Intelligent power plant inspection robot sensor fault tolerance control system and method

Abstract

The invention relates to the technical field of mobile robot control, in particular to an intelligent power plant inspection robot sensor fault tolerance control system, which is characterized in that a motion model construction module constructs a motion model; the disturbance observer design module defines an observation error and designs a disturbance observer, and the disturbance observer performs feedback adjustment by combining the observation error on the basis of the motion model; the augmentation modeling and self-adaptive sensor fault observation module is used for constructing an augmentation system model; designing an adaptive sensor fault observer based on the augmented system model; by means of an adaptive law, continuously optimizing the compensation amount for the interference observation error, and automatically counteracting the error generated in the interference estimation process; the fault-tolerant controller design module defines a tracking error between an actual track and an expected inspection track of the robot, designs a generalized sliding mode surface and a control law, and accelerates convergence of the sliding mode surface and reduces buffeting in combination with an exponential reaching law and a variable-speed reaching law.

Description

Technical Field

[0001] This invention relates to the field of mobile robot control technology, specifically a sensor fault-tolerant control system and method for a wheeled mobile inspection robot in a smart power plant, which is suitable for high-precision motion control in scenarios with external interference and sensor failure. Background Technology

[0002] In smart power plant scenarios, wheeled mobile inspection robots need to perform inspection tasks for extended periods in complex environments such as high temperatures, high electromagnetic interference, and uneven ground. Their core sensors (such as encoders for positioning and gyroscopes for attitude sensing) are susceptible to two types of malfunctions due to environmental factors: There are two main types of failures: one is complete sensor failure, where the sensor is completely inoperable and unable to output data; the other is loss failure, which manifests as incomplete failure issues such as data deviation and drift. These issues directly affect the robot's judgment and control accuracy regarding its own motion state. Traditional sensor fault-tolerant control methods are difficult to adapt to the actual needs of this scenario and have obvious limitations. Data-driven methods rely on a large amount of historical data to train fault diagnosis models, but real-time interference in the power plant environment (such as electromagnetic radiation generated by equipment operation and sudden changes in ground friction) is frequent and unpredictable. These methods do not fully consider the masking or false triggering effect of real-time interference on fault characteristics, resulting in a high misdiagnosis rate of sensor faults. While analytical model-based methods can construct residual signals to detect faults based on precise kinematic / dynamic mathematical models and measurable data of robots, they either fail to isolate the coupling relationship between external interference and residuals, or assume that the interference intensity is extremely small and negligible. However, in actual power plant scenarios with strong interference, interference can severely distort residual signals, significantly reducing the accuracy of fault detection and making it impossible to effectively distinguish whether the system deviation is caused by interference or sensor failure. At the same time, existing technologies still have three major problems: First, interference and sensor faults are not decoupled. When estimating faults, the fault observer may misjudge the interference signal as a fault component or have the true fault masked by the interference, resulting in low fault observation accuracy and inability to accurately quantify the degree of sensor faults (such as the magnitude of drift). Secondly, the controller is not robust enough to interference estimation errors. When the interference observer has estimation bias due to factors such as dynamic response delay, the controller cannot effectively compensate for the bias, which can easily lead to the control output deviating from the actual requirements and affect the stability of robot motion. Third, traditional sliding mode fault-tolerant control uses discontinuous control laws to ensure rapid system response, resulting in significant "chattering" and large tracking errors, with position tracking errors as high as 0.3586 and steering tracking errors as high as 0.1593. These issues make it difficult to meet the stringent requirements of smart power plant inspection for robot trajectory accuracy (such as precise movement along equipment pipelines and obstacle avoidance). These problems together mean that existing fault-tolerant control solutions cannot provide stable and high-precision motion guarantees for smart power plant inspection robots. In view of this, we propose a fault-tolerant control system and method for sensors in a smart power plant inspection robot. Summary of the Invention

[0003] The purpose of this invention is to solve the problems that when a wheeled mobile inspection robot for a smart power plant is running in a complex environment, the sensor is prone to failure (complete incompetence) or loss (data deviation, drift). Furthermore, traditional fault-tolerant control methods suffer from problems such as low fault observation accuracy due to the lack of decoupling between interference and sensor failure, insufficient robustness of the controller to interference estimation errors, significant chattering and high tracking errors in traditional sliding mode fault-tolerant control.

