AUV Actuator Fault Diagnosis Method Based on Fault Factors and Multiple Observers

Abstract

The present invention discloses an AUV actuator fault diagnosis method based on fault factors and multiple observers, and the main steps include: (1) using fault factors to describe the actuators of the AUV and modeling through state equations; (2) designing a group of fault isolation extended state observers (ESOs), where each observer is only sensitive to one type of fault, and judging the generated residual group through certain logical rules to achieve fault isolation of the actuators; (3) identifying faults in the rudder surface, propeller and steering gear in the actuators; (4) considering the deformation fault of the rudder surface and analyzing its deformation situation according to the positive and negative of the fault factors; (5) based on the output force and torque formulas of the propeller and steering gear, identifying and diagnosing their fault situations by analyzing the internal information of the fault factors.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis of underwater robot power systems, and relates to a fault diagnosis method for an AUV (Autonomous Underwater Vehicle) actuator based on fault factors and multiple observers. Background Art

[0002] In recent years, the discovery and exploration of marine resources have been continuously developed and gradually become the mainstream. As a tool for humans to explore and develop marine resources, the autonomous underwater vehicle (AUV) plays an important role in the marine field. Since the AUV needs to work in the ocean for a long time, the conditions in the ocean are extremely harsh and the surrounding environment is complex and changeable, and various faults may occur in the actuators of its motion control system. If the faults are not detected in time, the AUV will work in an unpredictable manner, which will not only shorten the service life of the AUV, but also affect its underwater work tasks, threatening the safety of personnel and equipment, and ultimately bringing catastrophic consequences. The actuator is the most common and important fault source of the AUV. As a carrier working in a complex marine environment, state monitoring and fault diagnosis are the foundation and core technologies for studying the safety issues of the AUV. Therefore, studying the fault diagnosis technology of the AUV actuator to improve the safety and reliability of the AUV has become an urgent research task and one of the hotspots in the current scientific research community.

[0003] The existing fault diagnosis technologies are generally divided into three categories: data-driven methods, analytical model-based methods, and knowledge-based methods. Among them, the AUV fault diagnosis method based on an analytical model has good diagnostic effects in a static water environment and is easy to achieve real-time diagnosis, directly providing useful information for the next-step fault-tolerant control or fault recovery. Different from other methods, this method relies on an accurate mathematical model of the object to be diagnosed and uses relevant parameters in the model for method research. Fault factors and multiple observers belong to the state estimation method in the analytical model. The state estimation method is a method that can estimate the internal state of a dynamic system by using the obtained measurement data. The data obtained by measuring the input and output of the system can only reflect the external characteristics of the system, while the dynamic law of the system needs to be described by internal state variables. The state estimation method first reconstructs the state of the controlled process, generates a residual sequence by comparing with measurable variables, then constructs an appropriate model and uses a statistical detection method to detect faults from the residual sequence and perform further separation, estimation, and decision-making.

[0004] At present, the fault diagnosis technology based on the state estimation method has the following defects: 1) The state estimation method needs to linearize the non-linear model part, and for the actual system with unstructured uncertainty, decoupling cannot be achieved; 2) The state estimation method requires a relatively accurate model, so the calculation amount is relatively large and the stability is poor; 3) The parameter calculation of the state estimation method is relatively large, and more parameters will affect the accuracy of fault diagnosis; 4) The ideas of fault isolation and identification are not clear, and a complete fault diagnosis mechanism cannot be formed between the two; 5) It is difficult to identify all the faults of the rudder surface, rudder machine and propeller in the AUV actuator by using the state estimation method, and there will be a large error in the identification result. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and provide a fault diagnosis method for an AUV actuator based on fault factors and multiple observers. The multiple extended state observers adopted by the method of the present invention reduce the calculation amount of parameters and avoid the problem of system decoupling; each observer is only sensitive to a certain kind of fault and insensitive to other faults, and has strong robustness; the output formulas of the rudder surface, rudder machine and propeller are further used for fault identification, and have high accuracy.

