JPH0456746A - Method for designing composition of steel product - Google Patents

Abstract

PURPOSE:To automatize the compositional design of a steel product and to effectively extract the most suitable compsn., at the time of deciding the compositional design of a steel product as manufacturing specifications against required specifications by limiting the range of the componental value of the elements to be added according to carbon equivalent properties and the properties of total elements. CONSTITUTION:According to product specifications (e.g., form and size, standards and designations) offered from customers to steel product makers, designing conditions such as the system of components (the constitution of the elements to be added), the range of allowable componental value, carbon equivalent properties and the properties of total elements are decided by the search of a data base in which the knowledge of past data and experts are sorted, and the case most similar to the product specifications is searched out to extract information corresponding to the designing conditions. In this way, the amounts to be added are set to the range narrower than the allowable componental range of the elements to be added, the compositional design most suitable for the product specifications is finally decided, and, in accordance with the above, the compositional design by the amt. of the elements to be added is decided.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to compositional design of steel products.

[Prior Art] Steel products have been the most widely used materials in the past, and various necessary functional conditions are presented by users to manufacturers in the form of required specifications depending on the form of use. Manufacturers determine manufacturing specifications for each steel product that has such required specifications, but determining stable and low-cost manufacturing specifications in a short period of time improves product reliability and promotes sales.

This is extremely important in terms of shortening delivery times. Composition design determines the composition and amount of added elements to steel products, and has a significant impact on the material properties such as strength and impact properties, which are the basic properties of steel products among manufacturing specifications. This is one of the most important design items since it also affects other manufacturing specifications.

As devices for supporting this type of composition design, there have conventionally been databases and search systems for steel product standards, and databases and search systems for past manufacturing results. However, these systems only support the acquisition of reference data and do not support the design itself.

In addition, since the manufacturing method of steel products itself has not yet been fully elucidated in terms of mathematics, physics, etc., in designing the composition, skilled experts have obtained references obtained using the above system. The reality is that designs are made by referring to data and references, and by making full use of the technical and experiential knowledge of experts.

[Problems to be Solved by the Invention] As mentioned above, conventional compositional design is highly dependent on the knowledge of experts, which tends to cause variations in design results depending on the level of knowledge, and takes a long time to design. It takes. There are problems such as difficulty in securing skilled experts, and automation of composition design has become an issue.

Furthermore, composition design involves a vast design space with diverse design conditions, many design parameters, and a wide range of possible values for each parameter. Usually, designs are made according to a specific procedure, and it is virtually difficult even for experts to design optimally in all cases, and a solution to this problem is also desired.

Regarding the automation of this composition design, it is possible to use knowledge engineering and expert system technology that applies it, but simply systemizing the knowledge of experts is insufficient to solve the aforementioned optimal needle setting problem. However, a major challenge is how to efficiently extract the optimal composition from the vast design space.

Therefore, the present invention aims to automate the composition design of steel products and efficiently extract the optimum composition.

[Means for Solving the Problems] In order to solve the above problems, in the present invention, based on input product specifications, at least a component system and an allowable component value range corresponding to the composition of added elements are determined.

Determine design conditions including Ce q characteristics and total element characteristics, and determine the range of component values to be considered as the component values of each additive element based on the above allowable component values based on the Ce q characteristics and total element characteristics. Generate component value candidates for each additive element within the above-mentioned limited component value range, and perform stepwise calculation according to the combination conditions indicated by the component system bundle corresponding to the component system of the design conditions. A composition candidate is generated by combining each component value candidate, and composition candidates that do not satisfy the design conditions are deleted, and a final composition is determined from among the composition candidates based on the results of evaluating the composition candidates under predetermined evaluation conditions. Select a candidate.

[Function] The input product specifications are, for example, mold size and standard name, and based on these, design conditions such as component system (composition of additive elements), allowable component value range, Ceq characteristics, total element characteristics, etc. 1. For example, this can be accomplished by searching a database that organizes past study performance data and expert knowledge, finding the case most similar to the product specifications, and obtaining information corresponding to the design conditions. Ru.

Next, the obtained component value range is further limited within the allowable component value range.

For example, let the component values of carbon, silicon, and manganese be X, Y, and 2, respectively, and regarding these component values.