[0004] To achieve the above objectives, this invention provides a fault-tolerant control system for a smart power plant inspection robot's sensor faults, comprising a motion model construction module, an interference observer design module, an augmented modeling and adaptive sensor fault observation module, and a fault-tolerant controller design module, wherein: The motion model construction module is based on the law of inertia and the motion characteristics of wheeled mobile robots, and focuses on building a motion model that can comprehensively depict the key information of the system. In the smart power plant scenario, the inspection robot needs to cope with complex environments such as high temperature, electromagnetic radiation, and uneven ground. Its motion state is easily affected by multiple factors. Therefore, the motion model construction module first clarifies the core motion state variables of the robot, including the position coordinates in the two-dimensional plane, the heading angle describing the direction of motion, the linear velocity along the heading, and the angular velocity controlling the steering, forming a basic state vector that accurately reflects the robot's real-time motion posture. Subsequently, the motion model construction module, combined with the differential steering principle of the wheeled robot, derives the kinematic relationships: the rate of position change is obtained by decomposing the linear velocity along the heading angle (i.e., the rate of change of directional position is related to the linear velocity and the cosine of the heading angle, and the rate of change of directional position is related to the linear velocity and the sine of the heading angle), the rate of change of the heading angle is directly equal to the angular velocity, and the changes in linear velocity and angular velocity are respectively related to the interaction of driving force, steering torque and viscous friction coefficient, ensuring that the model can accurately describe the motion law of the robot when there is no interference and no faults; Meanwhile, considering the unavoidable external interferences in the power plant environment (such as fluctuations in ground friction and electromagnetic interference generated by equipment operation) and sensor faults (such as encoder data drift and gyroscope measurement deviations), the module quantifies the interference into interference vectors (including motion direction interference and steering direction interference, which affect the stability of linear velocity and angular velocity, respectively), and abstracts sensor faults into fault vectors (position sensor faults and steering sensor faults, which cause deviations in position measurement values ​​and heading angle measurement values, respectively). Through the fusion of state equations and output equations, the constructed motion model simultaneously reflects the correlation between the robot's motion state, external interferences, and sensor faults—the state equations reflect the impact of interferences on the motion state, and the output equations reflect the effect of faults on the measurement values. This provides a precise mathematical framework for the interference estimation of the interference observer design module and the fault diagnosis of the augmented modeling and adaptive sensor fault observation module.

[0005] The disturbance observer design module, based on the motion model output by the motion model construction module, focuses on addressing the impact of external disturbances on robot control accuracy in a power plant environment. It achieves real-time disturbance estimation through a process of defining observation error, designing the observer, and feedback adjustment. First, the interference observer design module defines the observation error based on measurable state variables such as velocity and angular velocity in the motion model. This error is the difference between the actual measured value of the state variable and the preliminary estimated value of the observer. This error directly quantifies the system deviation caused by external interference. For example, when the robot experiences a sudden change in friction in its direction of motion due to uneven ground (enhanced external interference), the deviation between the actual measured value and the estimated value of the linear velocity will increase accordingly, providing a clear quantitative basis for interference estimation. Next, the core observation system of the disturbance observer is designed: based on the state evolution law of the motion model, the observation error is introduced into the observer as a feedback signal. By setting the observation gain (which must meet the system stability condition to ensure that the estimated value output by the observer can quickly converge to the true value), the observer can dynamically adjust the estimation of the disturbance according to the error. For example, when the observation error increases, the observer increases the correction amplitude of the disturbance estimate through feedback adjustment, so that the estimated value gradually approaches the true disturbance size; when the observation error decreases, the correction amplitude is reduced accordingly to avoid large fluctuations in the estimated value. Finally, the interference estimate output by the interference observer design module will be fed back to the augmented modeling and adaptive sensor fault observation module and the fault-tolerant controller design module in real time, providing key data support for subsequent offsetting of interference effects and improving the system's anti-interference capability. It also provides feedback for the model optimization of the motion model construction module. If the observed interference pattern deviates from the interference characteristics preset by the model, the interference quantization parameters in the model can be adjusted in reverse to further improve the model accuracy.

[0006] The augmented modeling and adaptive sensor fault observation module is the core of the fault diagnosis system. It is the key to achieving accurate identification and isolation of sensor faults. Through the three-step process of augmented modeling, adaptive observation and fault isolation, it solves the problems of interference and fault coupling and low observation accuracy in traditional methods. Moreover, it relies on the motion model of the motion model construction module and the interference estimate of the interference observer design module throughout the process to ensure the accuracy of fault diagnosis.

