Sensor fault tolerant control method for steam-water heat exchanger

Abstract

The application discloses a kind of sensor fault-tolerant control methods of steam-water heat exchanger, this method establishes the discrete data-driven steam-water heat exchanger system model containing disturbance, input constraint, designs dynamic event trigger mechanism, constructs radial basis neural network fault adaptive observer, obtains the water temperature estimation of adaptive observer under extended state, establishes fault detection mechanism, based on the design of sliding mode fault-tolerant controller of performance, consider the complex industrial working condition of sensor fault and disturbance coexistence under dynamic event trigger mechanism, based on the output fault-tolerant control decision scheme of sliding mode fault-tolerant controller, fault-tolerant control decision scheme is output to steam-water heat exchanger system model, realize steam-water heat exchanger online fault-tolerant control;It can quickly inhibit the influence of sensor fault on water temperature control, effectively reduce communication and computing resource consumption under the premise of guaranteeing control performance, solves the resource redundancy problem caused by traditional periodic sampling.

Description

Technical Field

[0001] This invention relates to the field of industrial process control and fault tolerance technology, and in particular to a sensor fault tolerance control method for a steam-water heat exchanger. Background Technology

[0002] Steam-water heat exchangers are core energy-efficient equipment in industrial sectors such as thermal power generation, chemical production, and district heating. The precise control of process parameters, such as outlet water temperature, directly determines the efficiency, safety, and energy consumption levels of industrial production. Temperature sensors, as a core component of the heat exchanger control system, are prone to drift and signal distortion under complex operating conditions. Without effective fault detection and tolerance mechanisms, control inaccuracies will occur, leading to decreased heat exchange efficiency, equipment downtime, and significant economic losses. Currently, most heat exchangers employ traditional time-triggered periodic sampling control modes. This not only results in significant redundant consumption of communication bandwidth and computing resources but also makes it difficult to respond promptly to sudden situations such as sensor failures and external disturbances. Furthermore, some fault alarm schemes rely on precise mathematical models of the system, which are difficult to adapt to the nonlinear characteristics of heat exchangers. This leads to insufficient system control accuracy and operational stability, failing to meet the high-efficiency and reliable control requirements of modern industry.

[0003] Therefore, existing technologies still need to be improved and enhanced. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this invention is to provide a sensor fault-tolerant control method for a steam-water heat exchanger, which aims to solve the problems of weak sensor fault tolerance and redundant communication resources in the existing control scheme.

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

[0006] A sensor fault-tolerant control method for a steam-water heat exchanger, characterized in that it includes:

[0007] Step 1: Based on the nonlinear dynamic characteristics of the steam-water heat exchanger and the sensor fault model, establish a discrete data-driven steam-water heat exchanger system model with disturbances and input constraints;

[0008] Step 2: Establish a dynamic event triggering mechanism for sensor data transmission, and selectively trigger sensor faults in the steam-water heat exchanger system model;

[0009] Step 3: Based on the adaptive parameter optimization and Lyapunov stability solution method, construct a radial basis neural network fault adaptive observer, and treat the disturbance as an extended state to obtain the water temperature estimate of the adaptive observer in the extended state.

[0010] Step 4: Obtain the estimation error based on the water temperature estimate of the adaptive observer, construct the fault detection threshold, establish the fault detection mechanism, and establish constraints on the fault-tolerant controller according to the descriptive performance of the steam-water heat exchanger, and design the sliding mode fault-tolerant controller.

[0011] Step 5: Considering the complex industrial conditions where sensor faults and disturbances coexist under the dynamic event triggering mechanism, output the fault-tolerant control decision scheme based on the sliding mode fault-tolerant controller, and output the fault-tolerant control decision scheme to the steam-water heat exchanger system model to realize online fault-tolerant control of the steam-water heat exchanger.

[0012] Furthermore, in step 1, the model of the steam-water heat exchanger system considering dual inputs and single outputs is represented as follows:

[0013] ;

[0014] in, , This represents the equivalent linear input corresponding to the steam flow rate and the cold water flow rate. , These are steam flow rates. and cold water flow rate The corresponding nonlinear function, This indicates the current measured process water temperature at the heat exchanger outlet. , These represent the outlet water temperatures measured at the previous sampling time and the two previous sampling times, respectively. , , , These represent the control signals applied to the heat exchanger actuator at the previous sampling time and the two previous sampling times for steam flow and cold water flow, respectively. A coefficient input for steam flow rate. The coefficient input for cold water flow rate. , , The coefficient representing the outlet process water temperature. It is a bounded external interference. It is a sensor fault signal.