If the following conditional expression exists, 0 ≦X, Y, Z ≦ 4 ... (1)
) From the equation, it is clear that Y + Z ≦ 8, so from the equation (2), it must be X≧2. That is, in this case, the range can be limited from the initial range of 0≦X≦4 to 2≦X≦4 (the same applies to Y and Z). Based on this idea, the allowable component value range is applied as the initial generation range to the design condition formula defined as the Ce q characteristic or total element characteristic, and the process is repeated until convergence. , the range is limited by deleting in advance a range that has no possibility of satisfying the current design conditions. This process can significantly reduce the amount of composition candidates generated in the next step.

Next, component value candidates for each additive element are generated within this limited component value range. For example, if the component value of carbon is X, and the component value range of X is 10≦X≦20, then
If the numerical precision is 2, 10°12.14, 16, 18
and 20 are generated as component value candidates for X.

For each component value candidate generated in this way,
These combinations are generated step by step according to the combination steps indicated by the component system bundle corresponding to the component system of the design conditions,
These are considered as composition candidates. The component system bundle here means a component system and a partial component system generated for it organized by their content relationships. For example, if the component system in the design condition is C-3t-M n -V. , the component system bundle is as shown in Figure 2. In the example in Figure 2, C, Si, Mn,...
First, C-8t using component value candidates for .

A combination of CM n +... is generated, and then a combination of C-3i-
The process proceeds step by step, such as generating the combinations Mn, C-S i -V', ``.

Further, among these combinations, those that do not satisfy the design conditions are deleted at an early stage to prevent generation of wasteful combinations. For example, as one of the design conditions, M
If the condition n/C≧a exists, then C-M
The combinations of n are monitored, and if a combination that satisfies Mn/C<a is generated, that combination is deleted as soon as it is generated, so that it is not used in combinations in subsequent steps. This eliminates waste and can efficiently generate composition candidates.

Finally, a final composition candidate with a high evaluation is selected from among the generated composition candidates by performing an evaluation using criteria such as alloy cost and ease of manufacture.

By implementing the above method on, for example, an expert system, it becomes possible to automate the composition design of steel products, and moreover, it is possible to efficiently extract the optimum composition.

Other objects and features of the present invention will become apparent from the following description of embodiments with reference to the drawings.

[Example] FIG. 3 shows the configuration of a composition designing apparatus that implements the present invention. Referring to FIG. 3, this device is comprised of a computer main body 21, an input device 26 and an output device 27 connected thereto. On the storage device of the computer main body 21, a knowledge base 22. A database 23 and an inference engine 25 are arranged as software. The inference engine 25 performs various inferences based on a large amount of knowledge data registered on the knowledge base 22, and processes the data input from the input device 26 while referring to the data on the database 23. The result is output to output bag N27. The knowledge base 22 contains a large number of data (
A set of rules) is maintained. The database 23 holds a large amount of data on the contents of various standards and past study results.

An outline of the processing in the apparatus shown in FIG. 3 is shown in FIGS. 1 and 4a. The operation of this device will be briefly explained. First, information including the mold size and standard name is input as product specifications from the input bag W126. There are specification constraints and manufacturing process constraints for the product. In other words, the fourth
As shown in figure a, the tensile properties and component value properties are constraints on the specifications.

Component Ceq characteristics and total element characteristics exist,
In addition, as process constraints, component upper and lower limit range constraints and strength Ce
There are q upper and lower range constraints. The role of the apparatus shown in FIG. 3 is to automatically design the composition so as to satisfy all of these conditions.

The result of the treatment is, for example, C in the column ``Composition design'' in Figure 4a.
=0.1810.15 (upper limit value/lower limit value), St=
0.40/ 0.30, M n = 1.35/
It is determined as the range (upper and lower limits) of each component value, as shown as 1.25.

FIG. 1 shows an outline of the processing procedure of the apparatus shown in FIG. The contents of each processing step in FIG. 1 will be explained below.

1: Here, the specifications of the steel product to be designed are input. This specification includes the mold size (for example, H-section steel, flange thickness, web thickness, flange width, and web height) and the standard name (for example, JIS, 5M41B).

