure difference changes before and after the target device and the pressure relief efficiency and pressure classification control rules corresponding to different opening and closing degrees of the bypass, the optimal solution is obtained through a linear coupling mapping algorithm. This maximizes the cleanliness of the fluid after the target device and bypass filtration while satisfying the pressure relief of the target device. The classification control opening and closing level-pressure relief efficiency mapping function and the third correlation mapping curve are obtained. Furthermore, the linear coupling mapping algorithm in this embodiment is preferably the Lagrange multiplier method.

[0082] The third reaction mapping delay is obtained based on the time point from the change in congestion level to the time point of bypass opening.

[0083] Based on the fluid viscosity co-variance curve, the blockage degree correlation mapping curve and the triangular correlation mapping and the corresponding curve, combined with the first reaction mapping delay, the second reaction mapping delay or the third reaction mapping delay, the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space is constructed through a multi-parameter coupling algorithm.

[0084] The above process in this embodiment achieves precise and adaptive bypass control by constructing a multi-dimensional parameter correlation system and a dynamic compensation mechanism. Its technical effectiveness stems from the synergistic effect of the following core methods: First, a dynamic delay compensation model based on the fluid viscosity variation curve and the blockage degree correlation mapping curve, including first and second reaction mapping delays, quantifies the nonlinear influence of temperature and cleanliness changes on viscosity and establishes its cross-correlation relationship with flow rate, solving the control lag problem caused by neglecting the dynamic changes in fluid properties in traditional methods. When temperature or particulate matter concentration changes abruptly, the system adjusts the bypass opening and closing command in advance through pre-calculated delay time, ensuring precise matching between the actual action time and the physical response process of fluid state changes, avoiding timing deviations caused by passive responses after sensors detect parameter changes. Second, the triangular correlation mapping model constructs three correlation curves by experimentally calibrating the coupling relationship between blockage degree, main / bypass pressure difference, and opening / closing degree. A closed-loop constraint is formed: the first correlation mapping curve reflects the direct impact of filter blockage on system resistance; the second correlation mapping curve characterizes the impact of the bypass's own state on pressure relief efficiency; and the third correlation mapping curve establishes a quantitative relationship between control quantity and effect. This model expands the originally isolated differential pressure parameter into a comprehensive evaluation system that includes equipment health status, bypass performance, and control effectiveness. This makes the calculation of bypass opening degree not only based on the current differential pressure threshold but also comprehensively considers filter life loss and bypass pipeline scaling risk. Furthermore, the linear mapping algorithm that couples pressure classification control rules with cleanliness targets introduces dual constraints when solving for optimal pressure relief efficiency: mathematically, it is transformed into a constrained convex optimization problem. The objective function simultaneously minimizes the main road differential pressure deviation and the cleanliness overscalar. The optimal opening degree under each pressure level is obtained through the Lagrange multiplier method, ensuring that system safety is prioritized under high differential pressure conditions while maintaining filtration quality in the medium and low differential pressure range. Furthermore, the LSTM prediction model extracts features from differential pressure time-series data through a long short-term memory network, enabling it to identify different blockage patterns, such as gradual particle accumulation and sudden large particle jamming, and the corresponding differences in differential pressure change trajectories. Combined with reinforcement learning for online evaluation of historical control effects, the prediction time window weights are dynamically adjusted, allowing the system to maintain prediction reliability even when facing unsteady flow. The above technology chain forms a closed-loop architecture of "parameter perception - state evaluation - prediction decision - execution feedback," the effects of which are: viscosity model compensation improves control stability under temperature disturbances, triangular mapping model enhances control robustness in multivariable coupled scenarios, and hierarchical optimization algorithm balances the contradictory requirements of safety and quality, ultimately enabling the bypass control system to have the adaptive capability to cope with complex operating conditions.