[0015] Furthermore, in step 2, triggering conditions containing time-varying adjustment variables are set, and the timing of data transmission is adaptively determined based on the state deviation of the steam-water heat exchanger system to establish a dynamic event triggering mechanism; the set of instantaneous events triggered when a sensor failure occurs is set as... The event-triggered input increment error is By introducing dynamically updated variables, the dynamically updated variable at the next moment triggered by the event is represented as follows:

[0016] ;

[0017] in, and It is an adjustable parameter. Dynamically update variables at the trigger time;

[0018] The event triggering function is constructed as follows:

[0019] ;

[0020] in, It is a constant;

[0021] Based on the event triggering function, the event triggering time sequence is obtained as follows: And the triggering time satisfy:

[0022] ;

[0023] Where N represents an integer, This is the event triggering condition used to determine the timing of data transmission;

[0024] The current temperature signal is sent to the back-end observer and controller only when the change in the outlet water temperature of the steam-water heat exchanger meets the triggering conditions at the triggering moment; otherwise, the data from the previous moment remains unchanged.

[0025] Furthermore, when the steam-water heat exchanger system does not meet the triggering conditions, i.e.

[0026] ;

[0027] Substitute it To dynamically update variables, you need to make ,Require Then we get:

[0028] ;

[0029] Based on the prior knowledge of series convergence, it is necessary to ,at this time satisfy:

[0030] ;

[0031] When the steam-water heat exchanger system meets the triggering conditions:

[0032] ;

[0033] get:

[0034] ;

[0035] Regardless of whether the triggering conditions are met, the event triggering increment error of the heat exchanger system is bounded.

[0036] Furthermore, in step 3, a radial basis neural network fault adaptive observer is constructed based on the adaptive parameter optimization and Lyapunov stability solution method, and the disturbance is equivalent to an extended state to obtain the water temperature estimate of the adaptive observer in the extended state.

[0037] To address the dynamic uncertainties inherent in nonlinear systems, an adaptive observer is combined with a pseudo-Jacobi matrix estimator to achieve coordinated estimation of the system's unknown state and time-varying dynamics. The adaptive observer is expressed as:

[0038] ;

[0039] in, This is an estimate of the pseudo-Jacobi matrix. For the gain of the observer to be designed, For sensor output, The state of the adaptive observer at the previous time step. To control the amount of input change, when hour, , ;

[0040] The update equation for the pseudo-Jacobi matrix estimation algorithm under event triggering is as follows:

[0041] ;

[0042] in, Step size factor The event triggering function is represented as:

[0043] ;

[0044] To improve the estimator's ability to track variables, a reset mechanism is proposed as follows:

[0045] ;

[0046] When the estimated value deviates from the actual dynamic, the reset mechanism ensures the accuracy of the observation, enabling the observer to adapt to the uncertainty of heat exchanger load changes and operating condition fluctuations.

[0047] Furthermore, in step 3, based on the adaptive observer, the dynamic equations of the steam-water heat exchanger system are updated as follows:

[0048] ;

[0049] in, Step size factor It is a composite interference. Estimate the error for the observer;

[0050] definition ,Right now:

[0051] ;

[0052] in, The output is an estimated value. The value is an estimate of the composite interference; the estimation error is:

[0053] ;

[0054] in, To trigger the output, during the trigger interval, the output from the previous moment is maintained; during the trigger moment, the output is the one used at the trigger time. To output the estimated value;

[0055] Design a discrete extended state observer to estimate the disturbance. The observer has the following form:

[0056] ;

[0057] in, and The observer gain; the estimated error value of the composite interference. satisfy:

[0058] ;

[0059] in, .

[0060] Further, in step 4, the adaptive observer from step 3 first provides an estimated value for the outlet water temperature of the steam-water heat exchanger. This estimated value is then differed from the actual measured value to obtain the estimation error. Finally, based on the system's boundedness, a fault detection threshold is derived, and a fault detection mechanism is established. The estimation error is expressed as:

[0061] ;

[0062] Further recursion leads to:

[0063] ;

[0064] Based on the known conditions , , , , , , , These represent the bounded values ​​of interference, the bounded values ​​of pseudo-Jacobi matrix estimation error, the bounded values ​​of sensor fault, and the bounded values ​​of input, respectively.

[0065] Taking the boundedness condition into account, we can obtain the following by taking the absolute value of the above expression and scaling it:

[0066] ;

[0067] When feedback gain satisfy At that time, Monotonically decreasing, therefore:

[0068] ;

[0069] The fault detection threshold is expressed as:

[0070] ;

[0071] When satisfied At that time, a sensor malfunction was detected in the system.

