A control system for a fan filter unit

Abstract

The application relates to the technical field of automation control, in particular to a control system of a fan filter unit (FFU). The control system comprises a safety storage module, a distributed execution module, a state monitoring module, a logic arbitration module, an envelope dynamic correction module and a mode switching execution module. The distributed execution module executes self-organizing control based on a safety envelope interval, outputs an uncertainty expected control quantity, the state monitoring module acquires operation feedback information, the logic arbitration module performs out-of-bound detection and confidence evaluation and outputs an arbitration result, the envelope dynamic correction module dynamically adjusts a safety envelope interval boundary value by calling a degradation function parameter set according to aging information and the arbitration result, and the mode switching execution module switches between self-organizing control and deterministic control according to the arbitration result. The application not only retains the high energy efficiency and flexibility of the distributed self-organizing control, but also ensures system compliance and reliability through physical red lines and deterministic switching, thereby improving the robustness and safety of clean room FFU control.

Description

Technical Field

[0001] This invention relates to the field of safe operation control technology, and in particular to a control system for a fan filter unit. Background Technology

[0002] With the rapid development of high-end industries such as semiconductor manufacturing, biomedicine, and aerospace, the requirements for air cleanliness in cleanrooms are becoming increasingly stringent. Fan filter units (FFUs), as the core circulating power unit in cleanrooms, have seen their application scale jump from hundreds to tens of thousands of units. However, existing FFU control systems have significant shortcomings: Control architecture defects: Traditional centralized control relies on manual adjustment or simple group start-stop, making it difficult to adapt to the high-speed communication and dynamic balance requirements of ultra-large-scale nodes, and it has a high risk of single-point failure and limited energy efficiency optimization capabilities. While distributed self-organizing control has high fault tolerance and energy-saving potential, its non-deterministic behavior makes it difficult to pass regulatory compliance reviews such as GMP, and it lacks traceable fault location methods. FFU system energy consumption accounts for 30%–50% of the total energy consumption of cleanroom air conditioning, and on-demand air supply technology is limited by low-speed control linearity. The complexity of temperature and dynamic differential pressure compensation makes it difficult for existing solutions to achieve deep energy savings while ensuring cleanliness and safety. The metal corrugated steel structure and static pressure box design of cleanrooms make wired buses susceptible to electromagnetic interference, while wireless solutions suffer from multipath attenuation. Furthermore, FFU group control systems exhibit nonlinear coupling—fine-tuning of local airflow can cause fluctuations in the entire airflow field, which traditional linear control struggles to handle. Current maintenance relies on manual inspections and passive sensor alarms, making it difficult to identify latent faults such as sensor drift, filter dust accumulation, and bearing wear early, and lacking the ability to dynamically adjust boundaries based on equipment health status. Therefore, existing technologies struggle to meet the stringent requirements of cleanrooms for reliability, traceability, and adaptive maintenance while simultaneously ensuring high energy efficiency and compliance with safety regulations in distributed control. Summary of the Invention

[0003] To overcome the above shortcomings, this invention provides a control system for a wind turbine filter unit, which aims to improve the problems of existing distributed self-organizing control, which is difficult to meet compliance requirements, lack of safety constraints and traceability, and traditional centralized control, which has low energy efficiency, poor flexibility and difficulty in adapting to large-scale dynamic wind field adjustment technology.

[0004] This invention provides the following technical solution: a control system for a fan filter unit, comprising: The secure storage module is used to store the safety envelope range, physical safety limits, deterministic control logic, and degradation function parameter set preset based on the inherent parameters of the wind turbine unit. A distributed execution module is used to execute a self-organizing control algorithm based on the safe envelope interval to output an uncertain expected control quantity, wherein the expected control quantity is limited by the safe envelope interval; The status monitoring module is used to acquire the actual operating status and group coordination status of FFU nodes through sensors and output operating feedback information. The logic arbitration module is used to perform out-of-bounds detection and calculate the confidence score based on the operation feedback information, and output the arbitration result based on the out-of-bounds detection result and the confidence score; The envelope dynamic correction module is used to dynamically adjust the boundary value of the safe envelope interval by calling the degradation function parameter set according to the operation feedback information and the arbitration result, and the adjusted boundary value is limited by the physical security limit; The mode switching execution module is used to switch control permissions between self-organizing control and deterministic control according to the arbitration result, and has the function of permission restoration.