[0006] To solve the above technical problems, the present invention adopts the following technical solutions:

[0007] A fault diagnosis method for an AUV actuator based on fault factors and multiple observers includes the following steps:

[0008] Step 1. Introduce multiplicative fault description factors and additive fault description factors to respectively describe the faults of the rudder surface, rudder machine and propeller in the AUV actuator, introduce the fault description factors into the state equation, and model the faults in the state equation;

[0009] Step 2. Use the state equation containing the actuator faults in Step 1 to divide the modeled fault vector into two parts, combine one part of the fault vector with the disturbance vector to form a new disturbance vector, and keep the other part of the fault vector as a prerequisite for Step 3;

[0010] Step 3. Use the scheme in Step 2 to design multiple extended state observers, and judge the residuals generated by the multiple extended state observers according to the set cooperation rules to realize the fault isolation of the rudder surface, rudder machine and propeller in the AUV actuator;

[0011] Step 4. When the rudder surface fault is isolated in Step 3, consider the small-angle deformation fault of the rudder surface, design a rudder surface output formula containing an additive fault description factor, and analyze and identify the deformation fault situation according to the positive and negative of the output force and torque;

[0012] Step 5. When the propeller and rudder failures are isolated in Step 3, an output formula for the propeller and the steering gear that includes a multiplicative fault description factor is designed, and the fault condition is analyzed and identified by analyzing the change of the multiplicative fault description factor.

[0013] Further, the specific content and method of Step 1 include the following steps:

[0014] Step 1.1 To describe the fault more accurately, two forms of description of the AUV actuator fault are given here.

[0015] (1) Additive description form:

[0016] u = u * + F + (1)

[0017] where u is the control input of the AUV actuator, and u * is the desired input value; F + represents the fault vector in u, f + is the additive fault description factor.

[0018] (2) Multiplicative description form:

[0019] u = F × u * (2)

[0020] where F × is the diagonal matrix representing the fault amplitude, and it is defined that when f × ≠ 1, the control input carries fault information, and at this time, f × is called the multiplicative fault description factor.

[0021] Step 1.2 Model the specific faults in the AUV actuator, and the method and steps are as follows:

[0022] (1) Power propulsion actuator fault:

[0023] The propeller is the actuator in the AUV that directly acts on the fluid to generate forward thrust. The relationship between the propeller output thrust and its rotational speed is introduced as:

[0024] T = K T n|n| + K v nV a (3)

[0025] where T is the thrust, n is the propeller rotational speed, and V a is the motor propulsion speed; the variables related to the propeller fault include: the thrust coefficient K T and the motor propulsion coefficient Kv .

[0026] (2) Attitude control actuator failure:

[0027] The attitude control system of the AUV mainly includes the rudder surface and the rudder machine. The failure of the rudder surface is a small-angle bending deformation at the edge. The deformed rudder surface interacts with the fluid to generate a force parallel to the vehicle coordinate system. The additive fault generated by its interaction with the fluid is:

[0028]

[0029] In the formula to are the additive faults of each degree of freedom; and F * are the fluid forces during failure and without failure respectively. In the attitude control system, the rudder machine is used to provide the torque for the rudder surface deflection. The given torque formula of the rudder machine is:

[0030] M = DK M n|n| + DK v nV a (5)

[0031] In the formula, M is the torque; D is the propeller diameter; the variables related to the rudder machine failure include: the torque coefficient K M and the motor propulsion coefficient K v .

[0032] For the force and torque formulas of the propeller and the rudder machine, the multiplicative fault description factor f * and the additive fault description factor f + are introduced, and we can get:

[0033]

[0034] Step 1.3 For the modeling and description of the AUV actuator system, consider the following form of nonlinear system:

[0035]

[0036] In the formula, x, y, u, d are the state vector, output vector, input vector and disturbance vector respectively; g(x) is the nonlinear term; A, B, C, D are constant coefficient matrices. Separating the changes caused by each fault in the actuator from the above state equation and rewriting the state equation in the form containing the fault description factor, we can get:

[0037]

[0038] Among them

[0039]

[0040] where \(G(\cdot)\) is a piecewise fault function, including additive faults, multiplicative faults, and input variables carrying these fault vectors; \(F = [F + F × = [F_1 F_2 \cdots F n is a fault distribution matrix, and the fault matrices therein can respectively characterize the rudder surface deformation fault, the propeller thrust fault, and the steering gear thrust moment fault.

[0041] Furthermore, the specific content and method of step 2 include the following steps:

[0042] Step 2.1 Using the state equation in step 1.3, define \(F a = [F a + F a × = [F i F i+1 \cdots F j , and \(1 \lt i \lt j \lt n\). The remaining fault vectors in \(G\) are combined into a new matrix defined as \(F b \), and removed from \(G\) to form a new distribution matrix with the disturbance \(D\), and the piecewise fault function \(G\) that only contains the fault matrix \(F a is called \(G_1\);

[0043] Step 2.2 Rewrite the state equation in step 1.3 as:

[0044]

[0045] where and are the new disturbance matrix and vector respectively,

[0046] The above structure is equivalent to classifying a part of the actuator fault vector as a new disturbance vector The goal of such a design is to make the residual generated by the observer not affected by the disturbance but only affected by the new piecewise fault function \(G_1\). Since the new disturbance vector contains the disturbance \(D\) and part of the fault \(F b ), the residual is insensitive to the disturbance and the actuator fault carrying the fault \(F b and sensitive to the actuator fault \(G_1\). According to the above idea, theoretically, fault isolation observers with the same number as the types of actuator faults can be designed.

