CN104933231B - A kind of flexible assemble production line type selecting layout method using the more knowledge models of cascade - Google Patents

Abstract

The invention discloses a type selection and layout method for a flexible assembly production line using a cascade multi-knowledge model. Establish a production capacity knowledge model, and select the equipment with the lowest production capacity as the equipment on the assembly line; establish a process knowledge model, arrange various equipment in the production line according to the process knowledge model; establish a structural knowledge model, and use the structural knowledge model to calculate and connect to the production line The parameters of the conveying device between different equipment in the production line; the spatial knowledge model is established to obtain the size and position of the buffer zone between each equipment in the production line. The invention integrates knowledge models in different fields to complete the selection and layout of the production line; improve the design speed and quality, which is beneficial to the sharing and transmission of production line information; reduce the cost of space use and enterprise investment; improve the work efficiency of the assembly line; Effective utilization of such equipment.

Description

A Type Selection and Layout Method of Flexible Assembly Production Line Using Cascade Multi-Knowledge Model

technical field

The invention relates to a type selection and layout method for a production line, in particular to a type selection and layout method for a flexible assembly production line using a cascade multi-knowledge model.

Background technique

Flexible automated assembly production line is a complex and highly automated system that organically combines technologies such as machinery, electronics, computer and system engineering, connects multiple processing and assembly equipment that can be combined and adjusted, and is equipped with an automatic delivery device The production line formed can satisfactorily solve the contradiction between high automation and high flexibility of mechanical manufacturing. The flexible automated production line can not only liberate manpower from heavy or repetitive labor, but also greatly reduce the cost of launching new products—that is, when a new product is launched, the factory only needs to replace the fixtures and other equipment of the new product. Yes, there is no need to rebuild or replace the entire production line, so as to realize the flexibility of production assembly and reduce the investment cost of fixed assets to the greatest extent.

In the technical system of flexible automated assembly production line, type selection and layout is the core technology. The structural design of the assembly line mainly refers to: the detailed design of the conveying module and the buffer zone, including the internal structural design of the conveying module, the design of parts, the design of the connection interface form between conveying modules, and between modules and equipment. Therefore, it is of great economic benefit and research value to design a targeted flexible automation assembly line layout plan for specific equipment and specific forms of the factory area.

The existing production line layout design process requires senior engineers with many years of work experience to use the rich experience in their minds to complete the design. At the same time, the design carried out in the two-dimensional CAD software system can only display the top view layout due to the relationship of dimensions, which lacks a sense of reality and is not easy to show the size change in the horizontal direction. In terms of operation, due to the conveyor chain, there are many lines in the top view. Even if you just copy the top view of a certain module and drag it, it is time-consuming and laborious. This results in excessive reliance on experience, error-prone and low design efficiency in the design process. Therefore, it is necessary to establish information resources that support the selection layout and provide support for the selection layout. This will not only improve the design speed and quality, but also facilitate the sharing and transmission of information within the enterprise.

Contents of the invention

In order to solve the above problems, the present invention proposes a flexible assembly production line selection and layout method using cascaded multi-knowledge models. Through the integrated knowledge model, the selection and layout of the production line is completed; the design speed and quality are improved, which is beneficial to the production line information. Sharing and transmission; reduce space use costs, reduce enterprise investment; improve the efficiency of assembly production lines; improve the effective utilization of various equipment.

Technical scheme of the present invention adopts following mode:

Establish a capacity knowledge model based on a decision tree, and select the equipment with the lowest capacity as the equipment on the assembly line;

Establish a framework-based process knowledge model, and arrange various types of equipment in the production line according to the process knowledge model;

Establish a structural knowledge model based on the neural network model, and calculate the parameters of the conveying device connected between different equipment in the production line through the structural knowledge model;

Establish a case-based spatial knowledge model to obtain the size and location of the buffer zone between each device in the production line.