[0085] Furthermore, the construction process of the enhanced predictive control model in this embodiment includes:

[0086] The input state of the enhanced predictive control model is constructed based on the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space, the pressure difference before and after the target device, the pressure difference threshold before and after the target device, the maximum upper limit of bypass shunting, and the preset pressure classification control rules.

[0087] Based on the constructed input state, the preset bypass opening and closing action trigger information is as follows:

[0088] If the pressure change prediction sub-model in the current enhanced predictive control model predicts that the change in the pressure difference across the target device at a certain future moment is greater than the preset threshold for the pressure difference across the target device, the first execution action information is triggered.

[0089] That is, based on the preset target device front and rear pressure difference threshold, pressure increase change value, and bypass pressure relief efficiency per unit time, the pressure relief rule of the corresponding level in the pressure classification control rule is triggered, and the bypass pressure relief opening degree is obtained by inversion.

[0090] When the first execution action information is triggered, the second execution action information is also triggered; that is, when the pressure difference change value before and after the target device is greater than the preset target device pressure difference threshold, the time point is combined with the first reaction mapping delay, the second reaction mapping delay or the third reaction mapping delay to obtain the corresponding pre-deduction time point for the bypass opening and closing command.

[0091] The reason for setting the first action information is to prevent equipment damage or operation interruption due to instantaneous pressure exceeding the limit. The predictive model identifies potential pressure risks in advance and initiates pressure relief actions.

[0092] The pressure change prediction sub-model constructed using the LSTM algorithm integrates time-series data such as fluid temperature, flow rate, particulate matter content, and current pressure difference to predict the pressure difference change trend at a future moment. When the predicted value exceeds the preset threshold, the required bypass opening degree is obtained by combining pressure classification control rules, such as the correlation mapping relationship between pressure relief level and bypass pressure relief efficiency, and the corresponding level of pressure relief strategy is activated in advance.

[0093] The pressure change prediction sub-model is obtained by pre-training with the LSTM algorithm using fluid temperature, filtration velocity, particulate matter content per unit volume of fluid, and pressure difference before and after the target device.

[0094] When the second execution action information is triggered, the third execution action information is triggered simultaneously, that is, real-time monitoring of the pressure difference before and after the target device, the bypass pressure difference, and the fluid cleanliness after the bypass opening command under the corresponding pressure relief level is issued at the corresponding pre-set time point.

[0095] The reason for setting the second action information is to compensate for system response delays, such as valve action delays and signal transmission delays, to ensure that the pressure relief command takes effect before the differential pressure reaches a dangerous value.

[0096] Based on the predicted time point of the first execution action, dynamic parameters such as the fluid viscosity co-change curve and the blockage degree correlation mapping curve are superimposed to obtain the impact of the first, second, and third reaction mapping delays on command execution; through the pre-deduction mechanism, the earliest issuance time of the bypass opening and closing command is determined, so that the actual effective time of the pressure relief action is synchronized with the predicted risk time point.

[0097] The reason for setting the third action information is to verify the effectiveness of the pressure relief action, dynamically adjust the strategy to balance differential pressure control and fluid cleanliness requirements, and avoid excessive pressure relief leading to insufficient filtration efficiency.

[0098] Real-time monitoring of the target device differential pressure change, bypass differential pressure, and fluid cleanliness after pressure relief. If the target device differential pressure continues to rise while the bypass differential pressure decreases, it indicates that the current pressure relief level is insufficient to suppress system pressure. In this case, the pressure relief level is recalculated using the third correlation mapping curve, and the first and second execution actions are triggered a second time to progressively increase the bypass opening degree, ensuring that the total cleanliness meets the preset standard while the differential pressure decreases. In this embodiment, the third correlation mapping curve is specifically the dynamic relationship curve between the bypass opening degree and the differential pressure change;

[0099] When the third execution action information is triggered, if the pressure difference before and after the target device is still increasing while the bypass pressure difference is decreasing, the fourth execution action information is triggered. That is, based on the pressure difference increase obtained from the second monitoring of the target device, the first and second execution action information are triggered for the second time. The pressure relief opening degree of the bypass is gradually adjusted through the third correlation mapping curve so that the pressure difference before and after the target device is decreasing while the total cleanliness of the target device and the bypass filtered fluid is lower than the preset fluid cleanliness standard value.