[0072] Furthermore, in step 4, when At this time, the system is in a fault-free phase, and a set of input and output data at the moment of event triggering is collected. ,in, , The time when the fault occurred;

[0073] Based on event sampling data during the fault-free phase, the output of the radial basis function neural network is represented as:

[0074] ;

[0075] in, It is the system's fault-free input vector. It is a weight row vector, where s is the number of nodes in the neural network. These are the basis functions of the hidden layer; The Gaussian function is selected as follows:

[0076] ;

[0077] in, and These are the center and width of the i-th node, respectively. Weight vector satisfy:

[0078] ;

[0079] in, and These are the lower bound vector and the upper bound vector of the weights, respectively. The update of the weight vector is as follows:

[0080] ;

[0081] in, The step size; within the event trigger interval The weights are zero-order preserved, that is:

[0082] ;

[0083] Since updates are only made at the moment the event is triggered, an event-triggered approximation error is introduced:

[0084] ;

[0085] Among them, traditional approximation error Bounded, that is ;

[0086] Within the event interval, the network output at the latest trigger moment is used to approximate the fault-free output:

[0087] ;

[0088] The sensor fault function is represented as follows:

[0089] .

[0090] Furthermore, in step 4, constraints are established for the fault-tolerant controller based on the descriptive performance of the steam-water heat exchanger, and the time-varying performance function is expressed as:

[0091] ;

[0092] in, For decay rate, This is the steady-state value;

[0093] Limited by the physical properties of the actuator The control input that is constrained is defined as follows: :

[0094] ;

[0095] in, and These are the upper and lower limits for the input amplitude constraint;

[0096] The compensated tracking error is expressed as:

[0097] ;

[0098] in, This is a compensation signal introduced to address the input constraint problem caused by the physical limitations of the actuator; the tracking error is confined within new boundaries.

[0099] ;

[0100] Using the hyperbolic tangent function Error transformation is performed, and it is expressed as:

[0101] ;

[0102] in, The conversion error can be expressed as follows:

[0103] .

[0104] Furthermore, in step 4, constraints are established for the fault-tolerant controller based on the described performance of the steam-water heat exchanger, and a sliding mode fault-tolerant controller is designed.

[0105] Design a first-order sliding mode function for the steam-water heat exchanger system model in step 1. :

[0106] ;

[0107] in, These are the parameters to be designed. Equivalent control input. Represented as:

[0108] ;

[0109] in, This is an estimate of the pseudo-Jacobi matrix. This is an estimate of the sensor fault function. For observer gain, , Indicates the corresponding number The compensation amount for each input; It is a diagonal matrix. , control The attenuation rate, To describe the performance boundary function, Switch control input Represented as: ;

[0110] in, For a sign function, when hour, ,when hour, , These are the parameters to be designed, reflecting the control strength of the switching control.

[0111] The control input of the fault-tolerant controller is:

[0112] .

[0113] The technical solution adopted in this invention has the following beneficial effects:

[0114] (1) In view of the characteristics of steam-water heat exchangers being nonlinear, difficult to model, and susceptible to disturbances and sensor failures, a data-driven approach is adopted to design the controller, and descriptive performance control is combined to quantitatively constrain the tracking error, ensuring that the tracking error of the heat exchanger outlet water temperature always converges within the preset boundary, thus achieving strict guarantee of convergence speed and steady-state accuracy, and making the control performance more stable and more accurate.

[0115] (2) The radial basis neural network is used to realize the online estimation and approximation of sensor faults, which can quickly suppress the impact of sensor faults on water temperature control. A dynamic event triggering mechanism is designed to adaptively determine the data transmission time based on the system state. Under the premise of ensuring control performance, the consumption of communication and computing resources is effectively reduced, which solves the resource redundancy problem caused by traditional periodic sampling and is more suitable for resource-constrained industrial network control systems. Attached Figure Description

[0116] Figure 1 This is a flowchart illustrating a specific implementation of the present invention;

[0117] Figure 2 This is a diagram of the fault detection scheme of the present invention;

[0118] Figure 3 This is a schematic diagram of the system output of the present invention;

[0119] Figure 4 This is a schematic diagram of the fault estimation method of the present invention;

[0120] Figure 5 This is a schematic diagram of the event triggering interval of the present invention. Detailed Implementation

[0121] The present invention will now be further described with reference to the accompanying drawings.

