阅读说明：本技术 基于刚度矢量空间投影极大化的大型高速回转装备测量与神经网络学习调控方法及其装置 (Large-scale high-speed rotation equipment measurement and neural network learning regulation and control method and device based on rigidity vector space projection maximization ) 是由 刘永猛 孙传智 谭久彬 于 2019-01-07 设计创作，主要内容包括：本发明提出了基于刚度矢量空间投影极大化的大型高速回转装备测量与神经网络学习调控方法及其装置,属于机械装配技术领域。所述方法利用包络滤波原理、二维点集S、最小二乘法和学习神经网络实现大型高速回转装备测量与调控；所述装置包括基座、气浮轴系、调心调倾工作台、精密力传感器、静平衡测量平台、左立柱,右立柱,左下横向测杆、左下伸缩式电感传感器、左上横向测杆、左上伸缩式电感传感器、右下横向测杆、右下杠杆式电感传感器、右上横向测杆和右上杠杆式电感传感器。所述方法和装置能够对大型高速回转装备进行有效的测量和精准的调控。(The invention provides a large-scale high-speed rotation equipment measurement and neural network learning regulation and control method and device based on rigidity vector space projection maximization, and belongs to the technical field of mechanical assembly. The method utilizes an envelope filtering principle, a two-dimensional point set S, a least square method and a learning neural network to realize measurement and regulation of large-scale high-speed rotation equipment; the device comprises a base, an air floatation shaft system, a centering and inclination adjusting workbench, a precise force sensor, a static balance measuring platform, a left upright post, a right upright post, a left lower transverse measuring rod, a left lower telescopic inductive sensor, a left upper transverse measuring rod, a left upper telescopic inductive sensor, a right lower transverse measuring rod, a right lower lever inductive sensor, a right upper transverse measuring rod and a right upper lever inductive sensor. The method and the device can be used for effectively measuring and accurately regulating and controlling large-scale high-speed rotation equipment.)

1. The large-scale high-speed rotation equipment measurement and neural network learning regulation and control method based on the maximization of the space projection of the stiffness vector is characterized by comprising the following steps of:

the method comprises the following steps of firstly, obtaining a morphological filter based on non-equal-interval sampling angles by utilizing an envelope filtering principle, and extracting a rotor circular contour from a functional angle; taking any point P from the two-dimensional point set S₁Will point P₁Is a starting point and said P₁Points at a distance less than 2 α constitute the subset S₁Wherein S is a two-dimensional space coordinate point set of the circular contour sampling points, and α is an alpha disc radius;

step two, in subset S₁Take an arbitrary point P₂Then there are two superp's with radius α₁And P₂The locus equation of the centers of the inner circle and the outer circle of the alpha disc at two points is as follows:

or

Wherein, P₀And P₀' are the centers of the two alpha disks respectively; and is provided with

ρ₀、ρ_0'、ρ₁、ρ₂、Are respectively a point P₀、P_0'、P₁、P₂The polar diameter and polar angle under polar coordinates;

x₀、y₀、x_0’、y_0’are respectively a point P₀、P_0’X and Y axis direction coordinates of (a);

step three, acquiring an alpha envelope boundary by using the trajectory equation in the step twoAnd sampling point polar coordinatesRelation, said alpha envelope boundaryAnd sampling point polar coordinatesThe relationship is expressed as:

wherein n is the number of round contour sampling points, rho_i、Are respectively a point P_iThe polar diameter and polar angle under polar coordinates; f is a non-equidistant morphological filter design rule based on the alphashape theory;

performing effectiveness processing on the circular contour acquired data through a non-equal interval filter to obtain effective circular contour data; then, fitting the rotor circular profile by using a least square method according to the effective circular profile data, and evaluating the offset of the rotor to obtain the offset of the single-stage rotor;

and step five, determining the accumulated offset of the kth-stage rotor after the n-stage rotor is assembled by using a multi-stage rotor vector stacking projection theory according to the offset of the single-stage rotor as follows:

wherein dx is_0-kCumulative offset dy of circle center of k-th-stage rotor measuring surface in X-axis direction after n-stage rotor assembly_0-kAccumulated offset S of circle center of n-th-stage rotor measuring surface in Y-axis direction after n-stage rotor assembly_xj-1For the reference plane of the rotor and stator of the j-1 st stage to rotate by theta around the X axis_xj-1A rotation matrix of angles; s_yj-1For the reference plane of the (j-1) th rotor and stator to rotate by theta around the Y axis_yj-1A rotation matrix of angles; p is a radical of_iAn ideal position vector of the circle center of the i-th-stage rotor radial measurement surface is obtained; dp_iA processing error vector of the circle center position of the i-th-stage rotor radial measurement surface is obtained; s_rj-1For the j-1 st rotor and stator to rotate by theta around the Z axis_rj-1A rotation matrix of angles; s_r1Is an identity matrix;

