Method for fault detection and diagnosis of a rotary machine

CROSS REFERENCE TO RELATED APPLICATIONS

Applicants claim priority under 35 U.S.C. §119 of European Application No. 07017478.4 filed on Sep. 6, 2007.

BACKGROUND OF THE INVENTION

The invention relates to a method for fault detection and diagnosis of a rotary machine, in particular a balancing machine, wherein a rotor having an imbalance is rotated and excites a vibration in the rotary machine due to the imbalance-caused force, and wherein the rotational speed of the rotor and the vibrations are measured in order to obtain input data quantitative for the rotational speed and the vibrations.

SUMMARY OF THE INVENTION

A method for fault detection and diagnosis of a rotary machine, in particular a balancing machine is provided. A rotor having an imbalance is rotated and excites a vibration in the rotary machine due to the imbalance-caused force. The rotational speed of the rotor and the vibrations are measured in order to obtain input data quantitative for the rotational speed and the vibrations.

In accordance with the method, it is assumed that a process of dynamic behavior of the rotary machine can be modeled by a linear system in the un-faulty state. An over-determined set of linear equations is formed, which contains input and output data of the process and unknown states of the assumed linear system. The number of states needed to accurately model the process is extracted by using mathematical operations such as orthogonal or oblique projections to form a matrix of which the rank equals the assumed linear system. Singular values are computed by using Singular Value Decomposition for obtaining an approximate indication for the order of the assumed linear state.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows a rotary system model for normal condition;

FIG. 2 shows the fault state of sliding friction in sensor connection;

FIG. 3 shows the fault state of sliding friction between bearing support socket and ground; and

FIG. 4 shows singular values for normal and faulty states.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
1 Introduction

Fault detection and diagnosis is increasingly important for modern rotary machines. Currently, mostly limit checking and periodic maintenance cycles are used to detect faults. Sometimes signal-based fault detection is applied. However, these methods mainly rely on the knowledge of experts. This situation can be improved by model-based fault detection. As more information (proper excitation, process model, several measurements) is used, more accurate fault detection can be performed. Standard rotary systems behave basically linear. In case of specific faults, e.g. sliding friction, loose bonds and motion blocks, linear relations no longer hold. An indication for the nonlinearity can be used to detect these faults. Taking into account the noisy environment the method presented will use subspace approaches to estimate singular values. These singular values can be used to detect these faults.

2 Modeling

In order to design a model-based fault detection and diagnosis system, the dynamic behavior of the rotor system needs to be modeled. In a first step, a general model with two degrees of freedom for stiff rotors is given. For lower rotary speeds a simplified model can be applied.

It is assumed that the rotor is not fully balanced, so that an imbalance force F_uand torque M_uexist. The rotor is situated on two independent bearing supports. Their movement speeds in horizontal plane are denoted by {dot over (x)}₁,{dot over (x)}₂. Plunger coil sensors are used to measure these speeds, resulting in measurement values {dot over (s)}₁,{dot over (s)}₂(see also FIG. 1).

2.1 Model for Stiff Rotors

Rotor and bearings are assumed to be stiff, the ground connection of the two bearing supports is modeled by two spring-damper systems. It is assumed furthermore that the sensor force feedback on the rotor movement can be neglected.

m_r>>m_s1m_r>>m_s2

c₁>>c_s1c₂>>c_s2

As long as the system stays in normal condition, it can be described by a linear state space system.

{dot over (x)}
_m(t)=A_mx_m(t)+B_mu_m(t) (1)

y
_m(t)=C_mx_m(t)