[0047] Furthermore, the specific content and method of step 3 include the following steps:

[0048] Step 3.1 Expand other system variables except the piecewise fault function \(G_1\) into a new state variable \(x_2\), and we can get:

[0049]

[0050] Thus, the form of the designed extended state observer is as follows:

[0051]

[0052] where \(i = 1, 2, \ldots, n\) is the observer corresponding to the \(i\)-th fault type; \(z\) 1,i , \(z\) 1,i are the estimates of the state variables \(x\) 1,i , \(x\) 1,i respectively, and \(l\) 1,i , \(l\) 1,i is the gain vector of the extended state observer;

[0053] Step 3.2 Calculate the gain of the extended state observer. The gain can usually be parameterized as:

[0054] \([l\) 1,i \(l\) 2,i = [\(\beta_1\omega_0\) \(\beta_2\omega_0\) 2 (13)

[0055] where \(\beta_1\) and \(\beta_2\) are selected parameters such that the characteristic polynomial \(\beta_2s+\beta_1\) is Hurwitz. Let

[0056] \(s\) (n+1) +\(\beta_1s\) n + \(\cdots\) + \(\beta\) n \(s+\beta\) n+1 = \((s + 1)\) (n+1) (14)

[0057] It should ensure that the roots of the characteristic polynomial are in the left half of the complex plane. Therefore, the parameters in equation (6) can be selected as:

[0058] \([l\) 1,i \(l\) 2,i = [\(2\omega_0\) \(\omega_0\) 2 (15)

[0059] where \(\omega_0\) is the bandwidth of the observer, and this bandwidth needs to be continuously changed in practice to ensure that the ESO can correctly estimate the state variables.

[0060] Furthermore, the specific content and method of step 4 include the following steps:

[0061] Step 4.1 Analyze the force on the faulty deformation part of the rudder surface. Consider the rudder surface in the actuator as a planar rectangle \(ABCD\), and the rudder surface angle at this time is \(\delta\) X, the area indicated by the dashed line is the part of the rudder surface with deformation failure. The deformation failure is represented as the triangle ABC rotating downward by an angle along the straight line AB, which is called the deformation angle δ f , and the angle generated on the plane is the fault plane angle r e . Subtracting the axial fluid force under the fault from that without fault, the fault forces in each axial direction can be obtained. Accordingly, the fault forces and moments generated by the deformed rudder surface ABC under the action of the fluid can be approximately calculated as follows:

[0062]

[0063] In the formula, x e is the vertical distance from the edge of the transverse rudder surface to the z-axis; P is the fluid pressure; l is the length of the deformed part of the rudder surface

[0064] Step 4.2 Further analyze Step 4.1. When different types of deformation failures occur on the rudder surface, there are situations where the fault effects are opposite in sign. Analyze the signs of the fault effects caused by all rudder surfaces under possible deformation conditions as shown in the following table:

[0065]

[0066] By arranging and combining the four fault forces and moments, 8 basic deformation failures can be obtained. Considering that only one rudder surface has a fault at the same time, and there are differences in the signs of the output effects of the deformation failures of different rudder surfaces. Thus, the identification of the faulty rudder surface can be achieved by combining the fault estimation results and the above table

[0067] Furthermore, the specific content and method of Step 5 include the following steps:

[0068] Step 5.1 Considering that there are two unknowns in the thrust formula and it cannot be solved only by the estimation data at one moment. And the fault degrees of the propeller and the rudder machine can be ignored in the short term. Therefore, take the measurement and estimation data of two adjacent time periods t1 and t2 with parameter changes, and convert Equation (6) into a matrix equation for solving the multiplicative fault description factor, and we can get:

[0069] (1) Thrust of the propeller:

[0070]

[0071] (2) Thrust moment of the rudder:

[0072]

[0073] Step 5.2 Use the multiplicative fault matrix F × (t) in Step 5.1, and judge the multiplicative fault description factor f ×(t) information to further identify the fault information of the propeller and rudder, as shown in the following table:

[0074]

[0075]

[0076] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0077] 1. The present invention uses fault description factors to divide the fault isolation and identification strategy into two levels, forming a complete fault diagnosis mechanism. In practical applications, a complete and logically correct fault diagnosis system can handle faults, avoiding many unnecessary troubles and reducing property losses during the process.