step six, determining an expression of the coaxiality of the assembled n-stage rotors, an expression of the sectional area S of the contact surfaces between the assembled rotors and an expression of the sectional inertia moment I of the assembled contact surfaces between the assembled rotors according to the definition of the ISO standard of the coaxiality, wherein the expressions are respectively as follows:

wherein R is the outer diameter of the contact surface, R is the inner diameter of the contact surface, and the bending rigidity is EI; eccentricity of the cylinderEccentricity angle d θ ═ arctan (dy)_0-n/dx_0-n) The tensile rigidity of the multistage rotor is ES, wherein E is the elastic modulus of the material;

seventhly, determining the nth-stage rotor unbalance amount caused by the offset of each stage of rotor, wherein the expression of the nth-stage rotor unbalance amount is as follows:

wherein, Ux_0-nThe unbalance amount of the n-th-stage rotor measuring surface in the X-axis direction after assembly is obtained; uy_0-nThe unbalance amount of the n-th-stage rotor measuring surface in the Y-axis direction after assembly is obtained; m is_0-nMass of the n-th-stage rotor after assembly; p is a radical of_iAn ideal position vector of the circle center of the i-th-stage rotor radial measurement surface is obtained; dp_iA processing error vector of the circle center position of the i-th-stage rotor radial measurement surface is obtained;

step eight, vector addition is carried out on the unbalance of the single-stage rotor and the unbalance introduced by the offset of the rotors at all stages in the assembling process, and the unbalance of any one-stage rotor after the multi-stage rotor is assembled is obtained; then, projecting the unbalance of each stage of rotor to two more front faces respectively, synthesizing the unbalance according to a dynamic balance formula, and establishing a prediction model of the unbalance of the multistage rotor;

step nine, establishing an intelligent learning neural network by using the prediction model obtained in the step eight and combining temperature and humidity assembly environment influence factors and simultaneously combining bolt tightening torque, tightening sequence and assembly process influence factors of the elastic modulus, height and contact surface radius of the material of the equipment; and finally, the measurement of the large-scale high-speed rotation equipment and the neural network learning regulation are completed by combining a prediction model and an intelligent learning neural network.

2. The method according to claim 1, wherein the neural network of step nine is a BP neural network prediction model with the rotor temperature, humidity, tightening torque, orientation error, positioning error, unbalance measurement error, cross section area, front position, bolt tightening sequence, material elastic modulus, rotor height and contact surface radius error factors of each stage as input quantities, and coaxiality, tensile stiffness, bending stiffness and unbalance quantities of the assembled rotor as output quantities, and with two hidden layers and 40 hidden layer nodes in each layer.

3. The method according to claim 1 or 2, wherein the neural network establishment process comprises neuron activation function selection, implicit layer network node number setting and implicit layer number setting.

4. The method of claim 3, wherein the neuron activation function adopts an activation function that selects a Tansig function as the hidden layer and a purelin function as the output layer activation function, and wherein the prototype of the Tansig function and the purelin function are as follows:

f(x)＝x。

5. the method of claim 2, wherein the implicit layer network node number setting and the implicit layer number setting are: setting the network prediction error to be 0.0001, and setting the number of hidden layer nodes to be 2 times that of input layer nodes; the number of network layers is 4; the number of the hidden layer nodes of the first layer is 30, and the number of the hidden layer nodes of the second layer is 30.