[0007] Construction of the augmented system model in the augmented modeling and adaptive sensor fault observation module: Considering that sensor faults (such as drift and deviation) are difficult to directly separate from the motion state, the augmented modeling and adaptive sensor fault observation module augments sensor faults as independent variables, integrating the robot's original motion state (position, heading angle, velocity, etc.) with sensor faults (position sensor faults, steering sensor faults) to form an augmented state vector. Based on this, the augmented system model is reconstructed using the motion model construction module as the foundation—the evolution law of faults (such as the trend of drift faults over time) is incorporated into the system state equation, making faults observable and estimable state variables. At the same time, the interference estimate output by the interference observer design module is introduced to preemptively deduct the influence of interference on the system state, avoiding residual interference that leads to fault estimation bias, and laying an accurate model foundation for subsequent fault observation. Adaptive sensor fault observer design: Based on the augmented system model, the augmented modeling and adaptive sensor fault observation module designs an adaptive sensor fault observer. Its core advantage lies in adaptive compensation for interference errors: the observer adjusts the output error (the difference between the actual output and the observed output) of the augmented system through the observation gain matrix, and quickly adjusts the estimation direction of the system state and fault. The observation gain must meet the matrix stability condition to ensure the convergence and estimation accuracy of the observer. For example, when the observed output error increases, the observation gain will guide the estimation direction to adjust in the direction of reducing the error, so that the estimated values ​​of the state and fault are closer to the true values. Meanwhile, by dynamically updating the compensation parameters with an adaptive law, the compensation amount for interference observation errors is continuously optimized: if there is a slight deviation in the interference estimate of the interference observer design module (due to dynamic changes in interference), the adaptive law will automatically adjust the compensation coefficient according to the output error of the augmented system model to offset the error generated in the interference estimation process and avoid interference contaminating the fault estimation results; for example, when the interference estimate is slightly smaller than the actual interference, the adaptive law will increase the compensation amount, which is equivalent to deducting the residual interference effect in the fault estimation, thereby significantly improving the accuracy of fault estimation, especially suitable for sensor fault diagnosis in power plant strong interference scenarios.

[0008] Fault Isolation: To prevent fault propagation from affecting the overall system operation, the augmented modeling and adaptive sensor fault observation module sets a minimum fault threshold (this threshold is determined through multiple power plant scenario experiments to ensure accurate differentiation between "minor faults that do not affect tracking performance" and "serious faults requiring intervention"). When the adaptive sensor fault observer acquires the degree of sensor fault, it compares the fault value with the minimum threshold: if the fault value is greater than the threshold, it indicates that the fault has seriously affected the robot's inspection trajectory tracking performance (e.g., a position sensor fault causing the robot to deviate from the expected trajectory beyond the safe range), and the decision-making body will immediately trigger the fault isolation mechanism to cut off the signal transmission of the faulty sensor and prevent erroneous data from entering the subsequent controller; if the fault value is less than the threshold, it is determined to be a minor fault, and the augmented modeling and adaptive sensor fault observation module can offset its impact through the control and adjustment of the fault-tolerant controller design module without interrupting the inspection task, thus achieving accurate judgment and on-demand isolation fault handling logic.

[0009] The fault-tolerant controller design module aims to solve the problems of significant chattering and large tracking errors in traditional sliding mode control. Through the design concept of "error definition - sliding surface design - control law optimization - approach law innovation", it achieves high-precision trajectory tracking under strong interference and sensor failure. Moreover, it relies on the output data of the preceding modules throughout the process to ensure the pertinence and effectiveness of the control strategy.

[0010] First, the fault-tolerant controller design module defines the tracking error between the robot's actual inspection trajectory and the desired trajectory—including position error (the difference between the actual position and the desired position), heading angle error (the difference between the actual heading angle and the desired heading angle), and speed error (the difference between the actual speed and the desired speed). These errors directly reflect the degree of deviation of the robot's trajectory, providing a clear target for control adjustment. Subsequently, a generalized sliding surface was designed: unlike traditional sliding surfaces that rely on only a single error, this sliding surface integrates the proportional, derivative, and integral terms of the tracking error. The proportional term ensures a rapid response to the current error, the derivative term suppresses the sudden trend of error (such as a brief error jump caused by fault isolation), and the integral term eliminates steady-state error (such as a fixed deviation caused by the accumulation of small disturbances during long-term operation). Through the synergy of these three terms, the sliding surface can comprehensively characterize the deviation state of the system, providing a precise adjustment basis for the design of the control law.

[0011] Based on the generalized sliding surface, a fault-tolerant control law is further designed: the control law incorporates the disturbance estimate and the disturbance observation error estimate output by the disturbance observer design module, and the disturbance effect is offset by feedforward compensation—for example, the driving force or steering torque is adjusted in advance according to the disturbance estimate to offset the influence of the disturbance on speed and angular velocity; at the same time, a robust term is introduced to enhance the controller's anti-interference ability against incompletely estimated disturbances and minor faults, and to avoid large fluctuations in the control output.