[0005] Furthermore, the secure storage module includes: The safety envelope is pre-set based on industry compliance standards or cleanroom class, including at least one of the upper and lower limits of differential pressure and upper and lower limits of air velocity; Physical safety limits are insurmountable boundary values ​​set based on the physical characteristics of the equipment or mandatory industry standards. The deterministic control logic consists of hard-coded execution rules in the controller's underlying layers, including at least one of constant speed, constant airflow, and fail-safe mode; the degradation function parameter set contains the mapping relationship between the degree of equipment aging and the envelope boundary adjustment, with the cumulative motor running time, shaft offset, and current RMS fluctuation rate as inputs.

[0006] Furthermore, in the distributed execution module, the step of executing the self-organizing control algorithm to output the uncertain expected control quantity includes: Each FFU node interacts with information through a distributed neighborhood communication protocol, and achieves local self-organized balance control of pressure difference and wind speed based on a group cooperation algorithm; Each node independently outputs an uncertain expected control quantity based on the local interaction results, and this control quantity is limited by the current safe envelope interval.

[0007] Furthermore, the swarm collaboration algorithm includes a swarm self-organization rule based on local interaction, and a consensus protocol as a convergence constraint of the swarm rule; each FFU node only exchanges state information with neighboring nodes, and independently calculates the adjustment step size and direction of the desired control quantity based on the local pressure difference or wind speed deviation, the number of neighboring nodes and its own aging degree, so that the swarm as a whole achieves a distributed balance between pressure difference and wind speed.

[0008] Furthermore, in the status monitoring module, the step of outputting operational feedback information includes: Continuously collect the uncertain expected control quantity output by the distributed execution module and the collaborative state variables between each FFU node; The collected data is time-aligned and features are extracted to generate operation feedback information with time-series labels. The operation feedback information includes the device aging status inferred from the node operation data.

[0009] Furthermore, in the logic arbitration module, the step of outputting the arbitration result based on the boundary detection result and the confidence score includes: The operation feedback information is obtained and compared with the boundary value of the current safe envelope interval. If it exceeds the boundary, an out-of-bounds flag is triggered. Based on the degree of deviation between the operational feedback information and the envelope boundary, the control consistency index between nodes, and historical switching statistics, the confidence score of the current self-organizing control is calculated. If the out-of-bounds flag is triggered or the confidence score is lower than the preset threshold, the output will switch the arbitration result; otherwise, the output will maintain the arbitration result.

[0010] Furthermore, in the envelope dynamic correction module, the step of dynamically adjusting the boundary values ​​of the safe envelope interval includes: Extract equipment aging parameters from the operational feedback information and obtain the historical arbitration results output by the logic arbitration module; The degradation function parameter set is invoked, and the adjustment amount of each boundary value of the safety envelope interval is calculated according to the equipment aging parameters and historical arbitration results. Based on the adjustment amount, perform a shrinking or shifting operation on the current safe envelope interval to generate candidate boundary values; The candidate boundary value is compared with the preset physical security limit. If it exceeds the limit, it is limited to the limit value, and the updated security envelope is used for subsequent distributed execution and logical arbitration.

[0011] Furthermore, in the mode switching execution module, the step of switching control permissions includes: When the arbitration result is a switch, the self-organizing control authority is forcibly revoked and taken over by the deterministic control logic; The permission recovery function includes restoring self-organizing control permissions after a forced switch, when the running feedback information enters the safe envelope interval and the confidence score is higher than the recovery threshold.

[0012] Furthermore, it also includes an arbitration log recording module, which is used to record arbitration and switching events. The events include timestamps, triggering reasons, operational feedback and envelope boundaries, and provide a read-only interface for compliance auditing or predictive maintenance.

[0013] The present invention has the following beneficial effects: This invention constructs a dual closed-loop architecture of distributed self-organization, secure envelope, and deterministic switching. Through physical security limits and mode switching mechanisms, it ensures that even when the distributed algorithm malfunctions or the equipment ages, it can still force a switch to deterministic control logic. It introduces a dynamic envelope correction module based on equipment aging parameters and arbitration history, enabling the safety boundary to adapt to changes throughout the entire lifecycle, such as filter clogging and motor wear, thus extending the effective service life of the equipment. It utilizes soft measurement technology to reduce the deployment of high-precision sensors and combines digital twins and arbitration log recording to achieve fault tracing, predictive maintenance, and compliance auditing. It supports one-click recovery of traditional control modes, reducing user switching risks and system integration difficulties. Attached Figure Description

[0014] Figure 1 This is a control system architecture diagram of a fan filter unit proposed in this invention; Figure 2 This is a flowchart of the envelope dynamic correction module proposed in this invention; Figure 3 This is a flowchart of the mode switching execution module proposed in this invention. Detailed Implementation

