Industrial interconnection system sensor fault tolerance control method based on artificial intelligence

Abstract

The invention aims to provide an industrial interconnection system sensor fault tolerance control method based on artificial intelligence, and the method comprises the steps: building a low-constraint switching architecture with a direction perception capability, and removing the unnecessary shutdown waiting constraint in a conventional method; a state observer and a self-adaptive compensation mechanism are utilized to reconstruct a real signal through a software algorithm under a sensor fault; predetermined performance control is realized through an error preprocessing mechanism; an unknown item of the system is approached by using an artificial intelligence technology based on a fuzzy logic system; and a dynamic surface signal processing technology is introduced, so that the computing power demand of the controller is greatly reduced. Therefore, efficient, safe and low-cost operation of the industrial interconnection system under complex working conditions is realized.

Description

Technical Field

[0001] This disclosure relates to the fields of industrial automation control technology and artificial intelligence application technology, and in particular to a low-constraint operating condition switching control method applicable to complex industrial scenarios such as multi-joint robotic arms and interconnected power systems, which has sensor fault tolerance control and predetermined performance capabilities. Background Technology

[0002] In modern intelligent manufacturing and the energy industry, large-scale collaborative interconnected systems such as multi-jointed robotic arms, flexible production lines, and multi-regional interconnected power grids play a central role. These systems are typically composed of multiple coupled physical components (subsystems) and need to frequently switch between various drastically different operating modes. For example, industrial robots switch between "high-speed handling" and "high-precision assembly" modes, or microgrids switch between "grid-connected power generation" and "independent power supply" modes. Due to the complex dynamic coupling between the various components, such frequent abrupt changes in operating conditions can easily cause mechanical vibrations or energy fluctuations in the equipment.

[0003] To prevent equipment failures caused by switching operating conditions, existing industrial control systems typically employ conservative delay-based switching strategies (such as average dwell time mechanisms). This strategy mandates that equipment pause or maintain its current state for a relatively long fixed period after each switching condition to allow the system to stabilize. However, this "forced waiting" mechanism severely restricts production efficiency. In high-cycle production lines, even when equipment switches from "vigorous movement" to "stable operation," the controller still forces the equipment to wait, increasing ineffective operation time and preventing the production cycle from meeting the requirements of high-speed production lines. Existing technologies lack intelligent matching of operation time based on the "direction of operating condition switching," and cannot flexibly adjust the waiting time according to the actual operating status, resulting in the equipment's performance not being fully utilized.

[0004] Furthermore, such industrial interconnected systems typically operate in harsh environments with high temperatures, vibrations, or severe electromagnetic interference. Core sensors, such as position encoders and current transformers, are highly susceptible to signal drift or gain failure after prolonged service. The drawback of existing technologies lies in the fact that traditional precision control algorithms often assume completely reliable sensor data. If a sensor experiences a hidden fault, the controller may output incorrect drive commands based on distorted feedback signals, potentially leading to robotic arm collisions, motor overheating and burnout, or grid instability and disconnection, resulting in serious industrial safety accidents.

[0005] In addition to the aforementioned stability and fault tolerance issues, high-precision industrial operations impose extremely stringent constraints on the dynamic response of the system. While traditional control methods can guarantee the eventual stability of the system, they cannot quantify and constrain instantaneous errors during the adjustment process, directly leading to uncontrollable production cycle time. Therefore, it is necessary to design a control strategy that can strictly limit the system state within a "preset safety envelope," i.e., predetermined performance control, to ensure that regardless of how drastically the operating conditions change, the dynamic error of the system is always forcibly constrained within an allowable physical range.

[0006] On the other hand, complex industrial interconnected systems often face significant model uncertainties in actual operation, such as unknown physical characteristics like friction and wear. Traditional adaptive control relies on linear parameterization assumptions, making it difficult to cope with unstructured uncertainties. To address this issue, artificial intelligence techniques based on general approximation principles, such as neural networks and fuzzy logic systems, have been widely introduced for online learning and approximation of these unknown nonlinear dynamics.

[0007] Meanwhile, for such multivariable, strongly coupled systems, the backstepping method is often used in academia for controller design. However, in engineering implementation, as the number of system modules increases, the amount of differential computation required by the backstepping method expands exponentially, a problem known as the "differential explosion." This necessitates that the controller be equipped with expensive, high-performance computing chips, significantly increasing hardware costs. Furthermore, on general-purpose embedded platforms with limited computing power (such as ARM or DSP), it is difficult to run such complex algorithms in real time.