[0122] A fault-tolerant control method for sensor faults in a steam-water heat exchanger, such as Figure 1 As shown, the control methods include:

[0123] Step 1: Based on the nonlinear dynamic characteristics of the steam-water heat exchanger and the sensor fault model, establish a discrete data-driven steam-water heat exchanger system model with disturbances and input constraints;

[0124] In step 1, the model of the steam-water heat exchanger system considering dual inputs and single outputs is represented as follows:

[0125] ;

[0126] in, , This represents the equivalent linear input corresponding to the steam flow rate and the cold water flow rate. , These are steam flow rates. and cold water flow rate The corresponding nonlinear function, This indicates the current measured process water temperature at the heat exchanger outlet. , These represent the outlet water temperatures measured at the previous sampling time and the two previous sampling times, respectively. , , , These represent the control signals applied to the heat exchanger actuator at the previous sampling time and the two previous sampling times for steam flow and cold water flow, respectively. A coefficient input for steam flow rate. The coefficient input for cold water flow rate. , , The coefficient representing the outlet process water temperature. It is a bounded external interference. It is a sensor fault signal.

[0127] Step 2: Establish a dynamic event triggering mechanism for sensor data transmission, and selectively trigger sensor faults in the steam-water heat exchanger system model;

[0128] In step 2, triggering conditions containing time-varying adjustment variables are set, and the timing of data transmission is adaptively determined based on the state deviation of the steam-water heat exchanger system, establishing a dynamic event triggering mechanism; the set of instantaneous events triggered when a sensor failure occurs is set as... The event-triggered input increment error is By introducing dynamically updated variables, the dynamically updated variable at the next moment triggered by the event is represented as follows:

[0129] ;

[0130] in, and It is an adjustable parameter. Dynamically update variables at the trigger time;

[0131] The event triggering function is constructed as follows:

[0132] ;

[0133] in, It is a constant;

[0134] Based on the event triggering function, the event triggering time sequence is obtained as follows: And the triggering time satisfy:

[0135] ;

[0136] Where N represents an integer, This is the event triggering condition used to determine the timing of data transmission;

[0137] The current temperature signal is sent to the back-end observer and controller only when the change in the outlet water temperature of the steam-water heat exchanger meets the triggering conditions at the triggering moment; otherwise, the data from the previous moment remains unchanged.

[0138] Furthermore, when the steam-water heat exchanger system does not meet the triggering conditions, i.e.

[0139] ;

[0140] Substitute it To dynamically update variables, you need to make ,Require Then we get:

[0141] ;

[0142] Based on the prior knowledge of series convergence, it is necessary to ,at this time satisfy:

[0143] ;

[0144] When the steam-water heat exchanger system meets the triggering conditions:

[0145] ;

[0146] get:

[0147] ;

[0148] Regardless of whether the triggering conditions are met, the event triggering increment error of the heat exchanger system is bounded.

[0149] Step 3: Based on the adaptive parameter optimization and Lyapunov stability solution method, construct a radial basis neural network fault adaptive observer, and treat the disturbance as an extended state to obtain the water temperature estimate of the adaptive observer in the extended state.

[0150] To address the dynamic uncertainties inherent in nonlinear systems, an adaptive observer is combined with a pseudo-Jacobi matrix estimator to achieve coordinated estimation of the system's unknown state and time-varying dynamics. The adaptive observer is expressed as:

[0151] ;

[0152] in, This is an estimate of the pseudo-Jacobi matrix. For the gain of the observer to be designed, For sensor output, The state of the adaptive observer at the previous time step. To control the amount of input change, when hour, , ;

[0153] The update equation for the pseudo-Jacobi matrix estimation algorithm under event triggering is as follows:

[0154] ;

[0155] in, Step size factor The event triggering function is represented as:

[0156] ;

[0157] To improve the estimator's ability to track variables, a reset mechanism is proposed as follows:

[0158] ;

[0159] When the estimated value deviates from the actual dynamic, the reset mechanism ensures the accuracy of the observation, enabling the observer to adapt to the uncertainty of heat exchanger load changes and operating condition fluctuations.

[0160] Based on the adaptive observer, the dynamic equations of the steam-water heat exchanger system are updated as follows:

[0161] ;

[0162] in, Step size factor It is a composite interference. Estimate the error for the observer;

[0163] definition ,Right now:

[0164] ;

[0165] in, The output is an estimated value. This is an estimate of the composite interference. The estimation error is:

[0166] ;

[0167] in, To trigger the output, during the trigger interval, the output from the previous moment is maintained; during the trigger moment, the output is the one used at the trigger time. To output the estimated value;

[0168] Design a discrete extended state observer to estimate the disturbance. The observer has the following form:

[0169] ;

[0170] in, and The observer gain. The estimated error value of the composite interference. satisfy:

[0171] ;

[0172] in, .