[0078] 2. Among the multiple extended state observers designed in the present invention, each observer is only sensitive to a certain type of fault and insensitive to other faults, having strong robustness. In practical applications, when a fault occurs in an AUV (Autonomous Underwater Vehicle), the above-mentioned fault isolation scheme can locate the fault in a timely manner, provide an alarm signal, and avoid accidents; if the fault is misjudged and no corresponding measures are taken, accidents will occur, causing great economic losses.

[0079] 3. The present invention combines the fault description factors with the output formulas of the rudder surface, rudder machine and propeller. The above method can identify the faults occurring in the system based on the historical data running in the system, and can analyze the faults in the system in real time online. In practical applications, the type of system faults can be pre-identified, and fault prevention and handling can be done in advance, saving operation costs, so it has important theoretical value and practical significance.

[0080] 4. The present invention can identify all the faults existing in the rudder surface, rudder machine and propeller in the AUV actuator. In practical applications, it can judge the damage degree of the propeller blades, how many specific blades are lost, and whether it reaches the degree that needs to be replaced; whether the deformation fault of the rudder surface is serious, and how many degrees the direction angle provided by the rudder machine is specifically offset due to the fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 is a process diagram of the fault diagnosis method based on the analytical model.

[0082] Figure 2 is a schematic diagram of an autonomous underwater vehicle in a coordinate system.

[0083] Figure 3 is the fault diagnosis flowchart of the AUV actuator based on the description factor and multiple observers

[0084] Figure 4 It is the first-step sub-flowchart of an embodiment of the present invention.

[0085] Figure 5 It is the schematic diagram of the AUV actuator fault isolation scheme.

[0086] Figure 6 It is the second- and third-step sub-flowcharts of an embodiment of the present invention.

[0087] Figure 7 It is the fourth-step sub-flowchart of an embodiment of the present invention.

[0088] Figure 8 It is the schematic diagram of the small-angle deformation of the rudder surface.

[0089] Figure 9 It is the fifth-step sub-flowchart of an embodiment of the present invention.

[0090] Figure 10a It is the fault isolation diagram of the rudder surface.

[0091] Figure 10b It is the fault identification diagram of the upward deformation of the left horizontal rudder surface.

[0092] Figure 10c It is the fault identification diagram of the rightward deformation of the upper horizontal rudder surface.

[0093] Figure 11a It is the fault isolation diagram of the propeller.

[0094] Figure 11b It is the identification diagram of three types of propeller faults.

[0095] Figure 12a It is the fault isolation diagram of the steering gear.

[0096] Figure 12b It is the identification diagram of three types of steering gear faults. Detailed implementation manners

[0097] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

[0098] Figure 1 It is the schematic diagram of the principle of the AUV actuator fault diagnosis method based on the analytical model of the present invention. The controller controls the actuator to obtain the state value, the extended state observer obtains the estimated value, and finally the residual is obtained by comparison, and the fault information is processed and decision-making in the residual sequence.

[0099] Figure 3 It is the flowchart of an embodiment of the AUV actuator fault diagnosis method based on the fault factor and multi-observer of the present invention. This embodiment specifically includes the following steps:

[0100] Step 1. Introduce multiplicative fault description factor and additive fault description factor, and respectively describe the faults of the rudder surface, steering gear and propeller in the AUV actuator. Introduce the fault description factor into the state equation and model the fault in the state equation;

[0101] In Step 1, the specific content and design steps of the fault description of the rudder surface, steering gear and propeller of the AUV actuator are as follows (as Figure 4 shown):

[0102] Step 1.1 To describe the fault more accurately, two forms of description of the AUV actuator fault are given here.

[0103] (1) Additive description form:

[0104] u = u * + F + (19)

[0105] In the formula, u is the control input of the AUV actuator, and u * is the desired input value; F + represents the fault vector in u, f + is the additive fault description factor.

[0106] (2) Multiplicative description form:

[0107] u = F × u * (20)

[0108] In the formula, F × is the diagonal matrix representing the fault amplitude, and it is defined that When f × ≠ 1, the control input carries fault information. At this time, f × is called the multiplicative fault description factor.

[0109] Step 1.2 Model the specific faults in the actuator as follows:

[0110] (1) Fault of the power propulsion actuator:

[0111] The propeller is the actuator in the AUV that directly acts on the fluid to generate forward thrust. The relationship between the output thrust of the propeller and its rotational speed is introduced as:

[0112] T = K T n|n| + K v nV a (21)

[0113] In the formula, T is the thrust, n is the rotational speed of the propeller, and V ais the motor propulsion speed; the variables related to the propeller failure include: the thrust coefficient K T and the motor propulsion coefficient K v .

