Data processing: artificial intelligence – Knowledge processing system
Reexamination Certificate
1998-01-23
2002-08-20
Davis, George B. (Department: 2121)
Data processing: artificial intelligence
Knowledge processing system
C706S046000, C706S047000, C706S048000, C706S059000
Reexamination Certificate
active
06438532
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to an adjustment rule generating method and apparatus for generating an adjustment rule for appropriately and easily adjusting an input to a multiple-input/output system having nonlinear characteristics to obtain a desired output from the system, and an adjustment control method and apparatus for adjusting the input to the system using the generated adjustment rule.
In an adjustment operation at a plant, a device production line, or a maintenance operation, when a certain element in the system is adjusted, a plurality of other elements vary upon adjustment, so it is often difficult to properly adjust all elements.
How to adjust an input to obtain a desired output is a general problem. To solve this problem, various means have been implemented conventionally.
In fact, the problem of an adjustment parameter (to be simply referred to as a parameter hereinafter) and the output is often concomitant with the original input/output relationship. For this reason, an effective result can hardly be obtained.
As reasons for this, the following three main factors are considered.
1. Complex correlation between the parameter and the output
2. Nonlinearity of the parameter and output
3. Maldistribution of data
Both the parameter and the output are generally multidimensional rather than one-dimensional (variable) and have complex causality. The relationship between the parameter and the output is not linear. Resultant data is small in quantity or maldistributed, so the characteristics between the parameter and the output cannot be sufficiently described using such data. It can be supposed that these factors make the problem difficult to solve.
To solve this problem, not only means based on the theory of linear mathematics but also means reflecting the farsighted knowledge or intuition of persons who have been concerned in actual adjustment have been used. For example, a method using fuzzy inference or qualitative causality reasoning is used.
The fuzzy inference can be effective for a system having nonlinear characteristics. However, the fuzzy inference is regarded eventually “successful” only when the membership function or adjustment rule can be appropriately defined.
Generally, the fuzzy theory is applied to a nonlinear system. However, analogical reasoning can hardly be made because the response from an object is not linear. In addition, trial and error in system identification also tends to be cumbersome. Even when the system can be identified using a nonlinear model, the input amount for adjustment (manipulated variable for control) is hard to calculate because of the nonlinear model. This results in a difficulty in setting the membership function or adjustment rule. Furthermore, it cannot be guaranteed that the initial rule is still effective for a variation in system characteristics.
Essentially, this also applies to qualitative causality reasoning. Once the causality is clarified, analysis is automatically performed by a computer. However, data in checking the causality depends on human determination, like the fuzzy inference. More specifically, even when data is to be semi-automatically processed and modeled, the human data determination reference must be defined in advance. In this respect, the qualitative causality reasoning is essentially identical to the fuzzy inference (e.g., Jpn. Pat. Appln. KOKAI Publication No. 7-191706).
In reasoning based on causality, normally, the current state is analyzed on the basis of past data (past events). This processing requires a large quantity of past data. This method is convenient when a relatively large plant (system) is operated for a long time. However, when an individual difference between objects is assumed as in adjusting parameters of individual products on a production line, or when adjustment is to be made in response to an environmental change, the number of data is limited because adjustment cannot always depend on other individual data. Therefore, adjustment can hardly be performed using the method based on the conventional event data.
Reasoning does not suffice for adjustment. Unlike system observation based on two references, e.g., faulty diagnosis for checking whether the interior of a system is faulty or not (subsequent processing is left to human operations), some action must be taken for the system after situation determination in the control system.
BRIEF SUMMARY OF THE INVENTION
The present invention has been made in consideration of the above situation, and has as its object to provide a system having the following characteristic features.
1. Adjustment is performed while sampling data
2. A large quantity of data is not required in advance
3. Nonlinear characteristics can be coped with.
More specifically, the present invention has as its object to provide an adjustment rule generating method and apparatus for generating an adjustment rule to adjust an object having multiple variables (multiple-input/output system) whose correlation has complex nonlinear characteristics.
It is another object of the present invention to provide an adjustment control method and apparatus for adjusting an object in accordance with a generated adjustment rule.
According to the present invention, the adjustment operation can be appropriately standardized and automated.
The adjustment rule generating method and apparatus of the present invention are characterized in that a table (dependency relationship table) representing qualitative characteristics is assumed in which inputs (to be referred to as manipulated variables hereinafter) are classified in units of change patterns of outputs (to be referred to as controlled variables hereinafter) having influence, and
an operation procedure (to be referred to as an adjustment rule hereinafter) for adjustment is generated.
The adjustment control method and apparatus according to the present invention are characterized in that it is determined whether the current object situation exhibits an exceptional behavior (vibration/saturation), on the basis of an instruction (selection of an adjusted controlled variable and a manipulated variable) obtained from an automatically generated adjustment rule and a past operation in response to an occasionally output deviation. If it is determined that no exceptional behavior is observed, the instruction of the generated adjustment rule is executed; otherwise, the correction amount of the manipulated variable which is input to the object to be adjusted is given assuming that a predetermined input operation is performed.
(1) An adjustment rule generating apparatus which determines the manipulated variable of the adjustment object or sets the value of a variable parameter (the variable parameter will not particularly be discriminated from the manipulated variable hereinafter) of an adjustment object such that a controlled variable within an allowable range can be obtained, is characterized by comprising
adjustable controlled variable selection means for receiving a change in controlled variable corresponding to each manipulated variable of the adjustment object and qualitative feature data of a change difference between controlled variables and defining some manipulated variables which can be independently adjusted from the feature data in units of controlled variables, and adjustment rule format generating means for converting adjustable controlled variable data output from the adjustable controlled variable selection means in units of manipulated variables into a predetermined format and outputting the format as an adjustment procedure.
(2) The adjustment rule generating apparatus of arrangement (1) is characterized in that the change in controlled variable corresponding to each manipulated variable of the adjustment object is defined by input data (manipulated variable characteristics and input/output dependency relationship table; to be referred to as a dependency table hereinafter) as binary data which describes whether each manipulated variable affects the controlled variable and binary data of a change pat
Davis George B.
Foley & Lardner
Kabushiki Kaisha Toshiba
LandOfFree
Adjustment rule generating and control method and apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Adjustment rule generating and control method and apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Adjustment rule generating and control method and apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2934436