[0173] Step 4: Obtain the estimation error based on the water temperature estimate of the adaptive observer, construct the fault detection threshold, establish the fault detection mechanism, and establish constraints on the fault-tolerant controller according to the descriptive performance of the steam-water heat exchanger, and design the sliding mode fault-tolerant controller.

[0174] First, the adaptive observer in step 3 provides an estimated value for the outlet water temperature of the steam-water heat exchanger. This estimated value is then compared to the actual measured value to obtain the estimation error. Next, based on the system's boundedness, a fault detection threshold is derived, and a fault detection mechanism is established. The estimation error is expressed as:

[0175] ;

[0176] Further recursion leads to:

[0177] ;

[0178] Based on the known conditions , , , , , , , These represent the bounded values ​​of interference, the bounded values ​​of pseudo-Jacobi matrix estimation error, the bounded values ​​of sensor fault, and the bounded values ​​of input, respectively.

[0179] Taking the boundedness condition into account, we can obtain the following by taking the absolute value of the above expression and scaling it:

[0180] ;

[0181] When feedback gain satisfy At that time, Monotonically decreasing, therefore:

[0182] ;

[0183] The fault detection threshold is expressed as:

[0184] ;

[0185] When satisfied At that time, a sensor malfunction was detected in the system.

[0186] when At this time, the system is in a fault-free phase, and a set of input and output data at the moment of event triggering is collected. ,in, , The time when the fault occurred;

[0187] Based on event sampling data during the fault-free phase, the output of the radial basis function neural network is represented as:

[0188] ;

[0189] in, It is the system's fault-free input vector. It is a weight row vector, where s is the number of nodes in the neural network. These are the basis functions of the hidden layer; The Gaussian function is selected as follows:

[0190] ;

[0191] in, and These are the center and width of the i-th node, respectively. Weight vector satisfy:

[0192] ;

[0193] in, and These are the lower bound vector and the upper bound vector of the weights, respectively. The update of the weight vector is as follows:

[0194] ;

[0195] in, The step size; within the event trigger interval The weights are zero-order preserved, that is:

[0196] ;

[0197] Since updates are only made at the moment the event is triggered, an event-triggered approximation error is introduced:

[0198] ;

[0199] Among them, traditional approximation error Bounded, that is ;

[0200] Within the event interval, the network output at the latest trigger moment is used to approximate the fault-free output:

[0201] ;

[0202] The sensor fault function is represented as follows:

[0203] .

[0204] Based on the descriptive performance of the steam-water heat exchanger, constraints are established for the fault-tolerant controller, and the time-varying performance function is expressed as:

[0205] ;

[0206] in, For decay rate, This is the steady-state value;

[0207] Limited by the physical properties of the actuator The control input that is constrained is defined as follows: :

[0208] ;

[0209] in, and These are the upper and lower limits for the input amplitude constraint;

[0210] The compensated tracking error is expressed as:

[0211] ;

[0212] in, This is a compensation signal introduced to address the input constraint problem caused by the physical limitations of the actuator; the tracking error is confined within new boundaries.

[0213] ;

[0214] Using the hyperbolic tangent function Error transformation is performed, and it is expressed as:

[0215] ;

[0216] in, The conversion error can be expressed as follows:

[0217] .

[0218] Based on the performance description of the steam-water heat exchanger, constraints are established for the fault-tolerant controller, and a sliding mode fault-tolerant controller is designed.

[0219] Design a first-order sliding mode function for the steam-water heat exchanger system model in step 1. :

[0220] ;

[0221] in, These are the parameters to be designed. Equivalent control input. Represented as:

[0222] ;

[0223] in, This is an estimate of the pseudo-Jacobi matrix. This is an estimate of the sensor fault function. For observer gain, , Indicates the corresponding number The compensation amount for each input; It is a diagonal matrix. , control The attenuation rate, To describe the performance boundary function, Switch control input Represented as:

[0224] ;

[0225] in, For a sign function, when hour, ,when hour, , These are the parameters to be designed, reflecting the control strength of the switching control.

[0226] The control input of the fault-tolerant controller is:

[0227] .

[0228] Step 5: Considering the complex industrial conditions where sensor faults and disturbances coexist under the dynamic event triggering mechanism, output the fault-tolerant control decision scheme based on the sliding mode fault-tolerant controller, and output the fault-tolerant control decision scheme to the steam-water heat exchanger system model to realize online fault-tolerant control of the steam-water heat exchanger.

[0229] The following simulation experiment is conducted using specific parameter values, based on the method described above:

[0230] In this embodiment, in step 1, the sensor fault model is considered as follows:

[0231] ;

[0232] in, This is an unknown sensor fault signal. Under ideal fault-free operating conditions, .