[0015] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] This invention provides a control system for a fan filter unit, such as... Figure 1 As shown, it includes: The secure storage module is used to store the safety envelope range, physical safety limits, deterministic control logic, and degradation function parameter set preset based on the inherent parameters of the wind turbine unit. Furthermore, the secure storage module includes: The safety envelope is pre-set based on industry compliance standards or cleanroom class, including at least one of the upper and lower limits of differential pressure and upper and lower limits of air velocity; Physical safety limits are insurmountable boundary values ​​set based on the physical characteristics of the equipment or mandatory industry standards. The deterministic control logic consists of hard-coded execution rules in the controller's underlying layers, including at least one of constant speed, constant airflow, and fail-safe mode; the degradation function parameter set contains the mapping relationship between the degree of equipment aging and the envelope boundary adjustment, with the cumulative motor running time, shaft offset, and current RMS fluctuation rate as inputs.

[0017] Specifically, this embodiment employs a non-volatile memory with hardware encryption capabilities to ensure that data is not lost in the event of power failure. Internally, it uses a structured database to store four types of core data: a secure envelope interval table, physical security limit constants, a hard-coded deterministic logic instruction set, and a degradation function coefficient matrix.

[0018] safe envelope interval This is the logical boundary preset for the cleanroom class based on the FFU deployment environment. Taking an ISO Class 5 cleanroom as an example, the specific parameters stored are the differential pressure envelope. Wind speed envelope and power envelope ,in Units are , Units are , Units are .

[0019] Physical safety limits These parameters are typically derived from the equipment's factory specifications or mandatory industry safety standards. Storage parameters include the absolute speed limit. Upper limit of effective current value and bus voltage threshold Among them, to prevent the impeller from breaking due to centrifugal force. No more than To prevent the windings from overheating and burning out No more than , In adjustment When dealing with boundary values, a verification function must be executed. This ensures that dynamic adjustments do not violate physical red lines.

[0020] Deterministic control logic Typically implemented as stored in The instruction set includes a fail-safe mode that forces a switch to constant speed mode when the communication bus heartbeat is lost for more than 500ms; and constant airflow logic that, based on a pre-stored pressure-airflow characteristic curve, compensates for the speed in real time according to sensor feedback to ensure basic air circulation.

[0021] Degenerate function parameter set This is used to calculate the impact of equipment aging on safety boundaries. Input parameters include: cumulative motor operating time. (hours, h), axis offset (Displacement amplitude, mm, obtained through built-in vibration sensor) and current RMS fluctuation rate (The ratio of the standard deviation to the mean of the current under steady-state conditions).

[0022] Establish a weighted scoring model to quantify the degree of aging and health score. The formula for the degradation evaluation model is: ; in, , , The weighted coefficients are stored in the parameter set of the degenerate function, and In this embodiment, the value is taken as... , , ; The theoretical lifespan designed for the equipment; and These are the limit offset and fluctuation thresholds for equipment in a scrapped state.

[0023] According to the calculation The value is dynamically adjusted to correct the boundary value of the safety envelope interval. This is based on the upper limit of wind speed. For example, the corrected boundary The calculation formula is ,in For the degradation adjustment function: ; in, The attenuation slope constant has a value of [value missing]. ; This is the critical fraction at which performance begins to degrade. This ensures that when the device ages... When the speed is reduced, the safety envelope range is automatically narrowed to prevent mechanical failures in older equipment at high speeds.

[0024] By introducing a dynamic adjustment mechanism for degradation functions based on aging conditions, a shift from post-maintenance to pre-prevention protection has been achieved. By actively limiting the output intensity under sub-health conditions, catastrophic physical damage has been effectively avoided.

[0025] A distributed execution module is used to execute a self-organizing control algorithm based on the safe envelope interval to output an uncertain expected control quantity, wherein the expected control quantity is limited by the safe envelope interval; Furthermore, in the distributed execution module, the step of executing the self-organizing control algorithm to output the uncertain expected control quantity includes: Each FFU node interacts with information through a distributed neighborhood communication protocol, and achieves local self-organized balance control of pressure difference and wind speed based on a group cooperation algorithm; Each node independently outputs an uncertain expected control quantity based on the local interaction results, and this control quantity is limited by the current safe envelope interval.

[0026] Specifically, in this embodiment, each FFU node establishes a scale-free network topology via an industrial Ethernet or wireless mesh network. The neighborhood state set is the node's... At any moment The maintained state vector is ,in Pressure difference, For wind speed, This is a degradation factor. It employs a push-pull interaction based on the Gossip protocol, with nodes... With random period To the nodes in the neighborhood list Send a status request. If the node... If online, return its current state vector. This mechanism ensures that the algorithm can still achieve global convergence through the remaining links when communication links between some nodes are interrupted.