[0012] The beneficial effects of this invention are: Improved fault diagnosis speed and accuracy: Convergence time for constant / time-varying fault estimation is less than 0.5s, and the error is less than 0.01. (e.g.) Figure 4-5 ); Interference suppression capability: The interference estimation error converges to zero (e.g., Figure 6 ); Improved fault tolerance: Compared to SMC and FTC, the fault diagnosis system proposed in this invention provides faster and more accurate diagnostic results for both constant and time-varying sensor faults in robots. It also exhibits excellent anti-interference capabilities against disturbances encountered during robot operation, resulting in superior tracking performance. Furthermore, to better illustrate the tracking effect, the root mean square error (RMSE) is used to quantitatively compare the differences between FTC, SMC, and the algorithm of this invention. The comparison results are as follows: Figures 2-3 The results show that the error of the state variables in the algorithm of this invention is reduced by 48.8% to 95%. Figure 2-3 ).

[0013] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description

[0014] Figure 1 Schematic diagram of a wheeled mobile inspection robot; Figure 2 : Schematic diagram of the implementation steps of this invention; Figure 3 AFTC system structure diagram (including fault diagnosis and controller); Figure 4 Comparison of constant fault estimation curves; Figure 5 Comparison of time-varying fault estimation curves; Figure 6 Comparison of interference estimation curves; Figure 7 Module diagram.

[0015] The meanings of the labels in the diagram are as follows: 100. Motion model construction module; 200. Interference observer design module; 300. Augmented modeling and adaptive sensor fault observation module; 400. Fault-tolerant controller design module. Detailed Implementation

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] refer to Figures 1-7 As shown, the fault-tolerant control system for the sensor of the intelligent power plant inspection robot includes: The motion model construction module 100 constructs the motion model; the interference observer design module 200 defines the observation error and designs the interference observer, which is based on the motion model and adjusted by feedback in combination with the observation error; the augmented modeling and adaptive sensor fault observation module 300 uses sensor fault as an independent variable for augmentation to construct an augmented system model; an adaptive sensor fault observer is designed based on the augmented system model; with the help of the adaptive law, the compensation amount for the interference observation error is continuously optimized to automatically offset the error generated in the interference estimation process; and a minimum threshold is set, and a decision-making mechanism is constructed according to the minimum threshold to isolate the fault; the fault-tolerant controller design module 400 defines the tracking error between the robot's actual trajectory and the expected inspection trajectory, designs a generalized sliding surface and control law, and combines the exponential approach law and the variable speed approach law to accelerate the convergence of the sliding surface and reduce chattering.

[0018] Specifically: The motion model construction module 100 utilizes the law of inertia (the change of an object's motion state is related to external forces) and the motion characteristics of wheeled mobile robots (differential drive enables straight-line movement and steering) to construct a motion model that can simultaneously reflect the robot's state, external interference, and sensor malfunctions. The motion model precisely quantifies the dynamic changes in the robot's position, velocity, heading angle, angular velocity, and other motion states. It also incorporates external interferences from the power plant environment, such as electromagnetic interference (e.g., electromagnetic disturbances to the drive motor causing power output deviations), and fluctuations in ground friction (e.g., changes in motion resistance caused by uneven ground). Furthermore, it considers the failure modes (including complete failure and partial performance loss) of position sensors (e.g., encoder signal drift and data loss) and steering sensors (e.g., gyroscope accuracy decay and data offset). This provides a fundamental mathematical and physical framework for subsequent design of interference observers to isolate external interference, adaptive fault observers to diagnose sensor faults, and fault-tolerant controllers to achieve precise control under fault conditions. This ensures a clear connection between the robot's own motion, external interferences, and sensor faults at the model level, providing underlying support for the stable inspection of the entire system in the complex environment of a smart power plant. The specific motion model is as follows: ; ; in This is the state vector of the robot's motion state. Let the coordinates be the robot's position coordinates on the plane. This refers to the heading angle (attitude direction). The rotational angular velocity about an axis perpendicular to the plane of motion; To control the input vector, For the control input of the left and right wheels (such as voltage or driving force); , This relates the control inputs for the left and right wheels to average control and differential control. This is the average control input (for forward control). This is the input for differential control (controlling steering); This is the external interference vector. Interference in the direction of motion (such as electromagnetic interference). For steering direction interference (such as fluctuations in ground friction); For the fault vector, This is due to a position sensor malfunction (such as drift). The problem is a steering sensor malfunction (e.g., misalignment). The system matrix represents the inherent damping characteristics of the robot's motion. The coefficient of viscous friction for linear motion; The coefficient of viscous friction for steering motion. The moment of inertia about the center of gravity, For the moment of inertia of the wheel, For robot quality, For the wheel radius, The coefficient of viscous friction is... The distance from the left and right wheels to the center of gravity. For driving gain; The control input matrix reflects the gain characteristics of the control input on the motion, where... Gain is driven by linear motion. For steering motion drive gain, To drive the gain, For wheel speed related quantities; To output the matrix, select the states to be measured (position and heading angle).