[0233] Typically, due to sensor malfunction, the signal acquired by the sensor often deviates from the actual system output. Assume the system operates within a finite interval [1, ...]. The interior is intact, but since the first... From a certain moment onwards, the sensor malfunctions, and the sensor fault signal is obtained. .

[0234] The coefficient for steam flow input is The coefficient for cold water flow input is The coefficient of the outlet process water temperature is , , External interference .

[0235] In this embodiment, in step 2, a triggering condition containing time-varying adjustment variables is set, and the timing of data transmission is adaptively determined based on the state deviation of the steam-water heat exchanger system to establish a dynamic event triggering mechanism.

[0236] The current temperature signal is sent to the back-end observer and controller only when the change in the outlet water temperature of the steam-water heat exchanger meets the triggering conditions at the triggering moment; otherwise, the data from the previous moment remains unchanged. This reduces the consumption of communication resources while ensuring the accuracy of water temperature control.

[0237] Tuning is performed based on the boundedness of the event-triggered incremental error and the heat exchanger response speed; the trigger parameter is selected as follows. , This ensures that the triggering frequency matches the actual dynamics of the heat exchanger, reducing the number of triggers when the water temperature is stable and increasing the triggering frequency when the water temperature changes abruptly or a fault occurs, thus balancing control performance and communication savings.

[0238] In this embodiment, in step 3, based on the adaptive parameter optimization and Lyapunov stability solution method, the step size factor can be obtained according to the least squares cost function. The regularization coefficient is According to Lyapunov's stability theorem, the observer gain Based on prior knowledge of the system's input and output sensitivity, the following was selected: The Lipschitz condition guarantees that the estimate is bounded.

[0239] When the estimated value deviates from the actual dynamic, the reset mechanism ensures the accuracy of the observation, enabling the observer to adapt to the uncertainty of heat exchanger load changes and operating condition fluctuations.

[0240] A discrete extended state observer is designed based on an adaptive observer to estimate disturbances, satisfying the stability constraint that the eigenvalues ​​of the error dynamic matrix lie within the unit circle. The parameters of the discrete extended state observer are determined by pole placement. , The upper bound of the composite interference is This ensures that disturbance estimation is fast and stable, providing a basis for disturbance compensation for subsequent fault-tolerant control and improving the robustness of the heat exchanger under complex operating conditions.

[0241] In this embodiment, in step 4, according to the fault detection mechanism, when At this time, the system is in a fault-free phase, and a set of input and output data at the moment of event triggering is collected. ,in, Take the time of the fault occurrence as The traditional approximation error satisfies s=8 is the number of nodes in the neural network. and These are the lower bound vector and the upper bound vector of the weights, respectively, with a step size of [missing value]. .

[0242] Once the fault detection mechanism determines that a fault has occurred, a radial basis function neural network is used to estimate and approximate the sensor fault signal online. The fault estimate is constructed by the difference between the neural network output and the actual output, and the magnitude of the sensor fault is obtained in real time.

[0243] Based on the performance characteristics of the steam-water heat exchanger, constraints are established for the fault-tolerant controller. By pre-setting a time-varying performance function, the convergence boundary of the water temperature tracking error is set as follows: This allows the error to converge rapidly from its initial value to the steady-state allowable range. The convergence rate is... , .

[0244] Simultaneously considering the physical constraints of the actuator valve opening, the amplitude of the control input is limited to prevent the valve from operating beyond its range. After error transformation, the constrained problem is transformed into an unconstrained control problem, which facilitates controller design and ensures that the heat exchanger water temperature control always meets the described performance indicators.

[0245] Based on the performance characteristics of the steam-water heat exchanger, constraints are established for the fault-tolerant controller, and a sliding mode fault-tolerant controller is designed, along with its control parameters. Switching control strength parameters .

[0246] In this embodiment, in step 5, the fault estimate and disturbance estimate are directly embedded into the control law for feedforward compensation to offset the impact of sensor faults and external disturbances on the outlet water temperature. Finally, the controller output acts on the steam valve and the cold water valve, enabling the heat exchanger to stably track the desired water temperature even under conditions of sensor faults, load fluctuations, and input limitations.

[0247] The fault detection scheme described in steps 2-4 of this invention is as follows: Figure 2 As shown, firstly, dynamic event triggering conditions are designed; secondly, an adaptive observer is constructed within the event triggering mechanism framework; and finally, the estimation error is directly used for fault detection decision-making, making the method more concise and intuitive.