6. A large-scale high-speed rotation equipment measurement and neural network learning regulation and control device for realizing the method of claim 1 is characterized by comprising a base (1), an air floatation shaft system (2), a centering and inclination adjusting workbench (3), precise force sensors (4a, 4b and 4c), a static balance measurement platform (5), a left upright post (6), a right upright post (7), a left lower transverse measuring rod (8), a left lower telescopic inductive sensor (9), a left upper transverse measuring rod (10), a left upper telescopic inductive sensor (11), a right lower transverse measuring rod (12), a right lower lever inductive sensor (13), a right upper transverse measuring rod (14) and a right upper lever inductive sensor (15); the air floatation shaft system (2) is embedded in the central position of the base (1), the aligning and inclination adjusting workbench (3) is arranged in the central position of the air floatation shaft system (2), and the three precise force sensors (4a, 4b and 4c) are uniformly arranged on the aligning and inclination adjusting workbench (3); the static balance measuring platform (5) is arranged on three precision force sensors (4a, 4b and 4 c); the left upright post (6) and the right upright post (7) are symmetrically distributed on two sides of the air floatation shaft system (2) and are fixedly arranged on the base (1); a left upper transverse measuring rod (10) and a left lower transverse measuring rod (8) are sleeved on the left upright post (6) from top to bottom in a movable and adjustable manner, and a left upper telescopic inductive sensor (11) is fixedly connected with the left upper transverse measuring rod (10); the lower left telescopic inductive sensor (9) is fixedly connected with the lower left transverse measuring rod (8); a right upper transverse measuring rod (14) and a right lower transverse measuring rod (12) are sleeved on the right upright post (7) from top to bottom in a movable and adjustable mode in sequence, and a right upper lever type inductive sensor (15) is fixedly connected with the right upper transverse measuring rod (14); and the right lower lever type inductive sensor (13) is fixedly connected with the right lower transverse measuring rod (12).

Technical Field

The invention relates to a large-scale high-speed rotation equipment measurement and neural network learning regulation and control method and device based on rigidity vector space projection maximization, and belongs to the technical field of mechanical assembly.

Background

With the continuous improvement of the performance requirements of the aero-engine in China, the assembly quality of the aero-engine is required to be higher and higher. The dynamic performance of the aircraft engine is directly influenced by the rotor assembly quality, and the geometric coaxiality, the rigidity and the initial unbalance amount after the assembly are core parameters for testing the multistage rotor assembly quality. When the aeroengine is at the working speed, such as the working speed of a civil aeroengine can reach over 12000rpm, the initial unbalance amount after the multi-stage rotor is assembled and the unbalance response caused by the coaxiality error can be amplified, so that the engine vibrates, and the blades and the casing are abraded. The smaller the overall stiffness of the rotor, the more severe the vibrations, which can even lead to serious failure of the aircraft engine. According to a dynamic equation, the coaxiality and the rigidity of the rotor structure are improved, the unbalance amount after assembly is reduced, and the method has important significance for reducing the weight of the rotor, improving the dynamic response characteristic of the rotor and improving the precision of an engine. The three-target optimization for realizing the coaxiality, the rigidity and the unbalance of the rotor is established on the basis of accurate prediction of the three parameters. Therefore, in order to improve and promote the dynamic performance of the high-pressure combined rotor of the existing aircraft engine and meet the development requirements of a new generation of advanced aircraft engine, the prediction of three parameters of the coaxiality, the rigidity and the unbalance of the high-pressure combined rotor is necessary. The existing prediction method basically has the problems that an assembly guidance model cannot be theoretically provided, the calculation process is complicated, the design cost is overhigh, and the like.

Disclosure of Invention

The invention provides a method and a device for measuring and regulating a large-scale high-speed rotation device and learning a neural network based on maximization of a stiffness vector space projection, aiming at solving the problems that an assembly guidance model cannot be theoretically provided in the existing prediction method, the calculation process is complicated, and the design cost is overhigh, and specifically comprises the following steps:

a large-scale high-speed rotation equipment measurement and neural network learning regulation and control method based on space projection maximization of bending rigidity and tensile rigidity adopts the following technical scheme:

the method comprises the following steps:

step one, obtaining a morphological filter based on non-equidistant sampling angles by utilizing an envelope filtering principle, and carrying out alignment from a functional angleExtracting the rotor circle outline; taking any point P from the two-dimensional point set S₁Will point P₁Is a starting point and said P₁Points at a distance less than 2 α constitute the subset S₁Wherein S is a two-dimensional space coordinate point set of the circular contour sampling points, and α is an alpha disc radius;

step two, in subset S₁Take an arbitrary point P₂Then there are two superp's with radius α₁And P₂The locus equation of the centers of the inner circle and the outer circle of the alpha disc at two points is as follows:

or

Wherein, P₀And P₀' are the centers of the two alpha disks respectively; and is provided with