[0027] Furthermore, the swarm collaboration algorithm includes a swarm self-organization rule based on local interaction, and a consensus protocol as a convergence constraint of the swarm rule; each FFU node only exchanges state information with neighboring nodes, and independently calculates the adjustment step size and direction of the desired control quantity based on the local pressure difference or wind speed deviation, the number of neighboring nodes and its own aging degree, so that the swarm as a whole achieves a distributed balance between pressure difference and wind speed.

[0028] Specifically, based on the local deviation assessment of the consensus protocol, the node By combining feedback from neighboring nodes, the consistency deviation vector of the overall trend is calculated. : ; in, For communication weighting factors, satisfying Defined by the spatial distance or logical correlation between nodes, this ensures that nodes can perceive the operating load of surrounding FFUs, providing a data foundation for collaborative compensation.

[0029] Based on the uncertainty search of bee colony rules, the algorithm introduces a random search mechanism to simulate the probing behavior of bee colonies when searching for nectar sources. This uncertainty is key to achieving optimal global energy efficiency. (Expected rotational speed increment) The calculation formula is: ; in, The following coefficient is set to 0.8, which guides the node to move closer to the node with the highest energy efficiency ratio and the highest pressure drop / power consumption in the neighborhood. For the uncertainty disturbance term, For the amplitude coefficient of uncertainty disturbance, , It is a random variable that conforms to the Lévy flight distribution. Its function is to allow the FFU to make subtle speed trials within the safe envelope in order to bypass local resistance fluctuations and find a better flow field equilibrium point.

[0030] Introducing a weighted adjustment based on the degree of aging to control the expected amount of final uncertainty. Weighting needs to be applied based on the health status of each node: ; in, Step size factor ; As a degradation suppression term, when node Degradation factors When the node is at a high level, meaning it is severely aged, its sensitivity to cooperative commands automatically decreases, and it assigns compensation tasks to healthy neighboring nodes, thus enabling the capable to do more work.

[0031] Calculated It does not directly drive the motor; it must go through hard-constraint arbitration logic. It retrieves the current time frame from the secure storage module in real-time via interval acquisition. ; Saturation cutoff Dynamic anti-shake: To prevent frequent motor speed adjustments caused by excessively high uncertainty search frequency, a threshold is set. (1%-5% or 10-50 RPM of rated speed), only when The PWM duty cycle is only updated at that time.

[0032] By utilizing swarm rules and Lévy flight random variables, the probability of searching for the globally optimal energy efficiency ratio in complex environments is improved. Combined with a degradation factor correction mechanism, intelligent load allocation based on device health is achieved. Finally, under the hard constraint of secure envelope mapping, the system is ensured to have extremely high resistance to single point of failure, real-time response sensitivity, and device lifespan during large-scale deployment.

[0033] The status monitoring module is used to acquire the actual operating status and group coordination status of FFU nodes through sensors and output operating feedback information. Furthermore, in the status monitoring module, the step of outputting operational feedback information includes: Continuously collect the uncertain expected control quantity output by the distributed execution module and the collaborative state variables between each FFU node; The collected data is time-aligned and features are extracted to generate operation feedback information with time-series labels. The operation feedback information includes the device aging status inferred from the node operation data.

[0034] Specifically, physical sensor data and logical intermediate variables are acquired in real time via a hardwired interface and communication bus. Node operation data includes outlet static pressure. Wind speed feedback Motor operating current and the acceleration amplitude output by the vibration sensor The consistency deviation vector of the collaborative state variables intercepted from the distributed execution module. The average load rate of neighboring nodes and the packet loss rate of the current communication link. Real-time capture of control command data includes the uncertain expected control quantity output by the distributed execution module. .

[0035] Because the sampling frequency and network latency differ among nodes in a distributed system, data preprocessing is performed. Based on sliding window time alignment, a bidirectional circular buffer is constructed in memory, and an interpolation completion algorithm is used. If sampling points from physical sensors are detected... The point of generation of control quantity There is a deviation Then, cubic spline interpolation is used to fill in the missing sample points, ensuring the generation of feature vectors with uniform temporal labels. .

[0036] Feature extraction and operating condition dimensionality reduction: To extract effective aging-related information from massive datasets, time-frequency domain feature transformation and current ripple analysis are performed. Perform Fast Fourier Transform to extract the amplitude of harmonic components at specific frequencies, such as the motor's fundamental frequency, to identify bearing wear characteristics; extract control gain sensitivity and calculate... This refers to the rate of change in differential pressure caused by unit control gain. This characteristic value is the most direct indicator for identifying filter blockage.