Applying Newton's law of motion for a rotary mass it follows

$\begin{matrix} u_{m} (t) = {(F_{u} (t) M_{u} (t))}^{T} & (2) \\ x_{m} (t) = {({\dot{x}}_{1} (t) x_{1} (t) {\dot{x}}_{2} (t) x_{2} (t))}^{T} & (3) \\ y_{m} (t) = {({\dot{s}}_{1} (t) {\dot{s}}_{2} (t))}^{T} & (4) \\ A_{m} = \frac{1}{m_{r} θ_{r}} (\begin{matrix} (- θ_{r} d_{1} - m_{r} d_{1} l_{1}^{2}) & (- m_{r} c_{1} l_{1}^{2} - θ_{r} d_{1}) \\ 0 & m_{r} θ_{r} \\ (l_{2} m_{r} l_{1} d_{1} - θ_{r} d_{1}) & (l_{2} m_{r} l_{1} c_{1} - θ_{r} c_{1}) \\ 0 & 0 \\ (- θ_{r} d_{2} + m_{r} l_{2} d_{2} l_{1}) & (- θ_{r} c_{2} + m_{r} l_{2} c_{2} l_{1}) \\ 0 & 0 \\ (- l_{2}^{2} m_{r} d_{2} - θ_{r} d_{2}) & (- θ_{r} c_{2} - l_{2}^{2} m_{r} c_{2}) \\ 0 & m_{r} θ_{r} \end{matrix}) & (5) \\ B_{m} = (\begin{matrix} 1 / m_{r} & - l_{1} / θ_{r} \\ 0 & 0 \\ 1 / m_{r} & l_{2} / θ_{r} \\ 0 & 0 \end{matrix}) & (6) \\ C_{m} = (\begin{matrix} 1 & 0 & 0 & 0 \\ 0 & 0 & 1 & 0 \end{matrix}) & (7) \end{matrix}$

However, this detailed model is not needed for the regarded rotary system. As it is operated in sub-critical region, certain simplifications can be made.

2.2 Reduced Model

Under the assumption that the machine is driven with sub-critical rotary speed, i.e. (ω_r<<ω_crit) where ω_ris the actual rotary speed and ω_critis the lower of the two critical speeds according to system (1) it can be assumed that

m₁{umlaut over (x)}₁<<d₁{dot over (x)}₁<<c₁x₁

m₂{umlaut over (x)}₂<<d₂{dot over (x)}₂<<c₂x₂

With these simplifications the model reduces to a model of order two:

$\begin{matrix} {\dot{x}}_{r} (t) = A_{r} x_{r} (t) + B_{r} u_{r} (t) & (8) \\ y_{r} (t) = C_{r} x_{r} (t) & (9) \\ u_{r} (t) = {({\dot{s}}_{1} (t) {\dot{s}}_{2} (t))}^{T} & (10) \\ x_{r} (t) = {(x_{1} (t) x_{2} (t))}^{T} & (11) \\ y_{r} (t) = {(F_{u} (t) M_{u} (t))}^{T} & (12) \\ A_{r} = (\begin{matrix} 0 & 0 \\ 0 & 0 \end{matrix}) & (13) \\ B_{r} = (\begin{matrix} 1 & 0 \\ 0 & 1 \end{matrix}) & (14) \\ C_{r} = (\begin{matrix} c_{1} & c_{2} \\ - c_{1} l_{1} & c_{2} l_{2} \end{matrix}) & (15) \end{matrix}$

The system is observable and controllable, the poles are on the stability limit. Inputs and outputs are exchanged in order to match the state space structure. Discretization with small sampling time T₀leads to

{dot over (x)}
_d(k+1)=A_dx_d(k)+B_du_d(k) (16)

y
_d(k)=C_dx_d(k) (17)

$\begin{matrix} u_{d} (k) = {({\dot{s}}_{1} (k) {\dot{s}}_{2} (k))}^{T} & (18) \\ x_{d} (k) = {(x_{1} (k) x_{2} (k))}^{T} & (19) \\ y_{d} (k) = {(F_{u} (k) M_{u} (k))}^{T} & (20) \\ A_{d} = (\begin{matrix} 1 & 0 \\ 0 & 1 \end{matrix}) & (21) \\ B_{d} = (\begin{matrix} T_{0} & 0 \\ 0 & T_{0} \end{matrix}) & (22) \\ C_{d} = (\begin{matrix} c_{1} & c_{2} \\ - c_{1} l_{1} & c_{2} l_{2} \end{matrix}) & (23) \end{matrix}$

This model will be used as basis for fault detection.

2.3 Nonlinear Fault States

The presented method is used to detect two fault states where sliding friction occurs.

2.3.1 Sliding Friction in Sensor Connection

If the sensor is not connected properly to the left moving rotor support, the force is propagated via sliding friction. The propagated force F_Rs=f({dot over (x)}₁−{dot over (s)}₁) is modeled as Coulomb dry friction. The sensor dynamics are described by the frequency response G_s(w).