[0008] In summary, there is an urgent need for an intelligent control method that can overcome the conservative constraints of traditional switching strategies, has sensor fault tolerance capabilities, can achieve predetermined performance control, and has low computational load and is easy to implement in engineering. Summary of the Invention

[0009] The purpose of this disclosure is to provide a fault-tolerant control method for sensors in industrial interconnected systems based on artificial intelligence, so as to solve at least one technical problem in the prior art.

[0010] The technical solution disclosed herein is:

[0011] A fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence, comprising:

[0012] Transform any industrial interconnected system in the physical world into a digital state equation that the controller can recognize, and establish a sensor fault model based on signal attenuation caused by sensor aging.

[0013] Construct a non-singular error scalar transformation function to convert the constrained tracking error into an unconstrained variable; and construct a barrier Lyapunov function to ensure that the transformation error state always remains within the constraint bounds.

[0014] Artificial intelligence technology based on fuzzy logic system is used to approximate unknown nonlinear continuous functions, transforming complex unknown nonlinear functions into the product of known basis functions and unknown weight vectors;

[0015] Based on the unmeasurability of system state and the characteristics of sensor gain failure, a state observer and a fault self-healing module are constructed.

[0016] Based on backstepping and dynamic surface control techniques, combined with the Lyapunov function of tangent obstacle, a controller is recursively constructed.

[0017] A logic judgment module based on the direction of operating condition switching is embedded in the controller to obtain the judgment result; and based on the judgment result, various types of operating condition mode switching are completed to improve fault tolerance.

[0018] The process of transforming any industrial interconnected system in the physical world into a digital state equation recognizable by the controller, and establishing a sensor fault model based on signal attenuation caused by sensor aging, includes:

[0019] For an industrial interconnected system consisting of M coupled subsystems, and considering the possibility of system switching conditions, a feedback nonlinear mathematical model for the i-th subsystem is established:

[0020]

[0021] in, This refers to the system status; For switching signals; This refers to the nonlinear physical characteristics existing within the system. To characterize the strong interconnection and coupling that exists between subsystems. It is a distractor. For control input; ; Indicates time; Represents the system's output vector; For the first The order of each subsystem.

[0022] Based on the signal attenuation caused by sensor aging in industrial settings, a sensor fault model is established:

[0023] ;

[0024] in, This is the actual measurement value from the sensor. The unknown sensor failure factor;

[0025] Introducing fault estimation parameters , will output the actual data Represented as: ;

[0026] in, For sensor measurement output, and These are the approximate errors and approximate values ​​of the fault parameters, respectively. Indicates time; This refers to the system sensor failure time.

[0027] The construction of the non-singular error scalar transformation function, which transforms the constrained tracking error into an unconstrained variable, includes:

[0028] Constructing a polynomial buffer term Non-singular error scalar transformation function :

[0029] ;

[0030] in, and , and For design parameters; ; t represents time, It is a positive design constant; The scheduled convergence time set by the user.

[0031] The construction of the barrier Lyapunov function ensures that the transformation error state always remains within the constraint limits, including:

[0032] Constructing tangent-type barrier Lyapunov functions to achieve full-state constraints Inside:

[0033] ;

[0034] in, This serves as the state error constraint boundary; These are unconstrained variables; It is a user-defined state constraint bound.

[0035] The method of approximating unknown nonlinear continuous functions using artificial intelligence technology based on fuzzy logic systems transforms complex unknown nonlinear functions into the product of known basis functions and unknown weight vectors, including:

[0036] Using the universal approximation property of fuzzy logic systems, fuzzy rules are constructed:

[0037] rule :if yes And... and yes ,So yes ;in For fuzzy rules;

[0038] The unknown nonlinear function is represented as:

[0039] ;

[0040] in, , The ideal fuzzy weight vector; For fuzzy basis function vectors; This represents the fuzzy approximation error.

[0041] Based on the unmeasurability of system state and the characteristics of sensor gain failure, a state observer is constructed, including:

[0042] Regarding the problem of unpredictable system state, when the first When a switching system is activated, the following observer is constructed to estimate the system state. :

[0043] ;

[0044] in, The derivative vector of the state estimate; It is a vector of state estimates; The observer gain matrix; ; ; It is the actual measurement value of the sensor; For the observer's correction term; These are estimated values ​​of sensor fault parameters; To estimate the error; ; This is an estimate of the ideal fuzzy weight vector; For fuzzy basis function vectors; To control the input; let and , making It is a strict Hurwitz matrix, and there exists a matrix... satisfy ,in It is a positive definite symmetric matrix.

[0045] The self-healing module includes:

[0046] Get fault parameters online The following adaptive update law is designed:

[0047] ;

[0048] in, For adaptive gain; ; It is a positive constant; It is the actual measurement value of the sensor; It is a normal number.