ρ₀、ρ_0'、ρ₁、ρ₂、Are respectively a point P₀、P_0'、P₁、P₂The polar diameter and polar angle under polar coordinates; x is the number of₀、y₀、x_0’、y_0’Are respectively a point P₀、P_0’X and Y axis direction coordinates of (a);

step three, acquiring an alpha envelope boundary by using the trajectory equation in the step twoAnd sampling point polar coordinatesRelation, said alpha envelope boundaryAnd sampling point polar coordinatesThe relationship is expressed as:

wherein n is the number of round contour sampling points, rho_i、Are respectively a point P_iThe polar diameter and polar angle under polar coordinates; f is a non-equidistant morphological filter design rule based on an alpha shape theory;

performing effectiveness processing on the circular contour acquired data through a non-equal interval filter to obtain effective circular contour data; then, fitting the rotor circular profile by using a least square method according to the effective circular profile data, and evaluating the offset of the rotor to obtain the offset of the single-stage rotor;

and step five, the multistage rotors are formed by sequentially assembling the single-stage rotors, and the accumulated offset of the kth-stage rotor after the n-stage rotors are assembled is determined to be expressed as:

wherein dx is_0-kCumulative offset dy of circle center of k-th-stage rotor measuring surface in X-axis direction after n-stage rotor assembly_0-kAccumulated offset S of circle center of n-th-stage rotor measuring surface in Y-axis direction after n-stage rotor assembly_xj-1For the reference plane of the rotor and stator of the j-1 st stage to rotate by theta around the X axis_xj-1A rotation matrix of angles; s_yj-1For the reference plane of the (j-1) th rotor and stator to rotate by theta around the Y axis_yj-1A rotation matrix of angles; p is a radical of_iAn ideal position vector of the circle center of the i-th-stage rotor radial measurement surface is obtained; dp_iA processing error vector of the circle center position of the i-th-stage rotor radial measurement surface is obtained; s_rj-1For the j-1 st rotor and stator to rotate by theta around the Z axis_rj-1A rotation matrix of angles; s_r1Is an identity matrix;

step six, determining an expression of the coaxiality of the assembled n-stage rotors, an expression of the sectional area S of the contact surfaces between the assembled rotors and an expression of the sectional inertia moment I of the assembled contact surfaces between the assembled rotors according to the definition of the ISO standard of the coaxiality, wherein the expressions are respectively as follows:

wherein R is the outer diameter of the contact surface, R is the inner diameter of the contact surface, and the bending rigidity is EI; eccentricity of the cylinderEccentricity angle d θ ═ arctan (dy)_0-n/dx_0-n) The multistage rotor tensile stiffness is ES, whereinE is the elastic modulus of the material;

step seven, in the assembly of the multi-stage rotor, the offset of the single-stage rotor can be transmitted and accumulated, and the unbalance after the assembly of the multi-stage rotor is influenced: determining the nth-stage rotor unbalance amount caused by each-stage rotor offset, wherein the expression of the nth-stage rotor unbalance amount is as follows:

wherein, Ux_0-nThe unbalance amount of the n-th-stage rotor measuring surface in the X-axis direction after assembly is obtained; uy_0-nThe unbalance amount of the n-th-stage rotor measuring surface in the Y-axis direction after assembly is obtained; m is_0-nMass of the n-th-stage rotor after assembly; p is a radical of_iAn ideal position vector of the circle center of the i-th-stage rotor radial measurement surface is obtained; dp_iA processing error vector of the circle center position of the i-th-stage rotor radial measurement surface is obtained;

step eight, vector addition is carried out on the unbalance of the single-stage rotor and the unbalance introduced by the offset of the rotors at all stages in the assembling process, and the unbalance of any one-stage rotor after the multi-stage rotor is assembled is obtained; then, projecting the unbalance of each stage of rotor to two more front faces respectively, synthesizing the unbalance according to a dynamic balance formula, and establishing a prediction model of the unbalance of the multistage rotor;

step nine, establishing an intelligent learning neural network by using the prediction model obtained in the step eight and combining temperature and humidity assembly environment influence factors and simultaneously combining bolt tightening torque, tightening sequence and assembly process influence factors of the elastic modulus, height and contact surface radius of the material of the equipment; and finally, the measurement of the large-scale high-speed rotation equipment and the neural network learning regulation are completed by combining a prediction model and an intelligent learning neural network.

And furthermore, the neural network in the ninth step is a BP neural network prediction model with the temperature, humidity, tightening torque, orientation error, positioning error, unbalance measurement error, sectional area, front position, bolt tightening sequence, material elastic modulus, rotor height and contact surface radius error factors of each stage of rotor as input quantities, coaxiality, tensile rigidity, bending rigidity and unbalance after rotor assembly as output quantities, and each layer is provided with two hidden layers and 40 hidden layer nodes.

Further, the neural network establishing process comprises neuron activation function selection, hidden layer network node number setting and hidden layer number setting.