$\begin{matrix} F_{Rs} = F_{cs} sign ({\dot{x}}_{1} - {\dot{s}}_{1}) & (24) \\ G_{s} (w) = \frac{s}{m_{s} s^{2} + c_{s}} & (25) \end{matrix}$

The sensor force feedback on the rotor movement is neglected. Thus, the relation is nonlinear in the outputs equation only.

$\begin{matrix} {\dot{x}}_{p} (t) = A_{p} x_{p} (t) + B_{p} u_{p} (t) y_{p} (t) = (\begin{matrix} f ({\dot{x}}_{1} (t), G_{s} (w), F_{cs}) \\ {\dot{x}}_{2} (t) \end{matrix}) & (26) \\ u_{p} (t) = {(F_{u} (t) M_{u} (t))}^{T} & (27) \\ x_{p} (t) = {({\dot{x}}_{1} (t) x_{1} (t) {\dot{x}}_{2} (t) x_{2} (t))}^{T} & (28) \\ y_{p} (t) = {({\dot{s}}_{1} (t) {\dot{s}}_{2} (t))}^{T} & (29) \\ A_{p} = A_{m} & (30) \\ B_{p} = B_{m} & (31) \end{matrix}$

2.3.2 Sliding Friction in Ground Connection

If the left rotor support is not properly connected to the ground, the bearing support socket (mass m_g) may move on the ground. FIG. (3) shows the dynamic behavior of this fault state. It is assumed that dry Coulomb friction persists between rotor support socket and the ground.

The dynamics may be described by a linear system with nonlinear feedback according to FIG. (3).

{dot over (x)}
₁(t)=A₁x₁(t)+B₁u₁(t)

y
₁(t)=C₁x₁(t) (32)

with

$\begin{matrix} u_{1} (t) = {(F_{u} (t) M_{u} (t) F_{R 1} (t))}^{T} & (33) \\ x_{1} (t) = {({\dot{x}}_{1} (t) x_{1} (t) {\dot{x}}_{2} (t) x_{2} (t) {\dot{x}}_{g} (t) x_{g} (t))}^{T} & (34) \\ y_{1} (t) = {({\dot{s}}_{1} (t) {\dot{s}}_{2} (t))}^{T} & (35) \\ A_{1} = (\begin{matrix} \frac{(m_{r} d_{1} l_{1}^{2} + ⊖_{r} d_{1})}{m_{r} ⊖_{r}} & \frac{(m_{r} c_{1} l_{1}^{2} + ⊖_{r} c_{1})}{m_{r} ⊖_{r}} \\ A_{m} & 0 & 0 \\ \frac{(- l_{2} m_{r} l_{1} d_{1} + ⊖_{r} d_{1})}{m_{r} ⊖_{r}} & \frac{(- l_{2} m_{r} l_{1} c_{1} + ⊖_{r} c_{1})}{m_{r} ⊖_{r}} \\ 0 & 0 \\ \frac{c_{1}}{m_{g}} & \frac{d_{1}}{m_{g}} & 0 & 0 & \frac{- c_{1}}{m_{g}} & \frac{- d_{1}}{m_{g}} \\ 0 & 0 & 0 & 0 & 1 & 0 \end{matrix}) & (36) \\ B_{1} = (\begin{matrix} 0 \\ B_{m} & 0 \\ 0 \\ 0 \\ 0 & 0 & 1 / m_{g} \\ 0 & 0 & 0 \end{matrix}) & (37) \\ C_{1} = (C_{m} \begin{matrix} 0 & 0 \\ 0 & 0 \end{matrix}) & (38) \\ F_{Ri} = F_{d} sign ({\dot{x}}_{g} (t)) & (39) \end{matrix}$

2.3.3 Summary

The two described fault states introduce nonlinear behavior into the state space relation, either in the output equation or in the system equation. An approximation of this behavior with the reduced model according to section 2.2 turns out to be inaccurate.