[0114] (2) Attitude control actuator failure:

[0115] The attitude control system of the AUV mainly includes the rudder surface and the steering gear. The failure of the rudder surface is a small-angle bending deformation at the edge. The deformed rudder surface acts on the fluid to generate a force parallel to the vehicle coordinate system. The additive failure generated by its action on the fluid is:

[0116]

[0117] In the formula to are the additive failures of each degree of freedom; and F * are the fluid forces during failure and without failure respectively. In the attitude control system, the steering gear is used to provide the torque for the rudder surface deflection. The given torque formula of the steering gear is:

[0118] M = DK M n|n| + DK v nV a (23)

[0119] In the formula, M is the torque; D is the propeller diameter; the variables related to the steering gear failure include: the torque coefficient K M and the motor propulsion coefficient K v .

[0120] For the force and torque formulas of the propeller and the steering gear, multiplicative failure description factor f * and additive failure description factor f + are introduced, and we can get:

[0121]

[0122] Step 1.3 For the modeling and description of the AUV actuator system, consider the following form of nonlinear system:

[0123]

[0124] In the formula, x, y, u, d are the state vector, output vector, input vector and disturbance vector respectively; g(x) is the nonlinear term; A, B, C, D are constant coefficient matrices. Separate the changes caused by each failure in the actuator from the above state equation, and rewrite the state equation into a form containing failure description factors, we can get:

[0125]

[0126] Among them

[0127]

[0128] Among them, G(·) is a piecewise fault function, including additive faults, multiplicative faults, and input variables carrying these fault vectors; F = [F + F × = [F1 F2 … F n is a fault distribution matrix, and the fault matrices therein can respectively characterize the rudder surface deformation fault, the propeller thrust fault, and the rudder actuator thrust moment fault.

[0129] Step 2. Use the state equation containing actuator faults in Step 1 to divide the modeled fault vectors into two parts, combine one part of the fault vectors with the disturbance vector to form a new disturbance vector, and retain the other part of the fault vectors as a prerequisite for Step 3;

[0130] In Step 2, the specific content and design steps of the fault isolation scheme are as follows (as Figure 6 shown):

[0131] Step 2.1 Use the state equation in Step 1.3 to define F a = [F a + F a × = [F i F i+1 … F j , and 1 < i < j < n. The remaining fault vectors in G are combined to form a new matrix defined as F b , and removed from G, and combined with the disturbance D to form a new distribution matrix. The piecewise fault function G that only contains the fault matrix F a is called G1;

[0132] Step 2.2 Rewrite the state equation in Step 1.3 as:

[0133]

[0134] Among them and are the new disturbance matrix and vector respectively,

[0135] The above structure is equivalent to classifying a part of the actuator fault vectors into the new disturbance vector The goal of such a design is to make the residual generated by the observer not affected by the disturbance but only affected by the new piecewise fault function G1. Since the new disturbance vector contains the disturbance D and part of the fault F b , so the residual is for the disturbance and the carried fault Fb is insensitive to the actuator fault G0, but sensitive to the actuator fault G1. According to the above idea, theoretically, fault isolation observers with the same number as the types of actuator faults can be designed (as shown in Figure 5 ).

[0136] Step 3. Using the scheme in Step 2, design multiple extended state observers, and discriminate the residuals generated by the multiple extended state observers according to the set cooperation rules to achieve fault isolation of the rudder surface, steering gear and propeller in the AUV actuator;

[0137] In Step 3, the specific content and design steps of the multiple extended state observers designed based on the fault isolation scheme are as follows:

[0138] Step 3.1 Expand other system variables except the piecewise fault function G1 into a new state variable x2, and we can get:

[0139]

[0140] Thus, the form of the extended state observer is designed as:

[0141]

[0142] where i = 1, 2,..., n is the observer corresponding to the i-th fault type; z 1,i , z 1,i are the estimates of the state variables x 1,i , x 1,i respectively, and l 1,i , l 1,i is the gain vector of the extended state observer;

[0143] Step 3.2 Calculate the gain of the extended state observer. The gain can usually be parameterized as:

[0144] [l 1,i l 2,i = [β1ω0 β2ω0 2 (31)

[0145] where β1, β2 are the selected parameters such that the characteristic polynomial β2s + β1 is Hurwitz. Let

[0146] s (n+1) + β1s n + … + β n s + β n+1 = (s + 1) (n+1) (32)

[0147] It should ensure that the roots of the characteristic polynomial are in the left half of the complex plane. Therefore, the parameters in Equation (6) can be selected as:

[0148] [l 1,i l 2,i = …2ω0 ω0 2 (33)

[0149] where ω0 is the bandwidth of the observer, and this bandwidth needs to be continuously changed in practice to ensure that the ESO can correctly estimate the state variables.