[0049] The controller, based on backstepping and dynamic surface control techniques and combined with the tangent obstacle Lyapunov function, is recursively constructed, including:

[0050] Combining the steps of constructing a non-singular error scalar transformation function to convert the constrained tracking error into an unconstrained variable, and the step of constructing a state observer based on the unmeasurability of the system state and the characteristics of sensor gain failure, the first-level transformation error surface is defined. We construct the barrier Lyapunov function and design virtual control and adaptive laws using backstepping techniques and Young's inequality.

[0051] Introduce a first-order low-pass filter: ;in, A positive constant. This is the filtered virtual control signal; For virtual controllers, the errors at each subsequent stage are: ;

[0052] Based on the Level error surface and barrier Lyapunov function, deriving practical control input and adaptive law:

[0053]

[0054] .

[0055] in, , and These are design parameters, and the control law simultaneously compensates for the effects of fuzzy approximation error, dynamic surface filtering error, and interconnection terms.

[0056] The controller incorporates a logic judgment module based on the direction of operating condition switching to obtain a judgment result; and based on the judgment result, it completes various types of operating condition mode switching, including:

[0057] Assume there exists a set of switching subsystems. Define a binary mapping. , and ,set up For subsystem families The system switching signal must meet the following dwell time conditions:

[0058] ;

[0059] in, The average residence time for binary dependencies; This is the initial time; T is the termination time; Represents the time interval Total runtime; It is a positive constant; Represents the time interval Total number of switches.

[0060] Assume there exists an element First, define an operator to collect the first element of the ordered pairs of switches in the group containing a specific switching subsystem:

[0061] .

[0062] For the A subsystem, defining a quadratic Lyapunov function. And its derivative satisfies the following dissipation inequality:

[0063] ;in, The attenuation rate; For constant terms;

[0064] For any switching pair The Lyapunov function satisfies the following jump condition:

[0065] ;in, ; ; It can be calculated; for The largest eigenvalue; yes The smallest eigenvalue.

[0066] Attenuation rate Defined as:

[0067] ;

[0068] constant term Defined as ;

[0069] in, for eigenvalues; It can be calculated according to claim 6; It is a matrix The largest eigenvalue; It is a positive constant; All of these are design parameters.

[0070] The beneficial effects of this disclosure include at least the following:

[0071] The fault-tolerant control method for sensor faults in industrial interconnection systems based on artificial intelligence disclosed herein is based on a low-constraint switching architecture with working condition direction perception. By intelligently identifying the switching path and matching the corresponding dwell time, it eliminates unnecessary long-term downtime constraints in traditional strategies, significantly improving equipment production cycle time while ensuring system stability. Furthermore, it constructs an active self-healing fault-tolerant mechanism and non-singular error transformation for sensor faults, utilizing a state observer and adaptive compensation law to reconstruct the failure signal online. This ensures that the equipment can maintain high-precision operation without emergency shutdown under sensor gain failure or drift conditions. Simultaneously, combined with non-singular error transformation technology, it ensures that the system can still achieve the predetermined performance control target under fault conditions. Moreover, this disclosure is adaptable to lightweight collision avoidance controllers with low-computing-power chips, introduces dynamic surface filtering technology to significantly reduce the algorithm's computational power requirements, and establishes virtual safety boundaries using obstacle functions to prevent physical over-limit accidents such as mechanical collisions from the algorithm's underlying layer. It has the advantages of predetermined performance control, low computational load, and ease of engineering implementation. Attached Figure Description

[0072] Figure 1 This is the control design flowchart of this disclosure;

[0073] Figure 2 The switching signal and tracking error changes of subsystem 1 under group 1 in the low-constraint handover implementation case of this disclosure;

[0074] Figure 3 The switching signal and tracking error changes of subsystem 1 under group 2 in the implementation case of this disclosure under low-constraint switching;

[0075] Figure 4 The switching signal and tracking error changes of subsystem 1 in group 3 under the low-constraint handover of the implementation case of this disclosure;

[0076] Figure 5 The switching signal and tracking error changes of subsystem 1 under group 4 in the implementation case of this disclosure under low-constraint switching;

[0077] Figure 6 The output and ideal trajectory of subsystem 1 in the implementation case of this disclosure;

[0078] Figure 7 The output and ideal trajectory of subsystem 2 in the implementation case of this disclosure;

[0079] Figure 8 State variables of subsystem 1 in the implementation case of this disclosure and its constraint boundary curves.