[0248] The system output diagram of the steam-water heat exchanger is shown below. Figure 3 As shown in the figure, during the fault-free phase, the output closely tracks the desired trajectory with no significant deviation. After the fault occurs, the output initially deviates momentarily due to the sensor malfunction, but quickly stabilizes and re-aligns with the desired trajectory.

[0249] Fault estimation, such as Figure 4 As shown in the figure, from start, In other words, the fault was successfully detected.

[0250] Event triggering time intervals such as Figure 5 As shown, from Figure 5 It can be seen that the system's sampling method is non-periodic. Compared with the traditional periodic sampling method, the event-triggered mechanism has a longer transmission time interval, which can effectively reduce signal transmission and reduce communication burden.

Claims

1. A sensor fault-tolerant control method for a steam-water heat exchanger, characterized in that, include: Step 1: Based on the nonlinear dynamic characteristics of the steam-water heat exchanger and the sensor fault model, establish a discrete data-driven steam-water heat exchanger system model with disturbances and input constraints; Step 2: Establish a dynamic event triggering mechanism for sensor data transmission, and selectively trigger sensor faults in the steam-water heat exchanger system model; Step 3: Based on the adaptive parameter optimization and Lyapunov stability solution method, construct a radial basis neural network fault adaptive observer, and treat the disturbance as an extended state to obtain the water temperature estimate of the adaptive observer in the extended state. Step 4: Obtain the estimation error based on the water temperature estimate of the adaptive observer, construct the fault detection threshold, establish the fault detection mechanism, and establish constraints on the fault-tolerant controller according to the descriptive performance of the steam-water heat exchanger, and design the sliding mode fault-tolerant controller. Step 5: Considering the complex industrial conditions where sensor faults and disturbances coexist under the dynamic event triggering mechanism, output the fault-tolerant control decision scheme based on the sliding mode fault-tolerant controller, and output the fault-tolerant control decision scheme to the steam-water heat exchanger system model to realize online fault-tolerant control of the steam-water heat exchanger.

2. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 1, characterized in that, In step 1, the model of the steam-water heat exchanger system considering dual inputs and single outputs is represented as follows: ； in, , This represents the equivalent linear input corresponding to the steam flow rate and the cold water flow rate. , These are steam flow rates. and cold water flow rate The corresponding nonlinear function, This indicates the current measured process water temperature at the heat exchanger outlet. , These represent the outlet water temperatures measured at the previous sampling time and the two previous sampling times, respectively. , , , These represent the control signals applied to the heat exchanger actuator at the previous sampling time and the two previous sampling times for steam flow and cold water flow, respectively. A coefficient input for steam flow rate. The coefficient input for cold water flow rate. , , The coefficient representing the outlet process water temperature. It is a bounded external interference. It is a sensor fault signal.

3. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 1, characterized in that, In step 2, triggering conditions containing time-varying adjustment variables are set, and the timing of data transmission is adaptively determined based on the state deviation of the steam-water heat exchanger system, establishing a dynamic event triggering mechanism; the set of instantaneous events triggered when a sensor failure occurs is set as... The event-triggered input increment error is By introducing dynamically updated variables, the dynamically updated variable at the next moment triggered by the event is represented as follows: ； in, and It is an adjustable parameter. Dynamically update variables at the trigger time; The event triggering function is constructed as follows: ； in, It is a constant; Based on the event triggering function, the event triggering time sequence is obtained as follows: And the triggering time satisfy: ； Where N represents an integer, This is the event triggering condition used to determine the timing of data transmission; The current temperature signal is sent to the back-end observer and controller only when the change in the outlet water temperature of the steam-water heat exchanger meets the triggering conditions at the triggering moment; otherwise, the data from the previous moment remains unchanged.

4. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 3, characterized in that, When the steam-water heat exchanger system does not meet the triggering conditions, i.e. ； Substitute it To dynamically update variables, you need to make ,Require Then we get: ； Based on the prior knowledge of series convergence, it is necessary to ,at this time satisfy: ； When the steam-water heat exchanger system meets the triggering conditions: ； get: ； Regardless of whether the triggering conditions are met, the event triggering increment error of the heat exchanger system is bounded.

5. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 1, characterized in that, In step 3, a radial basis function neural network fault adaptive observer is constructed based on the adaptive parameter optimization and Lyapunov stability solution method, and the disturbance is equivalent to an extended state to obtain the water temperature estimate of the adaptive observer in the extended state. To address the dynamic uncertainties inherent in nonlinear systems, an adaptive observer is combined with a pseudo-Jacobi matrix estimator to achieve coordinated estimation of the system's unknown state and time-varying dynamics. The adaptive observer is expressed as: ； in, This is an estimate of the pseudo-Jacobi matrix. For the gain of the observer to be designed, For sensor output, The state of the adaptive observer at the previous time step. To control the amount of input change, when hour, , ; The update equation for the pseudo-Jacobi matrix estimation algorithm under event triggering is as follows: ； in, Step size factor The event triggering function is represented as: ； To improve the estimator's ability to track variables, a reset mechanism is proposed as follows: ； When the estimated value deviates from the actual dynamic, the reset mechanism ensures the accuracy of the observation, enabling the observer to adapt to the uncertainty of heat exchanger load changes and operating condition fluctuations.

6. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 5, characterized in that, In step 3, based on the adaptive observer, the dynamic equations of the steam-water heat exchanger system are updated as follows: ； in, Step size factor It is a composite interference. Estimate the error for the observer; definition ,Right now: ； in, The output is an estimated value. The value is an estimate of the composite interference; the estimation error is: ； in, To trigger the output, during the trigger interval, the output from the previous moment is maintained; during the trigger moment, the output is the one used at the trigger time. To output the estimated value; Design a discrete extended state observer to estimate the disturbance. The observer has the following form: ； in, and The observer gain; the estimated error value of the composite interference. satisfy: ； in, .

7. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 1, characterized in that, In step 4, the adaptive observer from step 3 first provides an estimated value for the outlet water temperature of the steam-water heat exchanger. This estimated value is then compared to the actual measured value to obtain the estimation error. Next, based on the system's boundedness, a fault detection threshold is derived, and a fault detection mechanism is established. The estimation error is expressed as: ； Further recursion leads to: ； Based on the known conditions , , , , , , , These represent the bounded values ​​of interference, the bounded values ​​of pseudo-Jacobi matrix estimation error, the bounded values ​​of sensor fault, and the bounded values ​​of input, respectively. Taking the boundedness condition into account, we can obtain the following by taking the absolute value of the above expression and scaling it: ； When feedback gain satisfy At that time, Monotonically decreasing, therefore: ； The fault detection threshold is expressed as: ； When satisfied At that time, a sensor malfunction was detected in the system.

8. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 7, characterized in that, In step 4, when At this time, the system is in a fault-free phase, and a set of input and output data at the moment of event triggering is collected. ,in, , The time when the fault occurred; Based on event sampling data during the fault-free phase, the output of the radial basis function neural network is represented as: ； in, It is the system's fault-free input vector. It is a weight row vector, where s is the number of nodes in the neural network. These are the basis functions of the hidden layer; The Gaussian function is selected as follows: ； in, and These are the center and width of the i-th node, respectively; Weight vector satisfy: ； in, and These are the lower bound vector and the upper bound vector of the weights, respectively. The weight vector is updated as follows: ； in, The step size; within the event trigger interval The weights are zero-order preserved, that is: ； Since updates are only made at the moment the event is triggered, an event-triggered approximation error is introduced: ； Among them, traditional approximation error Bounded, that is ; Within the event interval, the network output at the latest trigger moment is used to approximate the fault-free output: ； The sensor fault function is represented as follows: 。 9. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 1, characterized in that, In step 4, constraints are established for the fault-tolerant controller based on the descriptive performance of the steam-water heat exchanger, and the time-varying performance function is expressed as: ； in, For decay rate, This is the steady-state value; Limited by the physical properties of the actuator The control input that is constrained is defined as follows: : ； in, and The upper and lower limits of the input amplitude constraint are used; the compensated tracking error is expressed as: ； in, This is a compensation signal introduced to address the input constraint problem caused by the physical limitations of the actuator; the tracking error is confined within new boundaries. ； Using the hyperbolic tangent function Error transformation is performed, and it is expressed as: ； in, The conversion error can be expressed as follows: 。 10. The sensor fault-tolerant control method for a steam-water heat exchanger according to claim 9, characterized in that, In step 4, constraints are established for the fault-tolerant controller based on the performance description of the steam-water heat exchanger, and a sliding mode fault-tolerant controller is designed. Design a first-order sliding mode function for the steam-water heat exchanger system model in step 1. : ； in, These are the parameters to be designed; Equivalent control input Represented as: ； in, This is an estimate of the pseudo-Jacobi matrix. This is an estimate of the sensor fault function. For observer gain, , Indicates the corresponding number The compensation amount for each input; It is a diagonal matrix. , control The attenuation rate, To describe the performance boundary function, ; Switch control input Represented as: ； in, For a sign function, when hour, ,when hour, , These are the parameters to be designed, reflecting the control strength of the switching control. The control input of the fault-tolerant controller is: 。

Images