The capacity knowledge model is specifically:

1.1) Establish energy consumption information matrix S=[s ₁ ,s ₂ ,…,s _n ] ^T :

S＝[s ₁ ,s ₂ ,...,s _n ] ^T (1)

Among them, s _i represents the energy consumption information of the i-th device on the production line, i=1,2,...,n, T represents the matrix transposition;

1.2) Calculate the attribute coefficient G of the information matrix S:

Among them, P _i is the frequency at which the information of the i-th device appears in S, and k is the gain coefficient;

1.3) Decompose the energy consumption information matrix S into two sub-matrices U and V, the expressions of the minimum value sub-matrix U and the maximum value sub-matrix V are:

Among them, min(·) represents the minimum value operation, max(·) represents the maximum value operation;

1.4) The energy consumption information matrix S is updated at regular intervals, and each update is calculated and decomposed using the above steps, and the operation is stopped until the number of updates N=50, and the minimum value sub-matrix U and the maximum value obtained from each update Combining the value sub-matrix V to obtain a decision tree as a production capacity knowledge model, the equipment corresponding to the minimum value of the minimum value sub-matrix U is regarded as the equipment with the lowest production capacity.

The process knowledge model is specifically:

2.1) Establish the type matrix W of all equipment in the production line:

W＝[w ₁ ,w ₂ ,...,w _n ] ^T (4)

Among them, w _i (i=1,2,...,n) represents the type of the i-th device on the production line, and T represents matrix transposition;

2.2) Give each device an attribute parameter matrix A under the type:

A＝f ₁ (w ₁ )+f ₂ (w ₂ )+...+f _n (w _n ) (5)

Among them, f _i (w _i ) is the attribute parameter of the i-th device under its own type;

2.3) According to the specific operation process of different equipment arrangements, the process knowledge matrix AG describing the equipment arrangement is obtained:

Among them, g _ij represents the j-th operation step in the i-th equipment operation process, j=1,2,...,m;

2.4) Combining type matrix W, attribute parameter matrix A and process knowledge matrix AG to obtain a complete process knowledge model.

The described structural knowledge model is specifically:

3.1) The following neural network model is established, and the parameters of the conveying device are input into the neural network model for training. The neural network model includes the input layer neuron function d(x), node function q(x) and error function e(x):

Among them, x is the input information of the neural network model, that is, the parameters of the delivery device, e is the natural logarithm, R is the number of neurons in the neural network model, r=1,2,...,R, y _r is the output of the neural network model Information, that is, the actual parameters of the delivery device after the training of the neural network model, and y ^* _r is the reference value of the output information of the neural network model, that is, the reference value of the actual parameters of the delivery device after the training of the neural network model;

3.2) By integrating the operating parameters of different equipment in the production line, the feedback learning matrix F is established:

Among them, x _a is the operating parameter of the a-th device, h _ab is the b-th learning function of the a-th device, a=1,2,...,n, b=1,2,...,z, z is the learning function number of functions;

3.3) The neural network model and the feedback learning matrix are combined to form a structural knowledge model, and then the parameters of the conveying device connecting different equipment in the production line are calculated.

The implementation of the spatial knowledge model is specifically:

4.1) Establish the initial case matrix G consisting of the set buffer cases, the buffer cases include buffer size and location:

G＝[(gd ₁ ,gz ₁ ),(gd ₂ ,gz ₂ ),...,(gd _n ,gz _n )] ^T (9)

Among them, gd _i is the size of the buffer of the i-th device, gz _i is the position of the buffer of the i-th device, and T represents matrix transposition;

4.2) For the current factory environment, use the k-nearest neighbor method to search from the initial case matrix G. If a buffer case suitable for the current factory environment is found, replace the buffer case with the original case matrix of the current factory environment In , a new case matrix is formed; if no buffer case suitable for the current factory environment is found, the original case matrix of the current factory environment is retained;

4.3) For the same factory environment, search iteratively until the number of searches reaches L=20, and use the final case matrix as the spatial knowledge model;

4.4) Therefore, the buffer size and position in the spatial knowledge model are used as the buffer size and position between each device in the production line.

The beneficial effects that the present invention has are:

The invention integrates knowledge models in different fields to complete the selection and layout of the production line; improve the design speed and quality, which is beneficial to the sharing and transmission of production line information; reduce the cost of space use and enterprise investment; Effective utilization of such equipment.

Description of drawings

Fig. 1 is a flowchart logic diagram of the method of the present invention.

detailed description

The present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments.

Step 1 of the present invention is based on a rule-based production capacity knowledge model, and selects by analyzing a set instance matrix, so as to realize low-power device selection and fast calculation of parameters.

Step 2) of the present invention is based on a frame-based process knowledge model, which is a flexible frame structure, which can realize the rational arrangement and layout of various equipment in the production line, and improve the efficient layout of various equipment in the production line.

