JPH07191705A - Method and device for automatically adjusting package - Google Patents

Abstract

PURPOSE:To provide a method and device for automatically adjusting a package, which easily and surely decides an adjusting quantity by deciding the rotational amount of a trimmer for the adjustment of an electronic circuit package to be the object of adjustment through the use of an expert system and a neural network and rotating the trimmer for adjustment. CONSTITUTION:A measuring part 1 measures an output value from the electronic circuit package by means of a measuring instrument. A mechanism part 2 moves a head for adjustment to the position of the trimmer for adjustment by a robot. A control part 3 decides the rotational amount of the trimmer for adjustment by means of an expert system expressing the relation between the output value and the rotational amount and a neural network absorbing an error which can not be expressed by the expert system, and controls the measuring part 1 and the mechanism part 2 so as to automatically adjust the electronic circuit package.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a package automatic adjusting method and device.

[0002]

2. Description of the Related Art As an example of a conventional automatic package adjusting method and device, there is an electronic circuit adjusting device described in Japanese Patent Laid-Open No. 02-092004. This electronic circuit adjustment device is adapted to obtain the adjustment amount of the gain and offset of the electronic circuit by fuzzy reasoning from the AC component and the DC component of the output of the electronic circuit when a sine wave input is applied, and perform the adjustment. ing.

[0003]

The above-mentioned conventional electronic circuit adjusting device has a problem in that, when performing fuzzy inference, it is necessary to perform an operation based on a rule every time to calculate an adjustment amount. There is.

An object of the present invention is to determine the adjustment amount by rotating the adjustment trimmer by determining the rotation amount of the adjustment trimmer of the electronic circuit package to be adjusted using an expert system and a neural network. An object of the present invention is to provide a simple and reliable package automatic adjusting method and device.

[0005]

According to the automatic package adjusting method of the present invention, the relationship between the output value from the electronic circuit package and the rotation amount of the adjusting trimmer of the electronic circuit package is expressed by an expert system, and the expert system is used. According to, the rotation amount of the trimmer for adjustment is obtained so that the output value from the electronic circuit package becomes a desired value,
The adjustment trimmer is rotated and coarsely adjusted according to the determined rotation amount of the adjustment trimmer, the error caused by the expert system is learned by a neural network, and the learned neural network is used to perform the adjustment from the electronic circuit package. The configuration is such that the rotation amount of the adjustment trimmer for causing the output value of 1 to match the desired value is obtained, and the corresponding adjustment trimmer is rotated according to the obtained rotation amount to perform fine adjustment.

The automatic package adjusting apparatus of the present invention comprises a measuring section for measuring an output value from an electronic circuit package, a mechanical section for rotating an adjusting trimmer, the measuring section and the measuring section for controlling the mechanical section. The relationship between the output value from the electronic circuit package and the rotation amount of the adjusting trimmer of the electronic circuit package measured by the above is expressed by an expert system, and the electronic circuit newly adjusted by using this expert system The amount of rotation of the adjustment trimmer is calculated so that the output value of the package becomes a desired value, and the mechanism unit rotates the adjustment trimmer of the electronic circuit package newly adjusted according to the obtained amount of rotation. Coarse adjustment is performed to learn the error caused by the expert system by a neural network, and The output value from the electronic circuit package that is newly adjusted by using a network, the rotation amount of the adjustment trimmer for matching the desired value is obtained, and according to the obtained rotation amount, This is a configuration including a control unit that actually rotates the adjustment trimmer by a mechanical unit to perform fine adjustment.

[0007]

Embodiments of the present invention will now be described with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention.

The measuring unit 1 measures the output value from the electronic circuit package with a measuring instrument, and the mechanical unit 2 moves the adjusting head to the position of the adjusting trimmer by the robot to rotate the adjusting trimmer, and the control unit 3 Controls the measuring unit 1 and the mechanical unit 2 by the control device.

FIG. 2 is an explanatory view for explaining the principle of the present invention.

From the expert system 12 expressing the relationship between the output value from the electronic circuit package and the rotation amount of the trimmer for adjusting the package, the n-dimensional target output value from the package X _D = (x _D1 , x _D2 , ... ., X _Dn ) ^{T, the} amount of rotation of the trimmer in the m-dimension Y _M = (y _M1 , y _M2 , ..., y _Mm ) ^T is calculated, and the trimmer is rotated according to the calculated rotation amount. adjust.

Next, the neural network 21 expressing the relationship of the error generated by the expert system 12
Thus, the n-dimensional target output value X _D = (x _D1 , x _D2 , ..., X _Dn ) ^T and the output value from the n-dimensional electronic circuit package _XP = (x _P1 , x _P2 , ... , X _Pn ) ^T and the rotation amount error of the m-dimensional trimmer ΔY _N = (Δy _N1 , Δy _N2 , ..., Δy _Nm ) ^T , and the trimmer Correct the rotation and make fine adjustments.