[0150] Step 4. When the actuator fault is isolated in Step 3, consider the small-angle deformation fault of the actuator surface, design an actuator surface output formula including an additive fault description factor, and analyze and identify the deformation fault situation according to the positive and negative of the output force and moment;

[0151] In Step 4, consider the small-angle deformation fault of the actuator surface, and analyze the deformation situation according to the positive and negative of the fault force and moment in the actuator surface output formula. The specific content and design steps are as ( Figure 7 shown):

[0152] Step 4.1 Analyze the forces on the fault deformation part of the actuator surface (as ( Figure 8 shown). Consider the actuator surface in the actuator as a planar rectangle ABCD, and the actuator surface angle at this time is δ X , and the area indicated by the dotted line is the part of the actuator surface with the deformation fault. The deformation fault is expressed as that triangle ABC rotates downward by an angle along line AB, called the deformation angle δ f , and the angle generated on the plane is the fault plane angle r e . Subtract the axial fluid force under the fault from that without fault to obtain the fault forces in each axial direction. Accordingly, the fault force and moment generated by the deformed actuator surface ABC under the action of the fluid can be approximately calculated:

[0153]

[0154] In the formula, x e is the vertical distance from the edge of the transverse actuator surface to the z-axis; P is the fluid pressure; l is the length of the deformed part of the actuator surface.

[0155] Step 4.2 Further analyze Step 4.1. When different types of deformation faults occur on the actuator surface, there are situations where the fault actions are opposite in positive and negative. Analyze the positive and negative of the fault actions caused by all possible deformations of the actuator surface as shown in the following table:

[0156] Deformation condition <![CDATA[f Y > <![CDATA[f Z > <![CDATA[f M > <![CDATA[f N > The left horizontal rudder surface deforms upward - + + + The left horizontal rudder surface deforms downward 0 0 0 1 The right horizontal rudder surface deforms upward 1 1 1 0 The right horizontal rudder surface deforms downward 1 0 0 0 The upper horizontal rudder surface deforms to the left 1 0 0 1 The upper horizontal rudder surface deforms to the right 0 0 1 1 The lower horizontal rudder surface deforms to the left 1 1 0 0 The lower horizontal rudder surface deforms to the right 0 1 1 0

[0157] By arranging and combining the four types of fault forces and torques, 8 basic deformation faults can be obtained. Considering that only one control surface has a fault at the same time, and there are differences in the positive and negative of the output effects of the deformation faults of different control surfaces. Therefore, combining the fault estimation results and the above table can achieve the identification of the faulty control surface.

[0158] Step 5. When the propeller and rudder faults are isolated in Step 3, design the output formulas of the propeller and the steering gear that include multiplicative fault description factors, and analyze and identify their fault conditions by analyzing the changes in the multiplicative fault description factors.

[0159] In Step 5, based on the output force and torque formulas of the propeller and the steering gear, identify their fault conditions by analyzing the internal information of the fault factors. The specific content and design steps are as follows (as Figure 9 shown):

[0160] Step 5.1 Considering that there are two unknowns in the thrust formula and it cannot be solved only by the estimation data at one moment. And the fault degrees of the propeller and the steering gear change negligibly in a short time. Therefore, take the measurement and estimation data of two adjacent time periods t1 and t2 with parameter changes, and convert Equation (6) into a matrix equation for solving the multiplicative fault description factor, and we can get:

[0161] (1) Thrust of the propeller:

[0162]

[0163] (2) Thrust moment of the rudder:

[0164]

[0165] Step 5.2 Use the multiplicative fault matrix F × (t) in Step 5.1 to further identify the fault information of the propeller and the rudder by discriminating the information of the multiplicative fault description factor f × (t) inside it, as shown in the following table:

[0166]

[0167]

[0168] The method of the present invention is verified by simulation as follows:

[0169] Step A. Use the MATLAB program to write a six-degree-of-freedom model of the autonomous underwater vehicle, and at the same time design three extended state observers, corresponding to the control surface, the steering gear and the propeller faults respectively, to achieve the fault isolation of the actuators;

[0170] Step B. Write the output formulas for the rudder surface, servo, and propeller in MATLAB, and identify faults by analyzing their respective fault description factors.

[0171] Figure 10a This is the fault isolation effect diagram for the rudder surface. It is not difficult to find from the figure that the observer designed based on the fault isolation scheme is only sensitive to a certain type of fault; therefore, the synchronous cooperation of multiple observers can effectively isolate faults in the AUV actuator.