Step 3) of the present invention is based on the structural knowledge model of the neural network model, which has high calculation accuracy and fast speed, can increase system functions, and can independently study the models of various equipment, and use the results for system design. The structural knowledge model is established accurately and can truly reflect the real working conditions of the conveying devices of different equipment, so as to obtain specific parameters reflecting the layout information of the conveying devices, and provide solutions for the rational configuration of these conveying devices.

Step 4) of the present invention is based on the spatial knowledge model of the case, by analyzing the case of the buffer zone between each equipment in the production line, and solving the currently targeted problem of the size and location of the buffer zone. The case-based spatial knowledge model can be continuously updated and replaced, thereby increasing the ability of the system to solve problems, and realizing the reliability of the size and position of the buffer zone between various equipment in the production line.

Embodiments of the present invention are as follows:

In the first aspect, a capacity knowledge model based on a decision tree is established, and the equipment with the lowest capacity is selected as the equipment on the assembly line.

1.1) Establish energy consumption information matrix S=[s ₁ ,s ₂ ,…,s _n ] ^T :

S＝[s ₁ ,s ₂ ,...,s _n ] ^T (1)

Among them, s _i represents the energy consumption information of the i-th equipment on the production line. The size of the electric power consumed in the alarm operation state, by counting and sorting out the size of the electric power consumed by these devices, the energy consumption information matrix S can be obtained, i=1,2,...,n, T represents the matrix transposition ;

1.2) Calculate the attribute coefficient G of the information matrix S:

Among them, P _i is the frequency at which the energy consumption information of the i-th equipment appears in S, and the frequency can be calculated by using a statistical method for a certain equipment (including CNC assembly machine tools, motors, conveyor belts, transfer machines, elevators, etc.) The number of times the energy consumption information of energy consumption appears in the consumption information matrix S is obtained, and k is the gain coefficient;

1.3) Decompose the energy consumption information matrix S into two sub-matrices U and V, the expressions of the minimum value sub-matrix U and the maximum value sub-matrix V are:

Among them, min(·) represents the minimum value operation, and max(·) represents the maximum value operation. In this step, on the basis of calculating the attribute coefficient G, the energy consumption information matrix S corresponding to the minimum value of the attribute coefficient G is taken as the minimum value The sub-matrix U uses the energy consumption information matrix S corresponding to the maximum value of the attribute coefficient G as the maximum value sub-matrix V.

1.4) The energy consumption information matrix S is updated at regular intervals, and each update is calculated and decomposed using the above steps, and the operation is stopped until the number of updates N=50, and the minimum value sub-matrix U and the maximum value obtained from each update Combining the value sub-matrix V to obtain a decision tree as a production capacity knowledge model, the equipment corresponding to the minimum value of the minimum value sub-matrix U is regarded as the equipment with the lowest production capacity.

The equipment involved above may be CNC assembly machine tools, motors, conveyor belts, transfer machines, elevators, etc.

In the second aspect, a framework-based process knowledge model is established, and various equipment in the production line are arranged according to the process knowledge model.

2.1) Establish the type matrix W of all equipment in the production line:

W＝[w ₁ ,w ₂ ,...,w _n ] ^T (4)

Among them, w _i (i=1,2,...,n) represents the type of the i-th equipment on the production line, these types include assembly type, power type (such as motor, etc.), transmission type (such as conveyor belt, transfer machine, etc.) , by classifying the types of all the equipment in the production line, the type matrix W can be obtained, and T represents the matrix transposition;

2.2) Give each device an attribute parameter matrix A under the type:

A＝f ₁ (w ₁ )+f ₂ (w ₂ )+...+f _n (w _n ) (5)

Among them, f _i (w _i ) is the attribute parameter of the i-th device under its own type; in practice, f _i ( _wi ) can be taken as the size value and rated power value used to describe multiple devices of the same type , floor area and other specific values that reflect the attributes of the equipment.

2.3) According to the specific operation process of different equipment arrangements, the process knowledge matrix AG describing the equipment arrangement is obtained:

Among them, g _ij represents the j-th operation step in the operation process of the i-th equipment. In practice, g _ij can be taken as specific operation steps such as handling, installation, commissioning, overhaul, maintenance, etc. j=1,2,..., m;

2.4) Combining the type matrix W, attribute parameter matrix A and process knowledge matrix AG to obtain a complete process knowledge model, so as to realize the arrangement of various equipment in the production line.