2.4 Imbalance Force and Torque

It is assumed that the rotor is not fully balanced. The remaining imbalances cause an imbalance force F_uand torque M_u. It is assumed that the rotor speed ω_rand rotor roll angle φ_rare known and the imbalance amplitudes A_u1, A_u2and angles φ_u1, φ_u2are measured or known. The imbalance force and torque can be modeled as

$\begin{matrix} F_{u} (t) = A_{u 1} ω_{r}^{2} (t) \cos (ϕ_{r} (t) + ϕ_{u 1}) & (40) \\ + A_{u 2} ω_{r}^{2} (t) \cos (ϕ_{r} (t) + ϕ_{u 2}) & (41) \\ M_{u} (t) = l_{1} A_{u 1} ω_{r}^{2} (t) \cos (ϕ_{r} (t) + ϕ_{u 1}) & (42) \\ - l_{2} A_{u 2} ω_{r}^{2} (t) \cos (ϕ_{r} (t) + ϕ_{u 2}) & (43) \end{matrix}$

3 Estimation of the Degree of Linearity

As an approximate indication for the degree of linearity, common subspace-based methods are well-suited. The mathematical approach that is used can be described as follows:

- It is assumed that the process can be modeled by a linear state space system in the un-faulty state.
- Given the input and output data of the process, an over-determined set of linear equations is constructed. The set of equations contains input data, output data and unknown states of the state space system.
- The number of states needed to accurately model the process is extracted by mathematic operations such as orthogonal or oblique projections. The number of observable and controllable states equals the system order.
- The problem of determining what number of states is needed to model the system is transformed to a matrix rank determination. This determination is performed approximately by computing the singular values via Singular Value Decomposition (SVD).

The computed singular values give an approximate indication for the order of the presumed state space system. If the given process strongly obeys the reduced model equations according to section (2.2), the method indicates a process of order two. If nonlinear behavior resides and the linear model does not fit, the indication becomes indistinct.

3.1 Linearity Indication

This subsection briefly describes the computation of the features for linearity indication. The used algorithm is partially known as MOESP (Multivariable Output-Error State Space).

3.1.1 State Space System

The input/output relation is assumed to match a linear state space relation according to equation (44). N samples of inputs and outputs are available.

x(k+1)=Ax(k)+Bu(k)+Bn(k) (44)

y(k)=Cx(k)+Cm(k) (45)

The system is observable and controllable of order n. m(k) and n(k) represent white noise sequences. Matrices A, B, C, D and the states x(k) can be transformed by a regular transformation to

Ā=T⁻¹AT

B=T⁻¹B

C=CT

x
(k)=T⁻¹x(k) (46)

3.1.2 Alignment in Block Hankel Matrices

The measured data is aligned in block Hankel Matrices

$\begin{matrix} U_{p} = (\begin{matrix} u (0) & u (1) & \dots & u (j - ) \\ u (1) & u (2) & \dots & u (j) \\ ⋮ & ⋮ & ⋱ & ⋮ \\ u ( - 1) & u () & \dots & u ( + j - 2) \end{matrix}) & (47) \\ U_{f} = (\begin{matrix} u () & u ( + 1) & \dots & u ( + j - 1) \\ u ( + 1) & u ( + 2) & \dots & u ( + j) \\ ⋮ & ⋮ & ⋱ & ⋮ \\ u (2  - 1) & u (2 ) & \dots & u (2  + j - 2) \end{matrix}) & (48) \\ Y_{p} = (\begin{matrix} y (0) & y (1) & \dots & y (j - ) \\ y (1) & y (2) & \dots & y (j) \\ ⋮ & ⋮ & ⋱ & ⋮ \\ y ( - 1) & y () & \dots & y ( + j - 2) \end{matrix}) & (49) \\ Y_{f} = (\begin{matrix} y () & y ( + 1) & \dots & y ( + j - 1) \\ y ( + 1) & y ( + 2) & \dots & y ( + j) \\ ⋮ & ⋮ & ⋱ & ⋮ \\ y (2  - 1) & y (2 ) & \dots & y (2  + j - 2) \end{matrix}) & (50) \end{matrix}$

with

u(k)=({dot over (s)}_i(k){dot over (s)}₂(k))^T

y(k)=(F_u(k)M_u(k))^T

N number of measurements 2i maximum order that can be indicated. User-chosen.

j=N−2i+1 if all measurements are used.