Further, the neuron activation function adopts an activation function which selects a Tansig function as a hidden layer and a purelin function as an output layer activation function, wherein prototypes of the Tansig function and the purelin function are respectively as follows:

f(x)＝x

further, the number of the hidden layer network nodes and the number of the hidden layers are set as follows: setting the network prediction error to be 0.0001, and setting the number of hidden layer nodes to be 2 times that of input layer nodes; the number of network layers is 4; the number of the hidden layer nodes of the first layer is 30, and the number of the hidden layer nodes of the second layer is 30.

A large-scale high-speed rotation equipment measurement and god network learning regulation and control device for realizing the method adopts the following technical scheme:

the device comprises a base 1, an air floatation shaft system 2, a centering and inclination adjusting workbench 3, precision force sensors 4a, 4b and 4c, a static balance measuring platform 5, a left upright post 6, a right upright post 7, a left lower transverse measuring rod 8, a left lower telescopic inductive sensor 9, a left upper transverse measuring rod 10, a left upper telescopic inductive sensor 11, a right lower transverse measuring rod 12, a right lower lever type inductive sensor 13, a right upper transverse measuring rod 14 and a right upper lever type inductive sensor 15; an air floatation shaft system 2 is nested on the central position of a base 1, an aligning and inclination adjusting workbench 3 is arranged on the central position of the air floatation shaft system 2, and three precise force sensors 4a, 4b and 4c are uniformly arranged on the aligning and inclination adjusting workbench 3; the static balance measuring platform 5 is arranged on three precision force sensors 4a, 4b and 4 c; the left upright post 6 and the right upright post 7 are symmetrically distributed on two sides of the air floatation shaft system 2 and are fixedly arranged on the base 1; a left upper transverse measuring rod 10 and a left lower transverse measuring rod 8 are sleeved on the left upright post 6 from top to bottom in a movable and adjustable manner, and a left upper telescopic inductive sensor 11 is fixedly connected with the left upper transverse measuring rod 10; the lower left telescopic inductive sensor 9 is fixedly connected with the lower left transverse measuring rod 8; an upper right transverse measuring rod 14 and a lower right transverse measuring rod 12 are sleeved on the right upright post 7 from top to bottom in a movable and adjustable manner, and an upper right lever type inductive sensor 15 is fixedly connected with the upper right transverse measuring rod 14; and the right lower lever type inductive sensor 13 is fixedly connected with the right lower transverse measuring rod 12.

The invention has the beneficial effects that:

the measurement and regulation method of the large-scale high-speed rotation equipment based on the spatial projection maximization of the bending stiffness and the tensile stiffness analyzes the sampling angle distribution characteristic and the measurement error of the single-stage rotor circular profile measurement, and performs functional filtering on the acquired circular profile data through the unequal interval morphological filter; obtaining the offset of the contact surface between the rotors at all levels according to the transmission relation of the rotors at multiple levels, and calculating the coaxiality prediction result according to a coaxiality formula; calculating the cross-sectional area and the cross-sectional moment of inertia of the contact surface, and obtaining a rigidity prediction result according to a tensile rigidity and bending rigidity formula; obtaining a prediction result of the unbalance amount of the rotor according to the error transfer relation of the rotor; and finally, predicting the performance of the multistage rotor of the aero-engine based on the unequal interval filtering technology. The measuring and regulating device for the large-sized high-speed rotating equipment based on the spatial projection maximization of the bending rigidity and the tensile rigidity, provided by the invention, can be used for effectively measuring and accurately regulating and controlling the large-sized high-speed rotating equipment.

Drawings

FIG. 1 is a schematic diagram of the spatial distribution of a two-dimensional point set S according to the present invention, where O is the center of a sampling contour, S is a two-dimensional coordinate point set of sampling points of a circular contour, and P is₁Is any point in a two-dimensional point set S, S₁To a point P₁Set of points starting at a distance of less than 2 α from it, α being the alpha disk radius, P₂As a set of points S₁At any point in, P₀And P₀' are the centers of the two alpha disks respectively;

FIG. 2 is a schematic structural diagram of the measuring and controlling device according to the present invention;

(1 is a base, 2 is an air flotation shaft system, 3 is an aligning and tilting workbench, 4a, 4b and 4c are respectively a precision force sensor, 5 is a static balance measuring platform, 6 is a left upright post, 7 is a right upright post, 8 is a left lower transverse measuring rod, 9 is a left lower telescopic inductive sensor, 10 is a left upper transverse measuring rod, 11 is a left upper telescopic inductive sensor, 12 is a right lower transverse measuring rod, 13 is a right lower lever inductive sensor, 14 is a right upper transverse measuring rod, and 15 is a right upper lever inductive sensor).

Detailed Description

The present invention will be further described with reference to the following specific examples, but the present invention is not limited to these examples.