In the third aspect, a structural knowledge model based on the neural network model is established, and the parameters of the conveying device connected between different devices in the production line are calculated through the structural knowledge model.

3.1) The following neural network model is established, and the parameters of the conveying device are input into the neural network model for training. The neural network model includes the input layer neuron function d(x), node function q(x) and error function e(x):

Among them, x is the input information of the neural network model, that is, the parameters of the conveying device. These parameters include the length, number of channels, and speed of the conveying device, e is the natural logarithm, and R is the number of neurons in the neural network model. The value may be determined by the training speed and training accuracy of the parameters, and usually can be taken as a value not greater than 20, which is taken as 10 in the present invention, r=1,2,...,R, y _r is the output of the neural network model Information, that is, the actual parameters of the delivery device after the training of the neural network model, and y ^* _r is the reference value of the output information of the neural network model, that is, the reference value of the actual parameters of the delivery device after the training of the neural network model;

3.2) By integrating the operating parameters of different equipment in the production line, the feedback learning matrix F is established:

Among them, x _a is the operating parameter of the a-th device, h _ab is the b-th learning function of the a-th device, a=1,2,...,n, b=1,2,...,z, h _ab is in In practice, it can be taken as an operation function such as a rounding operation function, a square root operation function, and an absolute value operation function for the operating parameter x _a , and z is the number of learning functions;

3.3) The neural network model and the feedback learning matrix are combined to form a structural knowledge model, and then the parameters of the conveying device connecting different equipment in the production line are calculated. After obtaining these parameters, the internal structure design and parts design of the conveying module can be carried out, so as to obtain a conveying device suitable for the factory environment.

The conveying device mentioned above may specifically be a conveying mechanism such as a conveyor belt, a transfer machine, or an elevator.

In the fourth aspect, a case-based spatial knowledge model is established to obtain the size and location of the buffer zone between each device in the production line.

4.1) Establish the initial case matrix G consisting of the set buffer cases, the buffer cases include buffer size and location:

G＝[(gd ₁ ,gz ₁ ),(gd ₂ ,gz ₂ ),...,(gd _n ,gz _n )] ^T (9)

Among them, gd _i is the buffer size of the i-th device, the specific value of gd _i is the buffer area value reflecting the buffer size, gz _i is the buffer position of the i-th device, and the specific value of gz _i is The value reflects the buffer position and the coordinate value of the buffer in the factory, and T represents the matrix transposition;

4.2) For the current factory environment, use the k-nearest neighbor method to search from the initial case matrix G,

If a buffer case suitable for the current factory environment is found, replace the buffer case into the original case matrix of the current factory environment to form a new case matrix;

If no buffer case suitable for the current factory environment is found, the original case matrix of the current factory environment is retained;

The above-mentioned k-nearest neighbor method is a relatively mature search algorithm, and its specific principle can be seen in "Yu Yi. K-Nearest Neighbor Method Text Classification Algorithm Analysis and Improvement [J]. Journal of Huazhong University of Science and Technology (Natural Science Edition), 2008, 33(4):143-145".

4.3) For the same factory environment, search iteratively until the number of searches reaches L=20, and use the final case matrix as the spatial knowledge model;

4.4) Therefore, the buffer size and position in the spatial knowledge model are used as the buffer size and position between each device in the production line.

Through the above design steps, the selection and layout results suitable for the current factory working environment and production assembly requirements can be finally obtained, that is, on the basis of obtaining the equipment model, quantity, parameters, and buffer size and location, the structural knowledge model and spatial knowledge are used The model acquires the layout method of these equipments to realize the selection and layout of the flexible production line.

The specific embodiments above are used to explain the present invention, rather than to limit the present invention. Within the spirit of the present invention and the protection scope of the claims, any modification and change made to the present invention will fall into the protection scope of the present invention.