2: Based on the specification information input in step 1, design conditions, that is, process constraints and customer/standard specifications shown in FIG. 4a are determined. Among the design conditions of Niku et al., the process constraints are constant, so constants registered in the system in advance are used for these. In addition, since the component value characteristics and total element characteristics are basically determined by the standard of the input specification, the standard database on the database 23 is searched using the standard name as a search key, and the component value characteristics and total element characteristics assigned to the standard are searched. Obtain information on total element characteristics. However, if a value that partially differs from the standard is specified as a customer specification, that customer specification will be given priority and replaced. Furthermore.

The characteristic expression and constraint values of the component Ce q characteristics are specified as customer specifications.

The component system to be the design target is determined as follows. That is, by referring to the design track record database (which holds data organizing past design records) provided on the database 23, and searching for a product specification that is most similar to the specification based on predetermined search conditions, The content is determined to be the same as the component data of the product.

Regarding the tensile properties, conditions regarding TS, 'YP, and EL defined by customer specifications or standards are converted into constraint expressions regarding the strength Ceq equation. Here, the intensity Ce q formula is
For example, intensity Ceq=C+Si/a+Mn/b+...where C: carbon component value.

Si: Component value of silicon.

Mn: component value of manganese.

at by...: constant This is a determination formula regarding intensity using component values.

The conditions regarding TS, YP, and EL are converted into intensity Ceq equations by the following process.

(1) Refer to the manufacturing record database (which holds data organized by past manufacturing records) provided on the database 23, and search for one or more product specifications similar to the specification based on predetermined search conditions. Explore.

(2) From the actual data of TS, YP, EL, and component values regarding the search result products, estimate the change characteristics of each of TS, YP, and EL with respect to changes in intensity Ceq during manufacturing.

(3) TS for changes in estimated intensity Ceq.

Using the change characteristics of each of YP and EL, a range of intensity Ce q that satisfies all the conditions regarding TS, YP, and EL defined by customer specifications or standards is determined, and this is used as a constraint expression.

3: Based on the concept using equations (1) and (2) above, the allowable component value range (component value characteristic in Figure 4a) is ) as the initial generation range and repeating the process until convergence, remove in advance the range that has no possibility of satisfying the current design conditions from the allowable component value range,
Limit the scope.

This process can significantly reduce the amount of composition candidates generated in the process described below.

In reality, the scope is limited by applying the following constraints.

First, as a design condition, the range is limited using one lower limit constraint expression shown in the following equation (3) and three upper limit constraint equations shown in the following equation (4).

ΣPikXk≧ F i win ; i =1.2
．． −・・I(k=1～K) ・・・(
3) ΣQjkXk: aGjmax; j=1%), -
---J (k=1~K) ...(
4) However, P xke Q jk "non-negative constraint equation coefficients Fimin, Gjmax: constraint values.

Xk: k=1.2. ...K is, for example, Xl: component value of carbon C, X2: component value of silicon Si, ...
It shows the component value of each element, such as Xkmin
:5Xk≦Xlt+ax: k=1,2,3
．． ...K... (5) Assume that there is a range constraint.

Equations (3) and (4) are the tensile characteristic equations in Figure 4a.

In the component Ceq characteristic equation or the total element characteristic equation, Equation (5) corresponds to the component value characteristic shown in the figure.

Here, if we transform equation (3), Xk≧(Fimin-(ΣPir
Xk)) / Pik (range of Σ: r=1~, P
ik≠O) ・(6) is obtained, but due to the range constraint of equation (5), the right side of equation (6) becomes XkL= (Fi win−(Σ Pir
rwax-Pik Xk wax)) /Pik
(Range of Σ: r=1 to K) ...(7) It has the minimum value XkL. Therefore, equation (6) and equation (7)
), a new range constraint such as Xk ≧ XkL (8) is generated. Here, X kL >
If Xk win, from equations (5) and (8), the range constraint regarding Xk is set to XkL:5Xk:5Xkmax:=1.
2. ...K ... can be narrowed down to (9).

Next, when formula (4) is transformed, Xk≦(Gj wax-(ΣQjr Xr-Qjk
k)) /Qjk (The range of Σ is r=l ~, Qj
k≠O) ---(10) is obtained, but the (5th)
Due to the range constraint of the equation, the right side of equation (10) is Xk)! = (Gj + wax-(ΣQjr Xr w
in-Qjk Xk win)) /Qjk (The range of Σ is r = 1 ~)
1) has a maximum value XkH. Therefore, a new range constraint is generated from equations (10) and (11) as follows: Xk ≦XkH...-(12). Here, X kH<
If X k wax, the formula (5) and the formula (12)
From the formula, the range constraint regarding X and k is set as Xkmin≦Xk≦XkH: k=1.2. ...K ...(13
) can be narrowed down to

That is, for the range constraint in Equation (5), the lower limit side is narrowed by using the lower limit constraint expression in Equation (3), and the upper limit side is narrowed by using the upper limit constraint expression in Equation (4). becomes possible.