The matrices contain all available data and therefore all available information. A set of linear equations is formed which contains these Hankel Matrices and the state Vectors x(k). To explain the procedure, the noise influence is set to zero at this stage.

Y
_p=Γ_iX_p+H_iU_p

Y
_f=Γ_iX_f+H_iU_f (51)

X
_f
=A
ⁱ
X
_p+Δ_iU_p

To develop this set of equations, following matrices are used:

$\begin{matrix} Γ_{i} = {(C^{T} {(CA)}^{T} {({CA}^{2})}^{T} \dots {({CA}^{i - 1})}^{T})}^{T} & (52) \\ Δ_{i} (A^{i - 1} B A^{i - 2} B \dots AB B) & (53) \\ H_{i} = (\begin{matrix} D & 0 & 0 & \dots & 0 \\ CB & D & 0 & \dots & 0 \\ CAB & CB & D & \dots & 0 \\ ⋮ & ⋮ & ⋮ & \dots & \dots \\ {CA}^{i - 2} B & {CA}^{i - 3} B & {CA}^{i - 4} B & \dots & D \end{matrix}) & (54) \end{matrix}$

The state matrices X_pand X_fare defined analogously to the input/output Hankel Matrices:

X
_p=(x(0)x(1) . . . x(j−1)) (55)

X
_f=(x(i)x(i+1) . . . x(i+j−1)) (56)

The set of equations (51) can easily be verified by direct insertion. By removing the unknown states from this set of equations the solution for Y_fyields:

$\begin{matrix} Y_{f} = (H_{f} Γ_{i} (Δ_{i} - A^{i} Γ_{i}^{†} H_{i}) Γ_{i} A^{i} Γ_{i}^{†}) (\begin{matrix} U_{f} \\ U_{p} \\ Y_{p} \end{matrix}) & (57) \end{matrix}$

The notation † stands for the Moore-Penrose-Pseudoinverse. Equation (57) yields direct information on the linearity indication features. If the model is purely linear, the row space of Y_fcan be fully described by the row spaces

$(\begin{matrix} U_{f} \\ U_{p} \\ Y_{p} \end{matrix}) .$

The row space of a matrix is the space spanned by its row vectors. If a matrix J is of full rank, its row space equals the row space of J_zif J=J_sJ_z.

3.1.3 Feature Extraction

For order extraction many different methods are known. The most common are N4SID, MOESP and CVA. As the underlying system is on the stability limit, the algorithm with the most direct order computation, MOESP, is used. Tests with real data as described in the following have approved this choice. MOESP uses a direct RQ decomposition of aligned Block Hankel Matrices:

$\begin{matrix} (\begin{matrix} U_{f} \\ U_{p} \\ Y_{p} \end{matrix}) = (\begin{matrix} R_{U_{f}} & 0 & 0 & 0 \\ 0 & R_{U_{p}} & 0 & 0 \\ 0 & 0 & R_{Y_{p}} & 0 \\ H_{f} R_{U_{f}} & Γ_{i} (Δ_{i} - A^{i} Γ_{i}^{†} H_{i}) R_{U_{p}} & Γ_{i} A^{i} Γ_{i}^{†} R_{Y_{p}} & 0 \end{matrix}) (\begin{matrix} Q_{U_{r}} \\ Q_{U_{p}} \\ Q_{Y_{p}} \\ Q_{Y_{f}} \end{matrix}) with U_{f} = R_{U_{f}} Q_{U_{f}}, U_{p} = R_{U_{p}} Q_{U_{p}}, Y_{p} = R_{Y_{p}} Q_{Y_{p}} & (58) \end{matrix}$

From the first part, a matrix β_r, with rank=system order is extracted.