[0037] Aging status can be assessed through performance residual analysis. Quantitative evaluation. The performance benchmark library storage module pre-stores the ideal characteristic curves of the FFU in its factory condition. The deviation is calculated. Aging factors The fusion calculation formula is as follows: ; in, , , These are weighting coefficients, corresponding to aerodynamic efficiency degradation, mechanical structure wear, and thermal aging, respectively. ,when When the value approaches 1, it indicates that the equipment is nearing the end of its physical lifespan or is experiencing severe blockage.

[0038] The generated operational feedback information is sent to the logic arbitration module and the secure storage module via the internal high-bandwidth bus. In addition to single-machine aging, this module also extracts the group cooperative coupling degree by calculating the local node's... Compared with the neighborhood average The cross-correlation function is used to assess whether a node has deviated from the group's coordination rhythm. If the synchronization entropy increases abnormally, the operational feedback information will include a node disconnection warning, triggering the system to restart the consensus protocol process.

[0039] The interpolation completion mechanism reduces the computational deviation caused by asynchronous sampling of heterogeneous data in the distributed architecture, ensuring high fidelity of state perception in the time domain; by using a quantization model based on performance residuals, complex physical losses are abstracted into standardized aging factors that can participate in the logical closed loop, realizing accurate quantitative monitoring of equipment health status.

[0040] The logic arbitration module is used to perform out-of-bounds detection and calculate the confidence score based on the operation feedback information, and output the arbitration result based on the out-of-bounds detection result and the confidence score; Furthermore, in the logic arbitration module, the step of outputting the arbitration result based on the boundary detection result and the confidence score includes: The operation feedback information is obtained and compared with the boundary value of the current safe envelope interval. If it exceeds the boundary, an out-of-bounds flag is triggered. Based on the degree of deviation between the operational feedback information and the envelope boundary, the control consistency index between nodes, and historical switching statistics, the confidence score of the current self-organizing control is calculated. If the out-of-bounds flag is triggered or the confidence score is lower than the preset threshold, the output will switch the arbitration result; otherwise, the output will maintain the arbitration result.

[0041] Specifically, it receives operational feedback information generated by the status monitoring module in real time. and compare it with the dynamic boundary in the secure storage module. Compare them. If or Immediately mark it as a transient anomaly. To prevent false cuts caused by sensor noise, adopt... Detection algorithms, in continuous If the number of out-of-bounds occurrences exceeds a certain limit within a sampling period. Then, the boundary crossing flag is officially triggered. Calculate the percentage of the current state that is far from the boundary. .when When this happens, the module enters early warning monitoring mode and increases the sampling frequency.

[0042] The confidence score is used to assess whether the current distributed control algorithm remains reliable. Its calculation consists of the deviation convergence component. Group consistency index and historical stability coefficient It is composed of weighted components from three dimensions.

[0043] Deviation convergence component It reflects the degree to which the actual state of a node closely approximates the desired distributed objective. A Gaussian function mapping is used. ,in This is the sensitivity coefficient, with a value of [value missing]. The greater the deviation, the more exponentially the confidence level decreases.

[0044] Group Consistency Index This reflects whether a node has deviated from the group's cooperative trajectory. The calculation formula based on the neighborhood deviation vector is as follows: ,in for The adjustment coefficient of the function takes the value of If neighboring nodes are all accelerating, but this node is unable to keep up due to an anomaly, As the confidence level increases, the confidence level decreases.

[0045] Historical stability coefficient Utilize the recent Number of mode switches within the time window Punish ,in This is the preset maximum number of allowed switching times.

[0046] Overall confidence formula: ; in, The preset weighting coefficients satisfy... .

[0047] Based on the combination of the boundary violation flag and the confidence score, the final result is output through the arbitration logic state machine. When the arbitration result is a switch, the motor speed is not changed instantaneously, but a safe takeover process is executed. The self-organizing mechanism is disconnected, and updates of the distributed execution module are immediately frozen. Deterministic logic is loaded, and fail-safe mode parameters, such as preset constant speed values, are retrieved from the safety storage module. A gradient transition is performed, using a ramp function to smoothly transition the current speed to the safe target speed within 1.5 seconds, preventing instantaneous negative pressure in the cleanroom or damage to the motor bearings due to sudden speed changes.