Utilizing these properties, the allowable component value range (component value characteristics in Figure 4a) is set as the initial generation range, and the new range is narrowed using the lower limit constraint equation (3). generate a new range, narrow the range using the upper limit constraint formula in equation (4), generate a new range, and repeat until the generated range converges to a constant value. By doing so, it is possible to obtain a component value candidate generation range in which a range that has no possibility of satisfying the design conditions is removed from the allowable component value range.

FIG. 5 is a flowchart showing the processing based on the above explanation. Referring to Figure 5, first step 3
In step 1, a design conditional expression used for range limitation is set, and in step 32, the allowable component value range is set as the initial value of the component value candidate generation range. however.

For elements that are not included in the component system of the design conditions, the only component value candidates are zero values, so the initial value of the generation range is set to zero to zero. Next, old memory of the generation range 33. Narrowing the range using the lower limit constraint equation 34. Narrowing the range by upper limit constraint equation 35. The processes of and convergence determination 36 are repeatedly executed until the generation range converges, and finally the generation range determination 37 is performed. Here, narrowing the range using the lower limit constraint equation 3
4, the reason why the upper limit value of the allowable component value range is included in the update formula for the lower limit value of the component value candidate generation range is to prevent the divergence of repeated processing that occurs when there is no component value range that satisfies the design condition formula. It's for a reason.

In range narrowing using upper limit value constraint expression 35, the lower limit value of the allowable component value range is included in the update expression for the upper limit value of the component value candidate generation range for the same reason as 6 or more, and the idea of range limitation in step 3. has been described for the case where all the design conditional expressions used for range limitation are linear and their coefficients are not negative, but range limitation is possible even when negative coefficients are included.

Furthermore, even if non-linear design conditional expressions are included, range limitation is possible as long as each component is a monotone function.

4: In this process, component value candidates for each additive element are generated within the component value range limited in the previous step 3. For example, let the component value of carbon be X, and the component value range of X is 1
If 0≦X≦20, and the numerical precision is 2,
Six values of 10, 12, 14, 16°18 and 20 are generated as component value candidates for X. However, in reality, for example [in this size and standard, the value of X is usually 14
The expert's knowledge is registered on the knowledge base in the form of "Set in the vicinity of ."Based on this, the priority of component value candidates is determined, and component value candidates are generated in descending order of priority. For example, when 14 has the highest priority,
X is generated in the order of 14, 16, 12 degrees, 18 degrees, 10 degrees, and 20 degrees.

The contents of this process are shown in the flowchart of FIG. 6, so please refer to it.

5: For each of the component value candidates generated in processing step 4, their combinations are generated step by step according to the combination steps indicated by the component system bundle corresponding to the component system of the design conditions, and these are considered as composition candidates. do. The component system bundle here means a component system and a partial component system generated for it organized by their content relationships. For example, the component system of design conditions is C-S i -M n -V. Then, the component system bundle becomes as shown in Figure 2. In the example of FIG. 2, C, Si,
Mn.

First, C-8t is obtained using component value candidates for .

Generate a combination of C-M n ,..., then C-3i
-Mn, C-3i -V, - is generated. Proceed with the process step by step.

The contents of this process are shown in FIG. 7, so please refer to it.

6: In this process, each composition candidate generated in the process of step 5 is evaluated. The actual processing details are as follows.

(1) For each composition candidate, obtain individual evaluation values for each evaluation scale.

(2) For each composition candidate, calculate the sum of the individual evaluation values obtained in (1) above, and use this as the overall evaluation value for that composition candidate. Examples of evaluation scales include the following.

(a) Alloy extra: Alloy extra is calculated using a predetermined calculation method, and composition candidates with a small amount are given priority.

(b) Upper and lower limit range of intensity Ceq: Find the upper and lower limit range values between the intensity Ceq value when using the upper limit value of each component of the composition candidate and that when using the lower limit value of each component, and calculate whether this upper and lower limit range value is Priority is given to those close to the predetermined value.