[0100] If the bypass opening level reaches its maximum and the voltage difference across the target device is still increasing, the bypass is determined to be in an abnormal or faulty state, and a feedback warning is issued.

[0101] The reason for setting the fourth action information is to identify bypass abnormalities or faults when the bypass system reaches its maximum pressure relief capacity but still cannot control the differential pressure, so as to prevent system collapse or secondary risks.

[0102] Specifically, when the bypass opening reaches its maximum and the differential pressure of the target device continues to rise, the bypass is determined to be faulty, such as valve jamming or pipeline blockage. By comparing the differential pressure monitoring with the bypass pressure relief effect, the abnormal state judgment logic is triggered, an early warning signal is sent to the control system, and an emergency protection mechanism is activated, such as switching to the backup bypass or shutting down for maintenance.

[0103] Based on the preset bypass opening and closing action trigger information, a reward and execution action corresponding to each execution action trigger information are constructed; further, the reward in this embodiment includes the deviation of the preceding time point, the deviation of the pressure relief opening and closing degree, the penalty reward corresponding to the number of times the fourth execution action is triggered, and the deviation value of the total cleanliness of the adjusted target device and the fluid after bypass filtration from the preset fluid cleanliness standard value; further, in this embodiment, the penalty reward corresponding to each trigger of the fourth execution action is determined by the pressure difference corresponding to the triggering information of the fourth execution action, which is specifically set by those skilled in the art based on historical experience.

[0104] Based on the input state of the reinforcement predictive control model, the preset bypass opening and closing action execution trigger information, the reward and execution action corresponding to each execution action trigger information, the reinforcement learning algorithm is combined with the simulation algorithm to train and obtain the trained reinforcement predictive control model.

[0105] The enhanced predictive control model in this embodiment achieves accurate decision-making and adaptive optimization of bypass control under complex operating conditions through the collaborative design of a multi-level triggering mechanism and a dynamic reward strategy. Its technical effects are reflected in the following aspects: First, the LSTM-based pressure change prediction sub-model, by analyzing the temporal correlation between temperature, flow rate, particulate matter concentration, and differential pressure, can identify the characteristic differences of different blockage modes. For example, in a hydraulic oil system, the gradual accumulation of fine particles manifests as a slow linear increase in differential pressure, while sudden blockage by large fibrous particles presents a step change in differential pressure. By capturing these differences in modes, the model can predict the time point when the differential pressure exceeds the limit and trigger the first action to execute the graded pressure relief rule. This means that the generation of control commands no longer depends on a single threshold trigger, but rather dynamically matches the opening and closing degree by combining the pressure increase rate and bypass pressure relief efficiency. Secondly, the multi-level delay compensation mechanism incorporates the inherent delays of sensor detection, fluid property response, and mechanical execution into the control timing planning. For example, when a sudden change in cleanliness is detected leading to an increase in viscosity, the system issues a bypass adjustment command in advance based on the pre-calculated fluid response delay, ensuring that the action execution is synchronized with the physical process of fluid state change, avoiding control lag caused by the accumulation of delays in traditional methods. Furthermore, the introduction of the fourth action constructs a closed-loop correction mechanism: when an abnormal increase in the main pipeline pressure difference is detected after the bypass is opened, the system uses the third correlation mapping curve to back-calculate the optimal opening adjustment amount and combines it with the cleanliness deviation to correct the control strategy in real time. For example, in a chemical pipeline, if the pressure difference still rises at a certain rate after the bypass is opened halfway for the first time, a secondary adjustment is triggered, gradually increasing the bypass diversion in increments of 'a%' until the pressure difference trend reverses. This process uses a penalty and reward mechanism, such as deducting points for frequently triggering the fourth action, to constrain over-adjustment behavior, minimizing cleanliness degradation while ensuring pressure relief. Furthermore, the reward function design under the reinforcement learning framework transforms multiple objectives such as time deviation, opening and closing error, and cleanliness deviation into a comprehensive reward signal, driving the model to explore the optimal control strategy in the simulation environment. For example, in the digital twin of the cooling water system of a nuclear power plant, the model learns through millions of iterations and autonomously discovers that under high-temperature conditions, a lower bypass opening degree should be prioritized and supplemented with temperature compensation, rather than simply following the pressure level rule.