[0019] The interference observer design module 200, based on the constructed motion model, first defines the velocity observation error. (Estimation of feedback adjustment speed and disturbance) and angular velocity observation error (Used for feedback adjustment of angular velocity and disturbance estimation); Velocity observation error The actual speed of the robot's movement Estimated velocity based on observation The difference, angular velocity observation error The actual rotational angular velocity of the robot Angular velocity estimated by observation The difference; Thus through velocity observation error and angular velocity observation error The system captures in real time the impact of external disturbances (such as electromagnetic interference in a power plant environment, fluctuations in ground friction, etc.) on the robot's motion state, and then dynamically adjusts the estimated value of external disturbances to decouple the disturbances from sensor faults, providing a clean input signal for subsequent sensor fault diagnosis and fault-tolerant control.

[0020] The interference observer design module 200 designs an interference observer. Based on a motion model, the interference observer incorporates feedback adjustments using the aforementioned observation errors (velocity observation error and angular velocity observation error). By monitoring changes in velocity and angular velocity observation errors in real time, it dynamically adjusts the estimated values ​​for external interference. The disturbance observer includes velocity observation and disturbance estimation equations, and angular velocity observation and disturbance estimation equations, specifically: The velocity observation and disturbance estimation equations are as follows:

[0021]

[0022] in: is the coefficient of viscous friction for linear motion in the motion model. For linear motion driving gain in the motion model, For inputs related to speed control, This is an estimate of the disturbance in the direction of motion. The observer gain (used to adjust the convergence characteristics of the error feedback); The equations for angular velocity observation and disturbance estimation are as follows: ; ; in: For the steering motion viscous friction coefficient in the motion model, This represents the steering motion drive gain in the motion model. For inputs related to angular velocity control, This is an estimate of the steering direction disturbance. For observer gain; More specifically: When external disturbances (such as additional friction caused by uneven ground) act on the robot, the robot's actual speed or angular velocity will deviate from the predicted value of the motion model, thus leading to speed observation errors. or angular velocity observation error The interference observer increases; while the interference observer dynamically adjusts the internal velocity observation and interference estimation equations and angular velocity observation and interference estimation equations by monitoring changes in observation errors in real time, thus increasing the interference estimate. and By gradually approaching the true interference value, the external interference is eventually separated from the total error, thus avoiding confusion with sensor failure. This enables the subsequent fault diagnosis system to determine whether the sensor is faulty based on clean error signals, avoiding the misjudgment of sensor faults by interference. At the same time, the error after removing interference is fed back to the controller, providing a basis for the controller to generate accurate control commands. This ensures that the robot can maintain a stable motion state even in the presence of external interference, thus guaranteeing the smooth progress of smart power plant inspection tasks.

[0023] The augmented modeling and adaptive sensor fault observation module 300 augments the sensor fault as an independent variable, integrating the robot's original motion state. Sensor malfunction For augmentation state An augmented system model was constructed, which includes both the dynamic characteristics of the robot's own motion and the influencing factors of sensor failure. Augmented state Medium robot status The robot's initial state (such as position, velocity, heading angle, angular velocity, etc.), sensor malfunction. The sensor fault vector (including position sensor faults and steering sensor faults) is used; by augmenting the faults as independent variables in the augmented system model, joint modeling of faults and system states is achieved. The state equations of the augmented system model are: ; in; To augment the system model matrix, The system matrix represents the original motion model of the robot, describing the evolution characteristics of the original state; To augment the control input matrix and reflect the coupling relationship between the original system characteristics and faults; This is the control input matrix for the robot's original motion model, describing the influence of the control input on the system. For robot control inputs, such as the drive voltage for the left and right wheels; The interference input matrix; The interference term, which includes the interference observation error, is used to describe the impact of interference on the augmented system model. To prevent observation errors before the interference observer converges; This is the adaptive term matrix; To augment the system model output equations The derivative is used to adaptively compensate for interference estimation errors and improve fault estimation accuracy; Output equations of augmented system model ,in To augment the system model's output vector, it includes sensor measurements of faulty output signals; To augment the output matrix, This is the output matrix of the robot's original motion model, used to select the states that need to be measured (such as position and heading angle).