[0080] Figure 9 The adaptive learning curve of the sensor fault parameters in the implementation case of this disclosure is a comparison curve with the true value.

[0081] Figure 10 The output signal variation curves of the controllers of subsystem 1 and subsystem 2 in the implementation examples of this disclosure are shown.

[0082] Figure 11 The curves showing the relationship between the tracking error and the predetermined performance envelope for subsystems 1 and 2 in the implementation examples of this disclosure are shown. Detailed Implementation

[0083] The present disclosure will now be further explained with reference to the accompanying drawings.

[0084] This disclosure addresses the pain points in the operation and control of existing complex industrial interconnected systems such as multi-joint robotic arms and interconnected power systems, namely: traditional switching strategies are too conservative, resulting in slow equipment production cycle; sensor aging and failure can easily lead to safety accidents; and high-order control algorithms have too much computational load to run on general-purpose chips. It proposes an artificial intelligence-based sensor fault-tolerant control method for industrial interconnected systems.

[0085] The core of this method lies in: constructing a low-constraint switching architecture with direction-aware capabilities, thus eliminating unnecessary downtime constraints in traditional methods; utilizing a state observer and adaptive compensation mechanism to reconstruct the real signal through software algorithms under sensor failure conditions; achieving predetermined performance control through an error preprocessing mechanism; approximating the system's unknowns using artificial intelligence technology based on fuzzy logic systems; and introducing dynamic surface signal processing technology, significantly reducing the controller's computing power requirements. This enables efficient, safe, and low-cost operation of industrial interconnected systems under complex operating conditions.

[0086] To achieve the above objectives, the present disclosure provides the following embodiments. Specific Implementation Example 1:

[0088] This disclosure provides an embodiment:

[0089] like Figure 1 A fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence, specifically including the following steps:

[0090] S1. Dynamic modeling and fault mode definition for complex industrial interconnected systems with operating condition switching.

[0091] A1. Mathematical model of complex industrial interconnection system.

[0092] For a complex industrial interconnected system consisting of M coupled subsystems, a rigorous feedback nonlinear mathematical state-space model of the i-th subsystem is established:

[0093] (1)

[0094] in, Number the subsystems. ; For system state variables; For switching signals; This refers to the nonlinear physical characteristics such as friction and damping that exist within the system. This refers to the strong interconnection and coupling between subsystems. This is an external disturbance. ; Indicates time; Represents the system's output vector; For the first The order of each subsystem.

[0095] A2. Sensor Fault Model.

[0096] Considering signal attenuation caused by sensor aging in industrial settings, a sensor fault model is established:

[0097] (2)

[0098] in, For sensor measurement output, For unknown failure factors, define failure estimation parameters. The actual output is represented as .

[0099] A3. Necessary Lemma and Assumptions.

[0100] To ensure the rigor of subsequent controller design and the stability of the closed-loop system, this disclosure is based on the following reasonable mathematical assumptions and lemmas:

[0101] Assumption 1: The Reference trajectory signal of each subsystem and until The derivative of the order is continuous and bounded.

[0102] Assumption 2: Unknown nonlinear interconnection terms between subsystems The following growth conditions must be met:

[0103] ; (3)

[0104] in, It is an unknown smooth nonlinear function.

[0105] Assumption 3: The estimation error caused by the fault and its rate of change are bounded. That is, there exist unknown normal numbers. and ,satisfy: and .

[0106] Lemma 1: For any set defined on a compact set arbitrary continuous nonlinear function on and any given approximation accuracy There always exists a fuzzy logic system , so that: , X∈Ω; where, For the ideal weight vector, is a fuzzy basis function vector.

[0107] S2. Construct a full-state safety constraint and error preprocessing mechanism.

[0108] To ensure that the system output tracking error strictly meets the preset convergence time and steady-state accuracy, and to avoid singularity problems in controller design, this embodiment constructs a novel non-singular error scalar transformation function, introduces dynamic surface signal processing technology and full-state safety boundary mechanism, and solves the problem that high-order control algorithms have large computational load and are difficult to run on general-purpose chips.

[0109] A1. Construct a non-singular error scalar transformation function Design the following piecewise smooth function as a performance scaling factor:

[0110] (4)

[0111] in: The predetermined convergence time set by the user, i.e. the time required for the system error to enter the steady-state region; and , These are the design parameters. The steady-state error limit is... ; It is a polynomial buffer term used to ensure a smooth transition of the function within its runtime. Its specific expression is:

[0112] ,

[0113] in, It is a constant, and the ingenious design of this function lies in that it is achieved through... The adjustment makes It is a finite constant, rather than an infinite or uncontrollable value as in traditional methods. This means that the controller will not produce singularities at startup, regardless of the initial state of the system.