[0172] Figure 10b This is the fault identification diagram for the upward deformation fault of the left horizontal rudder surface. Figure 10c This is the fault identification diagram for the rightward deformation fault of the upper horizontal rudder surface. By observing the changes in each curve in the figure, it can be seen that the simulation results are consistent with the designed fault types, thus verifying the effectiveness of the rudder surface deformation fault identification.

[0173] Figure 11a This is the fault isolation effect diagram for the propeller. It can be seen from the figure that the synchronous cooperation of multiple observations can effectively isolate propeller faults.

[0174] Figure 11b This is the identification effect diagram for three fault types of the propeller. It can be seen from the figure that the multiplicative fault description factor can further analyze the fault types of the propeller and achieve the effect of fault identification.

[0175] Figure 12a This is the fault isolation effect diagram for the servo. It can be seen from the figure that the synchronous cooperation of multiple observations can effectively isolate servo faults.

[0176] Figure 12b This is the identification effect diagram for three fault types of the servo. It can be seen from the figure that the multiplicative fault description factor also has a good effect on analyzing the specific fault types inside the servo.

[0177] As can be seen from the above figures, the method of the present invention can effectively achieve fault isolation and identification of the actuator of the autonomous underwater vehicle, effectively solve the problems of actuator fault diagnosis and its engineering application, which is of great significance for the safe operation of the autonomous underwater vehicle.

[0178] The above has described the embodiments of the present invention in detail with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those of ordinary skill in the art, various changes can be made without departing from the gist of the present invention.

Claims

1. A fault diagnosis method for AUV actuators based on fault factors and multi-observers, characterized in that It includes the following steps: Step 1. Introduce multiplicative fault description factors and additive fault description factors to describe the faults of the rudder surface, steering gear, and propeller in the AUV actuator respectively. Introduce the fault description factors into the state equation and model the faults in the state equation. For the modeling and description of the AUV actuator system, consider a nonlinear system in the following form: where x, y, u, and d are the state vector, output vector, input vector, and disturbance vector respectively; g(x) is the nonlinear term; A, B, C, and D are constant coefficient matrices. Separate the changes caused by each fault in the AUV actuator from the above state equation and rewrite the state equation in a form containing fault description factors, obtaining: where Among them, G(·) is a segmented fault function, including additive faults, multiplicative faults, and input variables carrying these fault vectors; F = [F + F × = [F1 F2 … F n is a fault distribution matrix, and the fault matrices therein respectively represent the deformational faults of the rudder surface, the thrust faults of the propeller, and the thrust torque faults of the steering gear; Step 2. Use the state equation containing actuator faults in Step 1 to divide the modeled fault vector into two parts. Combine one part of the fault vector with the disturbance vector to form a new disturbance vector, and keep the other part of the fault vector as a prerequisite for Step 3. Step 3. Use the scheme in Step 2 to design multiple extended state observers. Discriminate the residuals generated by the multiple extended state observers according to the set cooperation rules to achieve fault isolation of the rudder surface, steering gear, and propeller in the AUV actuator. Step 4. When the rudder surface fault is isolated in Step 3, consider the small-angle deformation fault of the rudder surface, design a rudder surface output formula containing an additive fault description factor, and analyze and identify its deformation fault situation according to the positive and negative of the output force and torque. Step 5. When the propeller and rudder faults are isolated in Step 3, design output formulas for the propeller and steering gear containing multiplicative fault description factors, and analyze and identify their fault situations by analyzing the changes in the multiplicative fault description factors.

2. The AUV actuator fault diagnosis method based on fault factors and multiple observers according to claim 1, characterized in that The two description forms for describing the faults of the rudder surface, steering gear, and propeller in the AUV actuator mentioned in Step 1 are respectively: (1) Additive description form: u = u * + F + (4) where \(u\) is the control input of the AUV actuator, and \(u\) * is the desired input value; \(F\) + represents the fault vector in \(u\), \(f\) + is the additive fault description factor; (2) Multiplicative description form: u = F × u * (5) where F × is a diagonal matrix representing the fault amplitude, defined as When f × ≠ 1, the control input carries fault information, and in this case, f × is called the multiplicative fault description factor.