$\begin{matrix} \begin{matrix} β_{r} = (Γ_{i} (Δ_{i} - A^{i} Γ_{i}^{t} H_{i}) R_{U_{p}} Γ_{i} A^{i} Γ_{i}^{†} R_{Y_{p}}) \\ = (Γ_{i} Γ_{i}) (\begin{matrix} (Δ_{i} - A^{i} Γ_{i}^{t} H_{i}) & 0 \\ 0 & A^{i} Γ_{i}^{t} R_{Y_{p}} \end{matrix}) \end{matrix} & \begin{matrix} (59) \\ (60) \end{matrix} \end{matrix}$

The rank of β_requals the number of linear independent vectors in its column space, which means that

rank(β_r)=rank(Γ_i)=n (61)

The column space of a matrix is the space spanned by its column vectors. If a matrix J is of full rank, its column space equals the column space of J_zif J=J_sJ_z.

A proper method can extract the ‘true’ order of the underlying system. In the case of a faulty, nonlinear behavior, the linearity indication differs from the described model of order two. To extract the matrix rank in an approximate way, Singular Value Decomposition (SVD) is used. The SVD of β_ryields 3 matrices U₁, S₁, V₁

$\begin{matrix} \begin{matrix} β_{r} = U_{1} S_{1} V_{1}^{T} \\ = U_{1} (\begin{matrix} σ_{1} & 0 & \dots & 0 \\ 0 & σ_{2} & \dots & 0 \\ ⋮ & ⋮ & ⋱ & ⋮ \\ 0 & 0 & \dots & σ_{i} \end{matrix}) V_{1}^{T} \end{matrix} & \begin{matrix} (62) \\ (63) \end{matrix} \end{matrix}$

U, and V₁are orthogonal matrices. S₁is a diagonal matrix which contains the singular values σ_i. In case of a fault-free, not noisy system, n singular values are nonzero while all other singular values are zero.

σ₁> . . . >σ_n>0 (64)

σ_n+1= . . . =σ_2i=0 (65)

3.2 Noise Influence on Feature Extraction

Under the influence of noise (process noise as well as measurement noise), the SVD no longer yields clear order decisions. To represent noise influences, it will be assumed that the measurements y(k) are contaminated by white noise n(k)=(n₁(k)n₂(k))^T. The excitations u(k) contain noise m(k)=(m₁(k)m₂(k))^T. With

$N_{p} = (\begin{matrix} n (0) & n (1) & \dots & n (j - ) \\ n (1) & n (2) & \dots & n (j) \\ ⋮ & ⋮ & ⋱ & ⋮ \\ n ( - 1) & n () & \dots & n ( + j - 2) \end{matrix})$

$N_{f} = (\begin{matrix} n () & n ( + 1) & \dots & n ( + j - 1) \\ n ( + 1) & n ( + 2) & \dots & n ( + j) \\ ⋮ & ⋮ & ⋱ & ⋮ \\ n (2  - 1) & n (2 ) & \dots & n (2  + j - 2) \end{matrix})$

and subsequent formation of M_pand M_f, equation (51) is enhanced to

Y
_p=Γ_iX_p+H_iU_p+H_iN_p+M_p

Y
_f=Γ_iX_f+H_iU_f+H_iN_f+M_f (66)

X
_f
=A
ⁱ
X
_p+Δ_iU_p+Δ_iN_p

The solution for Y_fthen yields

$\begin{matrix} Y_{f} = (H_{f} Γ_{i} (Δ_{i} - A_{}^{i} Γ_{i}^{†} H_{i}) Γ_{i} A^{i} Γ_{i}^{†}) (\begin{matrix} U_{f} \\ U_{p} \\ Y_{p} \end{matrix}) & (67) \\ + (H_{i} I Γ_{i} (Δ_{i} - A_{}^{i} Γ_{i}^{†} H_{i}) - Γ_{i} A^{i} Γ_{i}^{†}) (\begin{matrix} N_{f} \\ M_{f} \\ N_{p} \\ M_{p} \end{matrix}) & (68) \end{matrix}$

and the SVD is performed on

$\begin{matrix} β_{i} = β_{r} + (H_{i} I Γ_{i} (Δ_{i} - A^{i} Γ_{i}^{t} H_{i}) - Γ_{i} A^{i} Γ_{i}^{†}) (\begin{matrix} N_{f} \\ M_{f} \\ N_{p} \\ M_{p} \end{matrix}) / U_{f} \\ = β_{r} + E_{n} \end{matrix} (\begin{matrix} U_{p} \\ Y_{p} \end{matrix})$

where

β_i=2i×2i matrix with rank that has to be estimated

β_r=2i×2i matrix with rank n (n=system order)

E
_N=2i×2i Matrix with full rank (noise representation)

The notation

$/ U_{f} (\begin{matrix} U_{p} \\ Y_{p} \end{matrix})$

stands for the orthogonal projection onto the row space of

$(\begin{matrix} U_{f} \\ U_{p} \\ Y_{p} \end{matrix}) .$

where only the part lying in the row space of

$(\begin{matrix} U_{p} \\ Y_{p} \end{matrix})$

is considered. In literature, this projection is referred to as ‘Oblique Projection’.