Claims (1)

1. A method for selection and layout of a flexible assembly production line using a cascaded multi-knowledge model, characterized in that: comprising the following steps: Establish a production capacity knowledge model based on a decision tree, and select the equipment with the lowest production capacity as the equipment on the assembly line; the specific capacity knowledge model is: 1.1) Establish energy consumption information matrix S=[s ₁ ,s ₂ ,…,s _n ] ^T : S＝[s ₁ ,s ₂ ,…,s _n ] ^T (1) Among them, s _i represents the energy consumption information of the i-th device on the production line, i=1,2,...,n, T represents the matrix transposition; 1.2) Calculate the attribute coefficient G of the information matrix S: Among them, P _i is the frequency at which the information of the i-th device appears in S, and k is the gain coefficient; 1.3) Decompose the energy consumption information matrix S into two sub-matrices U and V, the expressions of the minimum value sub-matrix U and the maximum value sub-matrix V are: Among them, min(·) represents the minimum value operation, max(·) represents the maximum value operation; 1.4) The energy consumption information matrix S is updated at regular intervals, and each update is calculated and decomposed using the above steps, and the operation is stopped until the number of updates N=50, and the minimum value sub-matrix U and the maximum value obtained from each update Combining the value sub-matrix V to obtain a decision tree as a production capacity knowledge model, the equipment corresponding to the minimum value of the minimum value sub-matrix U is regarded as the equipment with the lowest production capacity; Establish a framework-based process knowledge model, and arrange various types of equipment in the production line according to the process knowledge model; The process knowledge model is specifically: 2.1) Establish the type matrix W of all equipment in the production line: W＝[w ₁ ,w ₂ ,...,w _n ] ^T (4) Among them, w _i (i=1,2,...,n) represents the type of the i-th device on the production line, and T represents matrix transposition; 2.2) Give each device an attribute parameter matrix A under the type: A＝f ₁ (w ₁ )+f ₂ (w ₂ )+...+f _n (w _n ) (5) Among them, f _i (w _i ) is the attribute parameter of the i-th device under its own type; 2.3) According to the specific operation process of different equipment arrangements, the process knowledge matrix AG describing the equipment arrangement is obtained: Among them, g _ij represents the j-th operation step in the i-th equipment operation process, j=1,2,...,m; 2.4) Combining the type matrix W, attribute parameter matrix A and process knowledge matrix AG to obtain a complete process knowledge model; Establish a structural knowledge model based on the neural network model, and calculate the parameters of the conveying device connected between different equipment in the production line through the structural knowledge model; The described structural knowledge model is specifically: 3.1) The following neural network model is established, and the parameters of the conveying device are input into the neural network model for training. The neural network model includes the input layer neuron function d(x), node function q(x) and error function e(x): Among them, x is the input information of the neural network model, that is, the parameters of the delivery device, e is the natural logarithm, R is the number of neurons in the neural network model, r=1,2,...,R, y _r is the output of the neural network model Information, that is, the actual parameters of the delivery device after the training of the neural network model, and y ^* _r is the reference value of the output information of the neural network model, that is, the reference value of the actual parameters of the delivery device after the training of the neural network model; 3.2) By integrating the operating parameters of different equipment in the production line, the feedback learning matrix F is established: Among them, x _a is the operating parameter of the a-th device, h _ab is the b-th learning function of the a-th device, a=1,2,...,n, b=1,2,...,z, z is the learning function number of functions; 3.3) Combine the neural network model and the feedback learning matrix to form a structural knowledge model, and then calculate the parameters of the conveying device connecting different equipment in the production line; Establish a case-based spatial knowledge model to obtain the size and location of buffer zones between equipment in the production line; The implementation of the spatial knowledge model is specifically: 4.1) Establish an initial case matrix G consisting of the set buffer cases, including buffer size and location: G=[(gd ₁ ,gz ₁ ),(gd ₂ ,gz ₂ ),... ,(gd _n ,gz _n )] _T (9) Among them, gdi is the size of the buffer of the i-th device, gzi is the position of the buffer of the i-th device, and T represents matrix transposition; 4.2) For the current factory environment, use the k-nearest neighbor method to search from the initial case matrix G, If a buffer case suitable for the current factory environment is found, replace the buffer case into the original case matrix of the current factory environment to form a new case matrix; If no buffer case suitable for the current factory environment is found, the original case matrix of the current factory environment is retained; 4.3) For the same factory environment, search iteratively until the number of searches reaches L=20, and use the final case matrix as the spatial knowledge model; 4.4) Therefore, the buffer size and position in the spatial knowledge model are used as the buffer size and position between each device in the production line.