[0048] By deeply integrating boundary violation detection and multidimensional confidence assessment, a technological leap has been achieved from reactive post-event response to proactive preventative monitoring. This not only enables the utilization of... The detection algorithm effectively filters sensor noise and eliminates false triggers. It can also provide early warning and intervention when the physical boundary is not exceeded but the algorithm's coordination deteriorates. Combined with its built-in smooth switching mechanism, it reduces the instantaneous impact of control mode changes on the actuator and the controlled environment, providing robust performance for continuous production under complex working conditions.

[0049] The envelope dynamic correction module is used to dynamically adjust the boundary value of the safe envelope interval by calling the degradation function parameter set according to the operation feedback information and the arbitration result, and the adjusted boundary value is limited by the physical security limit; Furthermore, such as Figure 2 As shown, in the envelope dynamic correction module, the step of dynamically adjusting the boundary values ​​of the safe envelope interval includes: Extract equipment aging parameters from the operational feedback information and obtain the historical arbitration results output by the logic arbitration module; The degradation function parameter set is invoked, and the adjustment amount of each boundary value of the safety envelope interval is calculated according to the equipment aging parameters and historical arbitration results. Based on the adjustment amount, perform a shrinking or shifting operation on the current safe envelope interval to generate candidate boundary values; The candidate boundary value is compared with the preset physical security limit. If it exceeds the limit, it is limited to the limit value, and the updated security envelope is used for subsequent distributed execution and logical arbitration.

[0050] Specifically, by subscribing to real-time operational feedback information from the status monitoring module, evaluation indicators are extracted from two core dimensions: physical health indicators and system stability indicators. Physical health indicators The degradation factor, synthesized from motor running time, vibration amplitude, and current fluctuation rate, has a value range of [value missing]. System stability indicators It is calculated by obtaining the historical arbitration results output by the logical arbitration module. The formula is: ; in, This represents the number of control mode switches in the past 24 hours. Frequent switching indicates that the current envelope settings are not effectively compatible with environmental disturbances.

[0051] Call the pre-stored degenerate function parameter set Calculate the safe envelope interval based on the above indicators. The adjustment vector is calculated. The shrinkage coefficient is calculated, and to reduce the risk of failure of aging equipment under high loads, the module calculates the boundary shrinkage ratio. : ; in, and The preset sensitivity coefficient has a typical value of [value missing]. and .

[0052] Calculate the translation amount. If the operational feedback shows that the actual state is consistently biased towards one side of the boundary and no arbitration switch is triggered, it indicates that the environmental baseline has shifted, and the translation amount needs to be calculated. : ; in, This is the translation step size factor. ; It is the median of the original envelope interval.

[0053] Based on adjustment amount and The following transformation is performed on the current safe envelope interval to generate candidate boundary values. The interval width is adjusted to approach the median, and a shrinking operation is performed. ,in The entire movement is centered on the translation operation. .

[0054] Candidate value generation , .

[0055] Generated candidate boundary values It cannot take effect directly; it must pass through physical security limits. Hard constraints.

[0056] By establishing a nonlinear mapping between health degradation indicators and control boundaries, the safety envelope range is made adaptive and dynamically evolved throughout the entire life cycle of the device. This not only proactively avoids the failure risk caused by hardware aging through geometric transformation logic and effectively reduces the frequent switching of control modes caused by environmental deviation or performance degradation, but also achieves a deep decoupling and synergistic balance between system operating performance and hardware physical reliability under complex operating conditions by combining hard constraint verification of physical safety limits.

[0057] The mode switching execution module is used to switch control permissions between self-organizing control and deterministic control according to the arbitration result, and has the function of permission restoration.

[0058] Furthermore, such as Figure 3 As shown, in the mode switching execution module, the steps for switching control permissions include: When the arbitration result is a switch, the self-organizing control authority is forcibly revoked and taken over by the deterministic control logic; The permission recovery function includes restoring self-organizing control permissions after a forced switch, when the running feedback information enters the safe envelope interval and the confidence score is higher than the recovery threshold.

[0059] Specifically, in this embodiment, the mode switching execution module acts as a circuit breaker and switcher for the system control flow. This module runs at the physical level within the processor's high-priority interrupt service routine, and its core logic is a bi-state switching switch driven by a logical arbitration result. The expected quantity generated by the input self-organizing control... Deterministic control logic Generated security value Logical arbitration mark The output terminal acts as the final control pulse for the motor driver. .