[0024] An adaptive sensor fault observer is designed based on the augmented system model. The adaptive sensor fault observer uses the observation gain matrix to provide feedback adjustment to the output error of the augmented system model, thereby quickly adjusting the estimation direction of the system state and fault. Adaptive sensor fault observer: ; ; ; in: in, For robot control inputs (such as left and right wheel drive voltages); This is the adaptive term matrix (used to compensate for interference estimation errors and improve fault estimation accuracy). To augment the output vector of the system model's output equations; The interference input matrix (used to describe the impact of external interference (such as electromagnetic interference from the power plant environment, fluctuations in ground friction, etc.) on the augmented system model); The external disturbance value estimated by the disturbance observer; This is a compensation term for interference observation errors (used to correct the bias in interference estimation). For adaptive weights, update according to the following adaptive law. , It is a very small positive number (to avoid the denominator being zero (ensuring the mathematical rationality of the formula), and at the same time, it plays a smoothing role, suppressing the chattering problem of traditional sliding mode control).

[0025] The augmented modeling and adaptive sensor fault observation module 300 utilizes an adaptive law to continuously optimize the compensation amount for interference observation errors, automatically offsetting errors generated during interference estimation, thereby significantly improving the estimation accuracy of sensor faults. The adaptive law is as follows: ; in: The update rate of the adaptive weights describes the adaptive weights. Patterns of change over time; It is a positive definite matrix (pre-designed to ensure the convergence of the adaptive law), which is a key parameter to ensure stability in adaptive control; The norm of the output error serves as the driving signal for the adaptive law—the larger the error, the greater the adaptive weights. The faster the update rate, the more timely the correction of deviations.

[0026] Augmented modeling and adaptive sensor fault observation module 300 fault isolation: After the adaptive sensor fault observer obtains the degree of sensor fault, it constructs a decision-making mechanism to isolate the fault according to the minimum threshold determined through multiple experiments. First, compare the observed fault severity with the minimum threshold. If the fault severity is greater than the threshold, it means that the fault will seriously reduce the robot's tracking performance, and the fault signal of the corresponding sensor needs to be shielded. If the severity of the fault is less than the threshold, it means that the fault has a negligible impact on the system and can be left unaddressed for the time being. By isolating faults, valid sensor signals that are not affected by faults (such as position and angular velocity data from fault-free sensors) are ultimately preserved, preventing fault signals from entering subsequent fault-tolerant controllers. This prevents the controllers from generating deviation control commands based on erroneous signals, ensuring that subsequent control links can achieve stable motion control of the robot based on clean and valid signals, and guaranteeing the smooth progress of smart power plant inspection tasks.

[0027] The 400 fault-tolerant controller design module defines the tracking error between the robot's actual trajectory and the desired inspection trajectory. It designs a generalized sliding surface and control law, and combines exponential and variable speed approach laws to accelerate sliding surface convergence and reduce chattering. Even with sensor failures and external interference, it can achieve accurate trajectory tracking and inspection tasks for the robot in the power plant.

[0028] Define the tracking error as The following generalized sliding surface is designed: ; The control law is designed as follows: ; ; in To prevent interference with observation errors The estimated value The update law in the controller is Furthermore, to accelerate the approach of the sliding surface to zero while reducing sliding chattering, a novel approach law is designed: Using the exponential reaching law and the law of speed convergence In combination, when the state variable is far from the sliding surface, the exponential reaching law dominates, ensuring a sufficiently large reaching rate. However, as the state variable gradually approaches the sliding surface, the variable reaching law dominates, with the rate of convergence increasing as the state variable moves along the sliding surface. The state variable decreases as the motion progresses until it reaches a stable point.

[0029] A fault-tolerant control method for sensor faults in a smart power plant inspection robot includes the following steps: S1. Utilize the laws of inertia and the motion characteristics of wheeled mobile robots to construct a motion model that can simultaneously reflect the robot's state, external disturbances, and sensor malfunctions. S2: Based on the established motion model, define the observation error; design a disturbance observer, which is based on the motion model and uses the observation error for feedback adjustment. S2: Augment the sensor fault as an independent variable, integrate the robot's original motion state with the sensor fault as the augmented state, and construct an augmented system model. S4: An adaptive sensor fault observer is designed based on an augmented system model. The adaptive sensor fault observer uses the observation gain matrix to provide feedback adjustment to the output error of the augmented system model, thereby quickly adjusting the estimation direction of the system state and fault. S5: By leveraging the adaptive law, the compensation amount for interference observation errors is continuously optimized, automatically offsetting the errors generated during the interference estimation process; S6: Set a minimum threshold. After the adaptive sensor fault observer obtains the degree of sensor fault, it constructs a decision-making mechanism to isolate the fault according to the minimum threshold. S7. Define the tracking error between the robot's actual trajectory and the desired inspection trajectory, design a generalized sliding surface and control law, and combine the exponential and variable speed approach laws to accelerate the convergence of the sliding surface and reduce chattering.