It is shown in [VOdM96] that E_n→0 for N→∞.

An infinite number of measurements is not achievable. However, if the Signal-to-Noise-Ratio (SNR) and the measurement number N is sufficiently high, the system impact on the Singular Values is larger than the noise influence (see [Lju99]) and E_n<<β_rholds and therefore rank (β_i)=rank (β_r). Normally, the influence of E_nis not fully negligible and the Singular Values computation yields

σ₁> . . . >σ_n>>σ_n+1, . . . , σ_2i (69)

The singular values representing the system structure dominate, all successive singular values represent the noise influence and are considerably lower.

4 Results

As test rig, a industrial production machine for rotors with mass m_r≈25 kg is used. It is equipped with 2 standard plunger coil sensors. The fundamentation can be adjusted by common clamps. The machine is driven in sub-critical rotary speed. The environment consists of normal industrial surrounding, e.g. other machines, noise etc.

Three different runs are analyzed:

- Un-faulty standard run
- Run with sliding friction in sensor connection
- Run with sliding friction in ground connection

The features described in the preceding section are computed for the 3 described states. All tests are done on the same machine with equal bearings and rotor. For each run, a timespan of 5 seconds is examined. The singular values are computed according to the preceding chapter. FIG. (4) gives an overview over the different features. The values are given in logarithmic scale, the standard deviation over all regarded runs is indicated.

The first two Singular Values, which represent the linear behavior, remain nearly equal. The third and fourth singular value refer to model inaccuracies and noise. In case of nonlinear behavior, their values are considerably higher than in the purely linear, un-faulty case. Principally, these values can be used as a feature for the appearance of nonlinear faults in linear systems.

The invention described above presents a subspace-based method to detect the occurrence of nonlinear fault states in linear systems. A rotor system was used as example. The fault-free state as well as two different fault cases have been modeled and tested on an industrial rotor balancing machine. It has been shown that the computed singular values are highly sensible to nonlinear fault states and are well-suited as features for the occurrence of the considered friction faults.

Symbols and Abbreviations

Symbol
Unit
Description

F_u
[N]
imbalance-caused force

M_u
[Nm]
imbalance-caused torque

{dot over (x)}₁
[m/sec]
movement speed of left rotor support

{dot over (x)}₂
[m/sec]
movement speed of right rotor support

{dot over (s)}₁, {dot over (s)}₂
[m/sec]
measurements by speed sensors

c₁, c₂
[N/m]
spring constants

d₁, d₂
[N/m]
damper constants

m_r
[kg]
rotor mass

Θ_r
[kgm²]
rotor moment of inertia

l₁, l₂
[m]
distance to the center of gravity

A_u1, A_u2
[l]
imbalance amplitude

φ_u1, φ_u2
[l]
imbalance amplitude

ω_r
[rad/sec]
rotation speed

φ_r
[rad]
rotor roll angle

n(k), m(k)

white noise sequences

FIG. 1: Rotary system model for normal condition. The rotor (mass m_r) movement in horizontal plane can be modeled with linear spring-damper systems (left drawing, birdview). The sensor plunger coils (masses m_s1,m_s2) are connected directly to the bearings. As example, the connection of plunger coil 1 is given in the right drawing (side view).

FIG. 2: Fault state of sliding friction in sensor connection

FIG. 3: Fault state of sliding friction between bearing support socket and ground.

FIG. 4: Singular values for normal and faulty states

Method for fault detection and diagnosis of a rotary machine

Information

Publication Number

Date Filed

Date Published

Inventors

CPC

US Classifications

International Classifications

Abstract

Description

Claims

Priority Claims (1)