[0114] A2. Based on the above transformation function, using sensor fault estimation parameters Define the first-level transformation error surface: By transforming the constrained original tracking error into an unconstrained variable... As long as it is guaranteed If the boundary is defined, it can be deduced that the original error always remains within the predetermined performance envelope. Within the neighborhood. Based on this, a tangent-type barrier Lyapunov function is constructed to simultaneously ensure that all states are constrained within the neighborhood. Inside: ,in This is the first-level state constraint bound. The property of this function is: when... Approaching When the controller output approaches infinity, it forces the controller to generate a huge reverse force, "pushing" the state back into the safe zone.

[0115] S3. Online approximation and compensation of unknown dynamic characteristics

[0116] Due to the internal dynamic function of the nonlinear switching interconnection system Since it is unknown, this embodiment uses artificial intelligence technology based on fuzzy logic system to approximate the unknown.

[0117] A1. Utilizing the universal approximation property of fuzzy logic systems, online approximation of unknown nonlinear functions is performed. The fuzzy rules take the following form:

[0118] rule :if yes And... and yes ,So yes ;in For fuzzy rules;

[0119] A2. Using single-point fuzzification, a product inference engine, and a centrally averaged defuzzifier, and applying Lemma 1, the unknown nonlinear function is expressed as:

[0120] , (5)

[0121] in: , The ideal fuzzy weight vector; For fuzzy basis function vectors, Gaussian function form is usually chosen; For the fuzzy approximation error, we assume it is bounded, i.e. This step transforms the complex "unknown nonlinear function" in the system into the "product of known basis functions and unknown weight vectors", thus transforming the control problem into an adaptive estimation problem of the parameters of the weight vectors.

[0122] S4. Design of State Observer and Fault Healing Module

[0123] To address the issues of unmeasurable system state and sensor gain failure, this step designs an observer and a fault estimation law to achieve synchronous reconstruction of state and fault.

[0124] A1. Based on the fuzzy approximation results from step S3, construct the following observer to estimate the system state. :

[0125] (6)

[0126] in: The derivative of the state estimate; and The observer gain matrix must be guaranteed during the design. It is a Herwitz matrix (i.e., all real parts of its eigenvalues ​​are negative), and To ensure convergence of observation errors; the observer's correction term adopts Instead of using directly .because Sensor fault parameters The estimated value of , Able to approximate real output This eliminates the impact of sensor failure on state estimation.

[0127] A2. In order to obtain fault parameters online The following adaptive update law is designed:

[0128] (7)

[0129] in, This is an adaptive gain; the adaptive law guarantees the estimated value. It always remains within a reasonable bounded range. ; It is a positive constant; It is the actual measurement value of the sensor; It is a normal number.

[0130] S5. Design a robust fault-tolerant controller with low computational power consumption and an adaptive parameter law.

[0131] Based on backstepping and dynamic surface control techniques, and combined with the Lyapunov function of tangential obstacle, a controller is recursively designed.

[0132] A1. Step 1 (Virtual Control Law Design): Define the first-level conversion error surface Constructing a barrier Lyapunov function

[0133] (8)

[0134] Using backstepping techniques and Young's inequality, design virtual control laws and adaptive laws:

[0135] (9)

[0136] , (10)

[0137] in, , and These are design parameters, and the control law ensures that the first-level state does not violate constraints.

[0138] A2, to avoid virtual control laws By performing analytical differentiation, a first-order low-pass filter is introduced:

[0139] (10)

[0140] in, This is the filtered virtual control signal. The error surfaces for subsequent stages are defined. Designing virtual controllers and adaptive laws at various stages using filters and backstepping techniques:

[0141] (11)

[0142] , (12)

[0143] in, , and These are design parameters.

[0144] A3. Final step (actual control law design): Based on the first-order error surface Using the barrier Lyapunov function, derive the actual control input and adaptive law:

[0145] , (13)

[0146] (14)

[0147] in, , and These are design parameters, and the control law simultaneously compensates for the effects of fuzzy approximation error, dynamic surface filtering error, and interconnection terms.

[0148] S6. Deployment of Low-Constraint Operating Condition Switching Strategy and Stability Analysis

[0149] In this embodiment, to address the problem that existing switching strategies (such as average dwell time, modality-dependent average dwell time, and allowable edge-dependent average dwell time) are too conservative and force devices to be shut down for extended periods, this embodiment constructs a low-constraint switching architecture based on the working condition switching direction with binary dependency average dwell time.