[0106] Example 2

[0107] Please see Figure 3 Another embodiment of the present invention provides: an opening control system for a pressure pipeline filter bypass, comprising: an enhanced control module and a control feedback module;

[0108] The enhanced control module, based on the preset target device temperature-fluid cleanliness-flow-pressure correlation mapping space and the preset pressure graded control rules, constructs an enhanced predictive control model to generate the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass.

[0109] The control feedback module performs real-time bypass control based on the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass. It also feeds back the controlled fluid cleanliness and target device pressure change values ​​to the associated mapping space to adjust the opening and closing time points and opening and closing degree in real time, so as to meet the target device pressure threshold while satisfying the fluid cleanliness.

[0110] 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, under the guidance of the present invention, can make changes, modifications, substitutions and variations to the above embodiments without departing from the spirit and scope of the claims. All of these variations are within the protection scope of the present invention.

Claims

1. A method for controlling the opening of a pressure line filter bypass, characterized in that, include: An enhanced predictive control model is constructed based on a preset target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space combined with a preset pressure graded control rule, generating the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass. Real-time bypass control is performed based on the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass, and the controlled fluid cleanliness and target device pressure change values ​​are fed back to the associated mapping space to adjust the opening and closing time points and opening and closing degree in real time. The process of constructing the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space includes: The fluid temperature, filtration flow rate, particulate matter content per unit volume of fluid, pressure difference before and after the target device, and bypass pressure difference are obtained and preprocessed to obtain the pressure relief control data sequence of the target device; A fluid cleanliness index is established based on the particulate matter content per unit volume of fluid in the pressure relief control data sequence of the target device. Based on fluid cleanliness index, fluid temperature and fluid viscosity, a fluid viscosity co-variance curve and viscosity coefficient scaling function are established. The first reaction mapping delay is obtained by monitoring the time point of change in fluid temperature or fluid cleanliness index and the time point of change in the fluid viscosity-coordinated change curve. Based on the cross-correlation coefficients between viscosity coefficient and filtration flow rate, a target device clogging degree index is established, and a clogging degree correlation mapping curve is obtained. The second reaction mapping delay is obtained by monitoring the time point of change in viscosity coefficient or filtration flow rate and the time point of change in the correlation curve with the degree of clogging. Based on the target device blockage degree index, the target device front and rear pressure difference value and bypass pressure difference value, and the sequence of target device front and rear pressure difference changes under different historical bypass opening degrees, the triangular correlation mapping and corresponding curves of blockage degree, target device front and rear pressure difference and bypass pressure difference are obtained. The triangular correlation mapping includes a first correlation mapping curve between the degree of blockage and the pressure difference before and after the target device, a second correlation mapping curve between the degree of blockage and the bypass pressure difference, and a third correlation mapping curve between the bypass opening degree and the change value of the pressure difference before and after the target device. The third reaction mapping delay is obtained based on the time point from the time point of change in the degree of congestion to the time point of bypass opening; Based on the fluid viscosity co-variance curve, the blockage degree correlation mapping curve and the triangular correlation mapping and the corresponding curve, combined with the first reaction mapping delay, the second reaction mapping delay or the third reaction mapping delay, the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space is constructed through a multi-parameter coupling algorithm. The construction process of the enhanced predictive control model includes: The input state of the enhanced predictive control model is constructed based on the target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space, the pressure difference before and after the target device, the pressure difference threshold before and after the target device, the maximum upper limit of bypass shunting, and the preset pressure classification control rules. Based on the constructed input state, the preset bypass opening and closing action trigger information is as follows: If the pressure change prediction sub-model in the current enhanced predictive control model predicts that the pressure difference change value before and after the target device at a certain future time is greater than the preset pressure difference threshold before and after the target device, the first execution action information is triggered; that is, the pressure relief rule of the corresponding level in the pressure classification control rule is triggered according to the preset pressure difference threshold before and after the target device, the pressure increase change value, and the pressure relief efficiency of the bypass per unit time, and the bypass pressure relief opening degree is obtained by inversion. When the first execution action information is triggered, the second execution action information is also triggered simultaneously; That is, the time point when the pressure difference change value before and after the target device is greater than the preset pressure difference threshold before and after the target device is predicted by the pressure change prediction sub-model, combined with the first reaction mapping delay, the second reaction mapping delay or the third reaction mapping delay, is used to calculate the corresponding pre-time point for issuing the bypass opening and closing command. The pressure change prediction sub-model is obtained by pre-training the LSTM algorithm with fluid temperature, filtration flow rate, particulate matter content in a unit volume of fluid, and pressure difference before and after the target device. When the second execution action information is triggered, the third execution action information is triggered simultaneously, that is, real-time monitoring of the pressure difference before and after the target device, the bypass pressure difference, and the fluid cleanliness after the bypass opening command under the corresponding pressure relief level is issued at the corresponding pre-set time point. When the third execution action information is triggered, if the pressure difference before and after the target device is still increasing while the bypass pressure difference is decreasing, the fourth execution action information is triggered. That is, based on the pressure difference increase obtained from the secondary monitoring of the target device, the first and second execution action information are triggered a second time. The pressure relief opening degree corresponding to the bypass is gradually adjusted through the third correlation mapping curve, so that the pressure difference before and after the target device is decreasing while the total cleanliness of the target device and the bypass filtered fluid is lower than the preset fluid cleanliness standard value.