[0060] When the logic arbitration module outputs a switching signal, the execution module immediately initiates a fallback takeover process. It immediately blocks the distributed execution module's write access to the PWM register to prevent unstable self-organizing calculation results from interfering with hardware execution. It then extracts the deterministic control logic from the read-only area of ​​the secure storage module. Typical This includes constant pressure control, which uses closed-loop feedback based on a preset static pressure target value; and safe speed control, which forces a switch to 60%-80% of rated power to ensure a basically positive pressure in the cleanroom. To avoid sudden torque changes caused by permission switching, this embodiment uses a first-order hysteresis transition algorithm. ; in, The smoothing coefficient has a value of [value missing]. ; This is the last self-organized output at the moment of switching. After takeover... Over a period of time, the controlled quantity gradually approaches the safe target value.

[0061] The permission recovery function aims to ensure that the system can return to an efficient self-organizing state after the environment stabilizes. To prevent control oscillations near the switching point, this module introduces a recovery determination based on hysteresis loops: Boundary regression verification only occurs when the runtime feedback information is received. Fully enter the safe envelope region Within the central 60% area, and for a duration exceeding the recovery observation window. Only then will the recovery process begin.

[0062] High confidence threshold verification defines the recovery threshold. With the downgrade threshold And satisfy ( This is the hysteresis difference, ranging from 0.1 to 0.2. It only applies when the current confidence score... When this happens, it indicates that the distributed algorithm has converged again.

[0063] Before restoring control, the mode switching module initiates a warm-start synchronization process by causing the distributed execution module to idle in the background for 3-5 cycles, pending its output. Once things are stable, control can be handed back to achieve a seamless transition.

[0064] By combining atomic privilege deprivation with a first-order hysteresis smoothing algorithm, the system ensures instantaneous takeover of control under abnormal conditions while eliminating the mechanical impact of torque mutations on hardware actuators. Combined with a high-threshold recovery mechanism based on hysteresis loops and dual verification via time windows, the system solves the problem of control oscillation under critical disturbance conditions.

[0065] Furthermore, it also includes an arbitration log recording module, which is used to record arbitration and switching events. The events include timestamps, triggering reasons, operational feedback and envelope boundaries, and provide a read-only interface for compliance auditing or predictive maintenance.

[0066] Specifically, in this embodiment, the arbitration log recording module employs a non-volatile memory independent of the system's running memory. The circular log area is divided into fixed-length recording slots. When the storage space is full, the module automatically overwrites the oldest record, ensuring that the latest system anomaly is preserved. A black box mechanism is included, where the module has power-loss protection, relying on onboard supercapacitors to maintain the final write voltage, ensuring that the last arbitration event is persisted before a catastrophic power outage occurs.

[0067] To meet compliance audit requirements, each record is encapsulated in a structured format. An event-driven logging strategy is employed to avoid the accumulation of massive amounts of invalid data.

[0068] Whenever the logic arbitration module issues a switching command or the mode switching execution module completes a restart, a full snapshot is immediately triggered. When the runtime feedback information enters the warning zone (less than 2% from the envelope boundary), the module initiates high-frequency temporary sampling and stores the extreme values ​​for that period in the log. At fixed intervals, the module forcibly records the current aging factor. The envelope interval serves as a benchmark for predictive maintenance.

[0069] The arbitration logging module provides a read-only interface protected by a physical switch or key. The regulator can export the logs to verify whether the FFU node has been operating strictly within the safety envelope over a period of time, demonstrating the controllability of air cleanliness. If the logs show that a node frequently triggers switches due to low confidence within a short period, even if its physical values ​​have not exceeded limits, the maintenance algorithm will still consider the information in the logs. The growth slope allows for the issuance of filter replacement or motor maintenance instructions in advance.

[0070] The arbitration log not only records events but also feeds back to the envelope dynamic correction module. By analyzing the statistical distribution of trigger causes in the log, the correction module finds that 90% of the switches are caused by the same boundary, and then automatically performs an envelope range shift operation to adapt to the trend changes in the production environment.

[0071] By adopting a recording strategy that combines event-driven and boundary warning, the accumulation of massive amounts of useless data is avoided, and full snapshot capture of abnormal system states is achieved. Through closed-loop linkage with predictive maintenance logic and envelope dynamic correction module, the probability of unplanned downtime in clean environment is effectively reduced and the service life of critical hardware is extended.