[0030] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A fault-tolerant control system for sensor faults in a smart power plant inspection robot, characterized in that, in: The motion model construction module (100) utilizes the law of inertia and the motion characteristics of wheeled mobile robots to construct a motion model that can simultaneously reflect the robot's state, external interference, and sensor failures. The interference observer design module (200) defines the observation error based on the constructed motion model; Design an interference observer that is based on a motion model and incorporates feedback adjustment based on observation errors; The augmented modeling and adaptive sensor fault observation module (300) uses sensor fault as an independent variable for augmentation, integrates the robot's original motion state with the sensor fault as the augmented state, and constructs an augmented system model; based on the augmented system model, an adaptive sensor fault observer is designed, and the adaptive sensor fault observer adjusts the output error of the augmented system model by observing the gain matrix. By leveraging the adaptive law, the compensation amount for interference observation errors is continuously optimized, automatically offsetting the errors generated during interference estimation; and by setting a minimum threshold, the adaptive sensor fault observer, after acquiring the degree of sensor fault, constructs a decision-making mechanism according to the minimum threshold to isolate the fault. The fault-tolerant controller design module (400) defines the tracking error between the robot's actual trajectory and the expected inspection trajectory, designs a generalized sliding surface and control law, and combines the exponential approach law and the variable speed approach law to accelerate the convergence of the sliding surface and reduce chattering.

2. The fault-tolerant control system for sensor faults of the smart power plant inspection robot according to claim 1, characterized in that: The motion model in the motion model construction module (100) is: ； ； in Let be the state vector of the robot's motion state. Let the coordinates be the robot's position coordinates on the plane. For heading angle, The rotational angular velocity about an axis perpendicular to the plane of motion; To control the input vector, For control input of the left and right wheels; This is the external interference vector. Interference with the direction of motion Interference with steering direction; For the fault vector, The problem is a position sensor malfunction. The problem is a steering sensor malfunction. For the system matrix, The coefficient of viscous friction for linear motion; It is the coefficient of viscous friction in steering motion, used to quantify the hindering effect of viscous friction on the motion state during steering. To control the input matrix, where Gain for linear motion drive For steering motion drive gain; This is the output matrix; The above, The coefficient of viscous friction, For friction correlation coefficient, The distance from the left and right wheels to the robot's center of gravity. Let the moment of inertia be the rotation about the robot's center of gravity. For the wheel radius, Let be the moment of inertia of the wheel. This is the correlation coefficient of the steering motion drive gain.

3. The fault-tolerant control system for sensor faults of the smart power plant inspection robot according to claim 1, characterized in that: The interference observer in the interference observer design module (200) includes velocity observation and interference estimation equations, and angular velocity observation and interference estimation equations; The velocity observation and disturbance estimation equations are as follows: ； ； in: To estimate the linear velocity, To estimate the linear velocity The first derivative, This is the robot's actual linear velocity. For linear motion control input, For linear motion disturbance, This is an estimate of the disturbance in the direction of motion. For linear velocity estimation error, is the coefficient of viscous friction for linear motion in the motion model. For linear motion driving gain in the motion model, For inputs related to speed control, This is an estimate of the disturbance in the direction of motion. For observer gain The equations for angular velocity observation and disturbance estimation are as follows: ； ； in: To estimate angular velocity, To estimate angular velocity The first derivative, For the control input of steering motion, For the amount of interference with steering motion The estimated value of ) For the steering motion viscous friction coefficient in the motion model, This represents the steering motion drive gain in the motion model. For inputs related to angular velocity control, This is an estimate of the steering direction disturbance. This is the observer gain.

4. The fault-tolerant control system for sensor faults of the smart power plant inspection robot according to claim 3, characterized in that: The interference observer design module (200) causes the robot's actual speed or angular velocity to deviate from the predicted value of the motion model when external interference is applied to the robot, thus resulting in speed observation error. or angular velocity observation error The interference observer increases; while the interference observer dynamically adjusts the internal velocity observation and interference estimation equations and angular velocity observation and interference estimation equations by monitoring changes in observation errors in real time, thus increasing the interference estimate. and By gradually approximating the true interference value, external interference is eventually separated from the total error.