[0150] A logic judgment module based on the direction of operating condition switching is embedded in the controller. Based on the judgment result, the dwell time parameter is dynamically matched. For benign switching, the long waiting constraint is automatically released, allowing the system to respond quickly and adopting a low-constraint switching strategy, namely a binary dependency average dwell time strategy.

[0151] Assume there exists a set of switching subsystems. Define a binary mapping. , and ,set up For subsystem families The system switching signal must meet the following dwell time conditions:

[0152] ; (15)

[0153] in, The average residence time for binary dependencies; This is the initial time; T is the termination time; Represents the time interval Total runtime; It is a positive constant; Represents the time interval Total number of switches.

[0154] Assume there exists an element First, define an operator to collect the first element of the ordered pairs of switches in the group containing a specific switching subsystem:

[0155] .

[0156] To ensure the absolute safety of the above strategies in engineering applications, for the first... A subsystem, defining a quadratic Lyapunov function. Based on the conclusion obtained in step S5, its derivative satisfies the following dissipation inequality:

[0157] (16)

[0158] Wherein: Attenuation rate Defined as:

[0159]

[0160] constant term Defined as .

[0161] For any switching pair That is, the system changes from mode Switch to mode The Lyapunov function satisfies the following jump condition:

[0162] (17)

[0163] in, , , for The largest eigenvalue; yes The smallest eigenvalue.

[0164] This strategy configures different residence times based on the stability differences between different mode transitions, and covers and outperforms traditional average residence time, mode-dependent average residence time, and edge-dependent average residence time strategies.

[0165] Assuming within the time interval The time when the switch occurs is ,and Along the solutions of the system, construct the following piecewise differentiable function:

[0166] (18)

[0167] In any continuous interval Inside, to Differentiation yields:

[0168] (19)

[0169] For the above formula Integrating, we get:

[0170] (20)

[0171] Consider switching time The jump relationship can be obtained recursively as follows:

[0172] , (twenty one)

[0173] in,

[0174] and .

[0175] After multiple iterations, until the final moment And substitute the BDADT switching conditions. Introducing intermediate variables and get:

[0176] (twenty two)

[0177] Based on the above analysis and As defined, all signals within the system are semi-globally consistent and eventually bounded. (Tracking error) At the scheduled time Then converged to the predetermined accuracy Within a compact neighborhood. Based on the properties of the tangent-type barrier Lyapunov function, through parameter configuration design, the full-state... Always stay within the constraints Inside.

[0178] In summary, the AI-based fault-tolerant control method for industrial interconnection systems proposed in this embodiment provides a complete solution to the two major pain points in traditional industrial control: "rigid switching strategies leading to low efficiency" and "sensor failures causing system paralysis".

[0179] First, an artificial intelligence approximation algorithm is introduced to actively learn the unknown nonlinear characteristics inside the system. Combined with the sensor's online self-healing mechanism and non-singular error transformation, the problem of the failure of traditional control algorithms for predetermined performance control when the system loses sensor signals or has excessive initial errors is solved.

[0180] Secondly, by utilizing the low-constraint switching architecture proposed in this disclosure, the system can intelligently identify the switching direction between operating modes and match the corresponding dwell time, thereby flexibly configuring the equipment waiting time and breaking the conservative limitations of traditional methods.

[0181] Furthermore, by combining lightweight dynamic surface control with a full-state hard constraint mechanism, this disclosure not only significantly reduces the algorithm's computational requirements on the chip, but also ensures that the system strictly meets physical safety constraints at all times through virtual safety boundaries. Simulation verification shows that this method possesses extremely strong robustness and high-precision tracking capabilities under complex fault environments, providing a reliable technical path for fault-tolerant control of equipment such as multi-joint robotic arms and interconnected power grids. Specific Implementation Example 2:

[0183] This disclosure also provides an embodiment:

[0184] As an example, this embodiment constructs a second-order complex interconnected system model containing two subsystems. This model simulates the core coupling characteristics of a dual-joint robotic arm or a dual-region interconnected microgrid in terms of dynamic properties. The mathematical dynamic equations of this model are as follows:

[0185] .

[0186] in, The control objective is to make the system output Track the reference signal separately and The tracking error is required to be within a predetermined time. Converging to predetermined accuracy Within its neighborhood.

[0187] Configure the system to handle a gain failure fault during operation: and .

[0188] The controller parameters are initialized as follows:

[0189]

[0190] The fuzzy membership function is a Gaussian function. .