[0072] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A control system for a fan filter unit, characterized in that, include: The secure storage module is used to store the safety envelope range, physical safety limits, deterministic control logic, and degradation function parameter set preset based on the inherent parameters of the wind turbine unit. A distributed execution module is used to execute a self-organizing control algorithm based on the safe envelope interval to output an uncertain expected control quantity, wherein the expected control quantity is limited by the safe envelope interval; The status monitoring module is used to acquire the actual operating status and group coordination status of FFU nodes through sensors and output operating feedback information. The logic arbitration module is used to perform out-of-bounds detection and calculate the confidence score based on the operation feedback information, and output the arbitration result based on the out-of-bounds detection result and the confidence score; The envelope dynamic correction module is used to dynamically adjust the boundary value of the safe envelope interval by calling the degradation function parameter set according to the operation feedback information and the arbitration result, and the adjusted boundary value is limited by the physical security limit; The mode switching execution module is used to switch control permissions between self-organizing control and deterministic control according to the arbitration result, and has the function of permission restoration.

2. The control system for a fan filter unit according to claim 1, characterized in that, The secure storage module includes: The safety envelope is pre-set based on industry compliance standards or cleanroom class, including at least one of the upper and lower limits of differential pressure and upper and lower limits of air velocity; Physical safety limits are insurmountable boundary values ​​set based on the physical characteristics of the equipment or mandatory industry standards. The deterministic control logic consists of hard-coded execution rules in the controller's underlying layers, including at least one of constant speed, constant airflow, and fail-safe mode; the degradation function parameter set contains the mapping relationship between the degree of equipment aging and the envelope boundary adjustment, with the cumulative motor running time, shaft offset, and current RMS fluctuation rate as inputs.

3. The control system for a fan filter unit according to claim 1, characterized in that, In the distributed execution module, the step of the self-organizing control algorithm outputting the uncertain expected control quantity includes: Each FFU node interacts with information through a distributed neighborhood communication protocol, and achieves local self-organized balance control of pressure difference and wind speed based on a group cooperation algorithm; Each node independently outputs an uncertain expected control quantity based on the local interaction results, and this control quantity is limited by the current safe envelope interval.

4. The control system for a fan filter unit according to claim 3, characterized in that, The swarm collaboration algorithm includes a swarm self-organization rule based on local interaction, and a consensus protocol as a convergence constraint for the swarm rule. Each FFU node only exchanges state information with neighboring nodes, and independently calculates the adjustment step size and direction of the desired control quantity based on the local pressure difference or wind speed deviation, the number of neighboring nodes and its own aging degree, so that the swarm as a whole can achieve a distributed balance between pressure difference and wind speed.

5. The control system for a fan filter unit according to claim 1, characterized in that, In the status monitoring module, the step of outputting operational feedback information includes: Continuously collect the uncertain expected control quantity output by the distributed execution module and the collaborative state variables between each FFU node; The collected data is time-aligned and features are extracted to generate operation feedback information with time-series labels. The operation feedback information includes the device aging status inferred from the node operation data.

6. The control system for a fan filter unit according to claim 1, characterized in that, In the logic arbitration module, the step of outputting the arbitration result based on the boundary detection result and the confidence score includes: The operation feedback information is obtained and compared with the boundary value of the current safe envelope interval. If it exceeds the boundary, an out-of-bounds flag is triggered. Based on the degree of deviation between the operational feedback information and the envelope boundary, the control consistency index between nodes, and historical switching statistics, the confidence score of the current self-organizing control is calculated. If the out-of-bounds flag is triggered or the confidence score is lower than the preset threshold, the output will switch the arbitration result; otherwise, the output will maintain the arbitration result.

7. The control system for a fan filter unit according to claim 1, characterized in that, In the envelope dynamic correction module, the step of dynamically adjusting the boundary values ​​of the safe envelope interval includes: Extract equipment aging parameters from the operational feedback information and obtain the historical arbitration results output by the logic arbitration module; The degradation function parameter set is invoked, and the adjustment amount of each boundary value of the safety envelope interval is calculated according to the equipment aging parameters and historical arbitration results. Based on the adjustment amount, perform a shrinking or shifting operation on the current safe envelope interval to generate candidate boundary values; The candidate boundary value is compared with the preset physical security limit. If it exceeds the limit, it is limited to the limit value, and the updated security envelope is used for subsequent distributed execution and logical arbitration.

8. The control system for a fan filter unit according to claim 1, characterized in that, In the mode switching execution module, the steps for switching control permissions include: When the arbitration result is a switch, the self-organizing control authority is forcibly revoked and taken over by the deterministic control logic; The permission recovery function includes restoring self-organizing control permissions after a forced switch, when the running feedback information enters the safe envelope interval and the confidence score is higher than the recovery threshold.

9. The control system for a fan filter unit according to claim 1, characterized in that, It also includes an arbitration log recording module, which is used to record arbitration and switching events. The events include timestamps, trigger reasons, operation feedback and envelope boundaries, and provide a read-only interface for compliance auditing or predictive maintenance.

Images