By the way, the expert system 12 is Such as the IF-THEN rule,
X ₁ , X ₂ , X ₃ , ... Are n-dimensional output values, Y ₁ ,
Y _2, Y _3, · · · represents the rotation amount of m dimension.

FIG. 3 is an explanatory diagram for explaining the neural network.

At the time of learning, an n-dimensional output error ΔX = (Δx ₁ , Δx ₂ , ..., Δx _n ) ^T from the electronic circuit package and an m-dimensional rotation amount error ΔY = of the trimmer for adjusting the package ΔY = The neural network 21 learns the relationship ΔY = F (ΔX) of (Δy ₁ , Δy ₂ , ..., Δy _m ) ^T. However, F represents an m × n dimensional function. Here, a hierarchical network is used for the neural network 21,
Learning is performed by backpropagation from an m-dimensional teacher signal ΔZ = (Δz ₁ , Δz ₂ , ..., Δz _m ) ^T based on the amount of rotation when a skilled worker actually makes an adjustment. .

At the time of adjustment, the rotation amount of the trimmer is calculated by using the learned neural network 21 so that the output value from the package completely matches the desired value, and according to the calculated rotation amount. , Rotate the trimmer to adjust.

A neural network, especially a hierarchical network and backpropagation is described in, for example, PDP model-February 27, 1989.
Cognitive Science and Search for Neuron Networks- (First Edition), D.I.
E. Ramelhart, J. L. McClelland, PDP Research Group, Translated by Shunichi Amari, Industrial Books.

[0018]

As described above, the present invention is an expert that expresses the rotation amount of the adjustment trimmer of the electronic circuit package as a relation between the output value from the electronic circuit package and the rotation amount of the adjustment trimmer of the package. Determined by the system, rotate the trimmer for adjustment, and then, by the neural network, determine the amount of rotation of the trimmer so as to absorb the error that cannot be expressed by the expert system, and perform automatic adjustment by rotating. The determination of the adjustment amount is simple and reliable, which has the effect of reducing the man-hours and time required for the adjustment. Also, pay special attention because the expert system expresses the relationship between the output value from the electronic circuit package and the rotation amount of the trimmer for adjusting the package, and the neural network absorbs errors that cannot be expressed by the expert system. There is also an effect that it can be applied to automation of adjustment of a wide and general electronic circuit package.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention.

FIG. 2 is an explanatory diagram illustrating the principle of the present invention.

FIG. 3 is an explanatory diagram illustrating a neural network.

[Explanation of symbols]

1 Measuring unit 2 Mechanism unit 3 Control unit 11 Electronic circuit package 12 Expert system 21 Neural network X _D Target output value X _P Rotation amount ΔY _N Rotation amount error ΔX Output error ΔY Rotation amount error ΔZ Teacher signal

Claims (2)

[Claims] 1. A relationship between an output value from an electronic circuit package and a rotation amount of an adjusting trimmer of the electronic circuit package is expressed by an expert system, and the output value from the electronic circuit package is a desired value by the expert system. The rotation amount of the adjustment trimmer is obtained so that the adjustment trimmer is rotated according to the obtained rotation amount of the adjustment trimmer to perform coarse adjustment, and the error generated by the expert system is learned by a neural network. , The rotation amount of the adjustment trimmer for making the output value from the electronic circuit package match the desired value by using the learned neural network, and the rotation of the corresponding adjustment trimmer according to the obtained rotation amount. A package automatic adjustment method characterized by performing a fine adjustment. 2. A measuring section for measuring an output value from an electronic circuit package, a mechanism section for rotating an adjusting trimmer, the measuring section and the electronic circuit for controlling the mechanism section and measured by the measuring section. The relationship between the output value from the package and the rotation amount of the trimmer for adjustment of the electronic circuit package is expressed by an expert system, and the output value of the electronic circuit package newly adjusted by using this expert system is desired. The rotation amount of the adjustment trimmer to be a value is obtained, and the adjustment trimmer of the electronic circuit package that has been newly adjusted by the mechanism unit is rotated according to the obtained rotation amount to perform rough adjustment, The error generated by the expert system is learned by a neural network, and the newly created by using the neural network after learning. The output value from the electronic circuit package that is the target of the adjustment is obtained by obtaining the rotation amount of the adjustment trimmer so that the output value from the electronic circuit package matches the desired value. An automatic package adjusting device, comprising: a controller for rotating a trimmer to perform fine adjustment.