[0191] The system design parameters are selected as follows:

[0192]

[0193] The low-constraint switching mechanism of this disclosure, namely the binary dependency average residence time strategy, is adopted, and different switching mapping groups, such as groups 1-4, are set to verify the flexibility of the strategy. Therefore, the calculated average residence time can be referred to Table 1. It can be seen that after configuring the parameters of the binary dependency average residence time (BDADT), the average residence time (ADT), modal dependency average residence time (MDADT), and allowable edge-dependent average residence time (AED-ADT) can actually be regarded as a special case of BDADT. That is, this low-constraint switching mechanism has strong scalability and practicality.

[0194] Depend on Figures 2-5As can be seen from the binary average dwell time calculated in Table 1, simulation results show that, under different BDADT switching groups, the system employing the strategy disclosed in this paper can complete multiple types of operating mode switching within the same time window, and the state curves are smooth and oscillating. This verifies that the proposed method successfully eliminates unnecessary downtime constraints and can significantly improve the equipment's action response speed and production cycle time in practical engineering.

[0195] Table 1: Average Resident Time of Binary Dependencies under Different Configuration Parameters

[0196]

[0197] according to Figure 6 and Figure 7 It is evident that even in the event of a severe sensor malfunction, the controller disclosed herein can still enable the system to output an accurate tracking reference trajectory. For example... Figure 8 As shown, system status Always maintain the preset constraint boundary During this period, the system state did not physically exceed limits, thus preventing mechanical collisions or overtravel accidents from occurring at the algorithm level. For example... Figure 9 As shown, the system's sensor fault parameters will quickly converge to their true values ​​after the system starts running, successfully repairing the distorted sensor signal at the software level and allowing for sensor gain failure faults. Figure 10 As shown, the controller outputs of subsystem 1 and subsystem 2 enable the system to operate normally without controller singularities. Figure 11 As shown, the tracking errors of both subsystem 1 and subsystem 2 can remain within the predetermined performance envelope within a predetermined time, thus achieving the preset convergence time and convergence accuracy.

[0198] In summary, this embodiment verifies that the AI-based sensor fault-tolerant control method for industrial interconnection systems described in Specific Embodiment 1 has excellent control performance and robustness when facing sensor faults, unpredictable states, and complex switching environments. Specific Implementation Example 3:

[0200] This disclosure also provides an embodiment:

[0201] A fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence includes: a storage medium and a processing unit; wherein, the storage medium is used to store a computer program; the processing unit exchanges data with the storage medium and executes the computer program through the processing unit to perform the steps of the fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence as described in Specific Embodiment 1. Specific Implementation Example 4:

[0203] This disclosure also provides an embodiment:

[0204] A computer-readable storage medium: the computer-readable storage medium stores a computer program;

[0205] When the computer program is running, it executes the steps of the sensor fault-tolerant control method for industrial interconnection systems based on artificial intelligence as described in Specific Embodiment 1.

[0206] It should be clarified that, in this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wireline, optical fiber, RF, etc., or any suitable combination thereof.

[0207] The above disclosures only cover a few specific implementation scenarios. However, this disclosure is not limited to these, and any variations that can be conceived by those skilled in the art should fall within the protection scope of this disclosure. The serial numbers in this disclosure are for descriptive purposes only and do not represent the superiority or inferiority of the implementation scenarios.

Claims

1. A fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence, characterized in that, include: Transform any industrial interconnected system in the physical world into a digital state equation that the controller can recognize, and establish a sensor fault model based on signal attenuation caused by sensor aging. Construct a non-singular error scalar transformation function to convert the constrained tracking error into an unconstrained variable; A barrier Lyapunov function is constructed to ensure that the transformation error state always remains within the constraint limits; Artificial intelligence technology based on fuzzy logic system is used to approximate unknown nonlinear continuous functions, transforming complex unknown nonlinear functions into the product of known basis functions and unknown weight vectors; Based on the unmeasurability of system state and the characteristics of sensor gain failure, a state observer and a fault self-healing module are constructed. Based on backstepping and dynamic surface control techniques, combined with the Lyapunov function of tangent obstacle, a controller is recursively constructed. A logic judgment module based on the direction of operating condition switching is embedded in the controller to obtain the judgment result; Based on the judgment results, it can complete the switching of various types of working condition modes and improve fault tolerance.

2. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 1, characterized in that, The process of transforming any industrial interconnected system in the physical world into a digital state equation recognizable by the controller, and establishing a sensor fault model based on signal attenuation caused by sensor aging, includes: For an industrial interconnected system consisting of M coupled subsystems, and considering the possibility of system switching conditions, a feedback nonlinear mathematical model for the i-th subsystem is established: ； in, This refers to the system status; For switching signals; This refers to the nonlinear physical characteristics existing within the system. To characterize the strong interconnection and coupling that exists between subsystems. It is a distractor. To control the input, ; Indicates time; Represents the system's output vector; For the first The order of the subsystems; Based on the signal attenuation caused by sensor aging in industrial settings, a sensor fault model is established: ； in, This is the actual measurement value from the sensor. The unknown sensor failure factor; Introducing fault parameters , will output the actual data Represented as: ； in, For sensor measurement output, and These are the approximate errors and approximate values ​​of the fault parameters, respectively. Indicates time; This refers to the system sensor failure time.

3. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 1, characterized in that, The construction of the non-singular error scalar transformation function, which transforms the constrained tracking error into an unconstrained variable, includes: Constructing a polynomial buffer term Non-singular error scalar transformation function : ； in, and , and For design parameters; ; Where t represents time. It is a positive design constant; The scheduled convergence time set by the user.

4. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 1, characterized in that, The construction of the barrier Lyapunov function ensures that the transformation error state always remains within the constraint limits, including: Constructing tangent-type barrier Lyapunov functions to achieve full-state constraints Inside: ； in, This serves as the state error constraint boundary; These are unconstrained variables; It is a user-defined state constraint bound.

5. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 1, characterized in that, The method of approximating unknown nonlinear continuous functions using artificial intelligence technology based on fuzzy logic systems transforms complex unknown nonlinear functions into the product of known basis functions and unknown weight vectors, including: Using the universal approximation property of fuzzy logic systems, fuzzy rules are constructed: rule :if yes And... and yes ,So yes ;in For fuzzy rules; The unknown nonlinear function is represented as: ； in, , The ideal fuzzy weight vector; For fuzzy basis function vectors; This represents the fuzzy approximation error.

6. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 1, characterized in that, Based on the unmeasurability of system state and the characteristics of sensor gain failure, a state observer is constructed, including: Regarding the problem of unpredictable system state, when the first When a switching system is activated, the following observer is constructed to estimate the system state. : ； in, The derivative vector of the state estimate; It is a vector of state estimates; The observer gain matrix; ; ; It is the actual measurement value of the sensor; For the observer's correction term; These are estimated values ​​of sensor fault parameters; To estimate the error; ; This is an estimate of the ideal fuzzy weight vector; For fuzzy basis function vectors; To control the input; let and , making It is a strict Hurwitz matrix, and there exists a matrix... satisfy ,in It is a positive definite symmetric matrix.

7. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 1, characterized in that, The self-healing module includes: Get fault parameters online The following adaptive update law is designed: ； in, For adaptive gain; ; and This can be derived from claim 6; It is a positive constant; It is the actual measurement value of the sensor; It is a normal number.

8. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 1, characterized in that, The controller, based on backstepping and dynamic surface control techniques and combined with the tangent obstacle Lyapunov function, is recursively constructed, including: Define the first-level transformation error surface We construct the barrier Lyapunov function and design virtual control and adaptive laws using backstepping techniques and Young's inequality. Introduce a first-order low-pass filter: ;in, A positive constant. This is the filtered virtual control signal; For virtual controllers, the errors at each subsequent stage are: ; Based on the Level error surface and barrier Lyapunov function, deriving practical control input and adaptive law: ； in, , and These are design parameters, and the control law simultaneously compensates for the effects of fuzzy approximation error, dynamic surface filtering error, and interconnection terms.

9. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 1, characterized in that, The logic judgment module based on the direction of working condition switching is implanted in the controller to obtain the judgment result; Based on the judgment results, various types of working condition mode switching are completed, including: Assume there exists a set of switching subsystems. Define a binary mapping. , and ,set up For subsystem families The system switching signal must meet the following dwell time conditions: ； in, The average residence time for binary dependencies; This is the initial time; T is the termination time; Represents the time interval Total runtime; It is a positive constant; Represents the time interval Total number of switching times; Assume there exists an element First, define an operator to collect the first element of the ordered pairs of switches in the group containing a specific switching subsystem: ； For the A subsystem, defining a quadratic Lyapunov function. And its derivative satisfies the following dissipation inequality: ;in, The attenuation rate; For constant terms; For any switching pair The Lyapunov function satisfies the following jump condition: ;in, ; According to claim 6, It can be calculated; for The largest eigenvalue; yes The smallest eigenvalue.

10. The fault-tolerant control method for sensors in an industrial interconnected system based on artificial intelligence according to claim 9, characterized in that: Attenuation rate Defined as: ； constant term Defined as ; in, for eigenvalues; It can be calculated according to claim 6; It is a matrix The largest eigenvalue; It is a positive constant; All of these are design parameters.

Images