5. The fault-tolerant control system for sensor faults of the smart power plant inspection robot according to claim 4, characterized in that: The augmented state in the augmented modeling and adaptive sensor fault observation module (300) includes the robot's original state and the sensor's fault vector. The state equations of the augmented system model are: ； in; To augment the system model matrix, The system matrix represents the original motion model of the robot. To augment the control input matrix; This is the control input matrix for the robot's original motion model; For the control input of the robot; For the interference input matrix, This represents a 4th-order interference input matrix, where all elements on the main diagonal are 1 and the rest are 0. The interference term, which includes the interference observation error, represents the impact of interference on the augmented system model. To prevent observation errors before the interference observer converges; This is the adaptive term matrix; To augment the system model output equations The derivative of This represents the second-order interference input matrix; The output equation of the augmented system model is ,in To augment the system model's output vector, it includes sensor measurements of faulty output signals; To augment the output matrix, This is the output matrix of the robot's original motion model.

6. The fault-tolerant control system for sensor faults of the smart power plant inspection robot according to claim 5, characterized in that: The adaptive sensor fault observer in the augmented modeling and adaptive sensor fault observation module (300) is: ； ； ； in: in, For the control input of the robot; This is the adaptive term matrix; To augment the output vector of the system model's output equations; The interference input matrix; The external disturbance value estimated by the disturbance observer; This is a compensation term for interference observation errors. For adaptive weights, update according to the following adaptive law. , It is a very small positive number. This is the correlation coefficient for the output error.

7. The fault-tolerant control system for sensor faults of the smart power plant inspection robot according to claim 1, characterized in that: The adaptive law in the augmented modeling and adaptive sensor fault observation module (300) is: ； in: The update rate for adaptive weights; It is a positive definite matrix; Let be the norm of the output error; The fault isolation: After the adaptive sensor fault observer obtains the degree of sensor fault, it constructs a decision-making mechanism to isolate the fault according to the minimum threshold determined through multiple experiments. First, compare the observed fault severity with the minimum threshold. If the fault severity is greater than the threshold, then the fault signal of the corresponding sensor is blocked; if the fault severity is less than the threshold, then no action is taken.

8. The fault-tolerant control system for sensor faults of the smart power plant inspection robot according to claim 1, characterized in that: The fault-tolerant controller design module (400) defines the tracking error as follows: The following generalized sliding surface is designed: ； in For the first The sliding surface function of each channel is the tracking error. Error derivative A linear combination of the error integral; For the first The integral term of the channel error; The control law is designed as follows: ； ； in To prevent interference with observation errors The estimated value The update law in the controller is ; These are the control input, control input gain, system matrix elements, state variables, estimated disturbance values, and second derivatives of the desired state for the k-th control channel, respectively. Let be the error feedback gain coefficient of the k-th channel; Let be the first derivative of the error of the k-th control channel, the sliding surface gain coefficient, the sliding surface function, the variable speed reaching law gain coefficient, the variable speed reaching law function, the sign function, and the estimated value of the uncertainty term, respectively.

9. The fault-tolerant control system for sensor faults of the smart power plant inspection robot according to claim 8, characterized in that: The fault-tolerant controller design module (400) sets a new type of reaching law: ; Using the exponential reaching law and the law of convergence of speed In combination, when the state variable is far from the sliding surface, the exponential reaching law dominates; while when the state variable gradually approaches the vicinity of the sliding surface, the variable reaching law dominates. As the state variable moves on the sliding surface... As the state variables decrease, the motion diminishes until a stable point is reached.

10. A fault-tolerant control method for sensor faults in a smart power plant inspection robot, applied to the fault-tolerant control system for sensor faults in a smart power plant inspection robot as described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Utilize the laws of inertia and the motion characteristics of wheeled mobile robots to construct a motion model that can simultaneously reflect the robot's state, external disturbances, and sensor malfunctions. S2: Define the observation error based on the established motion model; Design an interference observer that is based on a motion model and incorporates feedback adjustment based on observation errors; S2: Augment the sensor fault as an independent variable, integrate the robot's original motion state with the sensor fault as the augmented state, and construct an augmented system model. S4: An adaptive sensor fault observer is designed based on an augmented system model. The adaptive sensor fault observer uses the observation gain matrix to provide feedback adjustment to the output error of the augmented system model, thereby quickly adjusting the estimation direction of the system state and fault. S5: By leveraging the adaptive law, the compensation amount for interference observation errors is continuously optimized, automatically offsetting the errors generated during the interference estimation process; S6: Set a minimum threshold. After the adaptive sensor fault observer obtains the degree of sensor fault, it constructs a decision-making mechanism to isolate the fault according to the minimum threshold. S7. Define the tracking error between the robot's actual trajectory and the desired inspection trajectory, design a generalized sliding surface and control law, and combine the exponential and variable speed approach laws to accelerate the convergence of the sliding surface and reduce chattering.

Images