Evaluation formula = 1 to the absolute value of K I
Average intensity Ceq value: Find the average value of the intensity Ceq value when using the upper limit value of each component of the composition candidate and that when using the lower limit value of each component,
Priority is given to the average value located near the center of the intensity Ceq range constraint of the design conditions.

Evaluation formula = 2 for the absolute value of K 2 is used to select the best composition candidate as the design result.

Please refer to the flowchart of FIG. 9 for details of the above processing step 6.7.

8: This process checks whether or not the composition candidate satisfies the content of the conditional expression to be satisfied, and if it does not satisfy the content, the generated composition candidate is deleted from the candidate storage device 9. . This process is performed in parallel (substantially simultaneously) with process step 5, so that deletion is performed immediately after it is generated. The outline of the conditional expression is as follows.

(1) Customer/standard specification conditional expression This is the design conditional expression determined in step 2 above, and the fourth
(a) Tensile characteristic equation (strength Ceq range constraint equation L
(b) Component Ceq characteristic equation, (C) Component value characteristic equation.

(d) Includes total element characteristic equations.

(2) Process constraint expression This is the design condition expression determined in step 2 above, and the fourth
(a) Component value upper and lower limit range constraint equation shown in figure a.

(b) Intensity Ceq upper and lower limit range constraint expressions are included.

(3) Conditional expression based on expert knowledge This is a conditional expression based on expert knowledge regarding 1-composition design, and is used as appropriate depending on the product specifications to be designed. For example, there is the following conditional expression.

(a) If the product specifications specify shock value guarantee 20 ≦
Cr5ax However, C: C component value candidate value Cmax: Predetermined upper limit value according to product specifications (b) If the product specifications have shock value guarantee specification: Mn/CiHMNC+
win However, C: the value of the component value candidate of C: Mn: the value of the component value candidate of Mn: MNCmin: the predetermined lower limit value 10 according to Mr. NCm:
This process is executed in parallel (substantially simultaneously) with the component value candidate generation process 4 and the composition candidate generation process 5, and monitors the candidate generation status on the candidate storage device 19 to determine the number of candidates. When the predetermined upper limit is reached, the component value candidate generation process 4 or the composition candidate generation process 5 is stopped. This makes it possible to avoid problems such as insufficient capacity in the candidate storage device 9 when the number of candidates satisfying the design conditions is enormous.

In the candidate generation stop process 10, even if the component value candidate generation process 4 or the composition candidate generation process 5 is stopped midway, in this embodiment, the probability calculation is performed based on the knowledge data as described above. Since candidates with high scores are preferentially generated, there is no problem in which the process ends without generating candidates that should have been highly evaluated in the composition candidate evaluation process 6.

[Effects of the Invention] As described above, if the method of the present invention is applied to an expert system or the like, compositional design of steel products can be automated. Moreover, the optimum composition can be efficiently extracted.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an outline of the processing procedure in FIG. 3, and FIG. 2 is a block diagram showing an example of a component bundle. FIG. 3 is a block diagram showing the configuration of an apparatus implementing the present invention. FIG. 4a is a block diagram showing an example of design constraints and design results, and FIG. 4b is a block diagram showing an outline of the flow of processing from component value candidate generation to evaluation. Figures 5, 6, 7, 8, and 9 are, respectively,
2 is a flowchart showing in detail a portion of the process of FIG. 1; 21: Computer body 22: Knowledge base 23: Database 25: Inference engine 26 Two-person device
27: Output device 1st diagram

Claims (1)

[Claims] Based on the input product specifications, determine the design conditions, including at least the component system, allowable component value range, Ceq characteristics, and total element characteristics corresponding to the composition of the additive elements, and consider the component values of each additive element. The target component value range is limited to a narrower range than the allowable component value range based on the Ceq characteristics and total element characteristics, and the component value candidates for each additive element are within the limited component value range. and generate composition candidates by combining each component value candidate in stages according to the combination conditions indicated by the component system bundle corresponding to the component system of the design conditions, and delete composition candidates that do not satisfy the design conditions, A method for designing a composition of a steel product, wherein a final composition candidate is selected from among the composition candidates based on a result of evaluating the composition candidates under predetermined evaluation conditions.