2. The opening control method for a pressure pipeline filter bypass as described in claim 1, characterized in that, The process of constructing the enhanced predictive control model also includes: If the bypass opening level reaches its maximum and the voltage difference across the target device continues to rise, the bypass is determined to be in an abnormal or faulty state, and a feedback warning is issued; the process of constructing the third correlation mapping curve includes: Based on the monitored pressure difference changes before and after the target device and the pressure relief efficiency and pressure classification control rules corresponding to different opening and closing degrees of the bypass, the optimal solution is obtained through a linear coupling mapping algorithm. This maximizes the cleanliness of the target device and the fluid filtered by the bypass under the premise of satisfying the pressure relief of the target device, and obtains the classification control opening and closing level-pressure relief efficiency mapping function and the third correlation mapping curve. The pressure grading control rule is constructed by mapping the pressure level obtained through evaluation to the preset bypass opening degree.

3. A control system for opening a pressure line filter bypass, used to implement the opening control method for opening a pressure line filter bypass as described in any one of claims 1-2, characterized in that, include: Strengthen the control module and control feedback module; The enhanced control module generates the opening and closing time points and opening and closing degree control commands corresponding to the bypass of the target device based on the enhanced predictive control model constructed by combining the preset target device temperature-fluid cleanliness-flow rate-pressure correlation mapping space with the preset pressure graded control rules. The control feedback module performs real-time bypass control based on the opening and closing time points and opening and closing degree control commands corresponding to the target device bypass, and feeds back the controlled fluid cleanliness and target device pressure change values ​​to the associated mapping space to adjust the opening and closing time points and opening and closing degree in real time.

Images