3. The AUV actuator fault diagnosis method based on fault factors and multiple observers according to claim 2, characterized in that The specific content and method steps for modeling faults in the state equation are as follows: (1) Power propulsion actuator fault: The propeller is the actuator in the AUV that directly acts on the fluid to generate forward thrust. Introduce the relationship between the propeller output thrust and its rotational speed as: T = K T n|n| + K v nV a (6) where T is the thrust, n is the propeller speed, and V a is the motor propulsion speed; the variables related to the propeller failure include: the thrust coefficient K T and the motor propulsion coefficient K v ; (2) Attitude control actuator fault: The attitude control system of the AUV includes the rudder surface and the steering gear. The rudder surface fault is a small-angle bending deformation at the edge. The deformed rudder surface acts on the fluid to generate a force parallel to the body coordinate system, and its additive fault generated by the fluid action is: wherein to are additive faults for respective degrees of freedom; and F * are the fluid forces during fault and non-fault respectively, In the attitude control system of the AUV, the steering gear is used to provide the torque for the rudder surface deflection. The given torque formula for the steering gear is: M = DK M n|n| + DK v nV a (8) where M is the torque; D is the propeller diameter; the variables related to the steering gear failure include: the torque coefficient K M and the motor propulsion coefficient K v , Introduce the multiplicative fault description factor f for the force and moment formulas of the propeller and the rudder servo * and the additive fault description factor f + , and we can get: 。 4. A method for fault diagnosis of an AUV actuator based on fault factors and multiple observers according to claim 1, characterized in that The specific content and method steps of Step 2 include: Step 2.1 Using formula (1), define and 1 < i < j < n, and define the remaining fault vectors in G as a new matrix F b , and remove it from G, and form a new distribution matrix with the interference D, which only contains the fault matrix F a The piecewise fault function G of is called G1; Step 2.2 Rewrite formula (1) as: wherein and are a new interference matrix and vector respectively, Part of the fault vectors in the AUV actuator fault vector are classified as a new disturbance vector The goal of such a design is to make the residuals generated by the observer free from interference and only affected by the new piecewise fault function G1. Since the new disturbance vector contains the interference D and part of the fault F b , the residuals are insensitive to the interference and the actuator fault carrying F b and sensitive to the actuator fault G1, thus designing a fault isolation observer with the same number of types of actuator faults 5. A fault diagnosis method for an AUV actuator based on fault factors and multiple observers according to claim 1, characterized in that The specific content and method steps of Step 3 include: Step 3.1 Expand other system variables outside the piecewise fault function G1 into a new state variable x2, obtaining: Thus, the form of the designed extended state observer is: where \(i = 1, 2, \ldots, n\) are the observers corresponding to the \(i\)-th fault type; \(z\) 1,i , \(z\) 2,i are the estimates of the state variables \(x\) 1,i , \(x\) 2,i respectively, and \(l\) 1,i , \(l\) 2,i is the gain vector of the extended state observer; Step 3.2 Calculate the gain of the extended state observer. The gain can usually be parameterized as: [l 1,i l 2,i = [β1ω0 β2ω0 2 (13) where β1, β2 are chosen parameters such that the characteristic polynomial β2s + β1 is Hurwitz. Let s (n+1) +β1s n +…+β n s+β n+1 =(s + 1) (n+1) (14) It should ensure that the roots of the characteristic polynomial are in the left half of the complex plane. Therefore, the parameters in Equation (6) are selected as follows: [l 1,i l 2,i = [2ω0 ω0 2 (15) where ω0 is the bandwidth of the observer, and this bandwidth needs to be continuously changed in practice to ensure that the ESO can correctly estimate the state variables.

6. A fault diagnosis method for the actuator of an AUV based on fault factors and multiple observers according to claim 1, characterized in that The specific content and method steps of Step 4 include: Step 4.1 Analyze the force on the faulty deformation part of the control surface. Consider the control surface in the actuator as a planar rectangle ABCD, and let the control surface angle be δ X , and the area is the part of the control surface where the deformation fault occurs. The deformation fault is represented as triangle ABC rotating downward by an angle along line AB, which is called the deformation angle δ f , and the angle generated on the plane is the fault plane angle r e , Subtract the axial fluid force when there is no fault from the axial fluid force under the fault to obtain the fault force in each axis. Based on this, calculate the fault force and moment generated by the deformed control surface ABC under the action of the fluid: where x e is the vertical distance from the edge of the transverse rudder surface to the z-axis; P is the fluid pressure; l is the length of the deformed part of the rudder surface, Step 4.2 further analyzes Step 4.

1. When different types of deformation faults occur in the control surface, there are cases where the resulting fault effects are opposite in sign. Analyze the signs of the fault effects caused by all control surfaces under deformation as follows: The left horizontal rudder surface deforms upward: Left horizontal rudder face deforms downward: The right rudder surface deforms upward: The right rudder surface deforms downward: The upper rudder surface deforms to the left: The upper rudder surface deforms to the right: The lower rudder surface deforms to the left: The lower rudder surface deforms to the right: By arranging and combining the four fault forces and torques, 8 basic deformation faults are obtained. Considering that only one control surface has a fault at the same time, and there are differences in the signs of the output effects of the deformation faults of different control surfaces, the faulty control surface can be identified by combining the fault estimation results.

Images