CN107292007A - A kind of product Design Decision Making based on effectiveness supports system and method - Google Patents

Abstract

The invention provides a utility-based product design decision support system and method. The decision support system includes a knowledge base module, a single utility function module, a multiple utility function module and an overall expected utility module, and each module has an independent input and output interface. . The evaluation index of the alternative design scheme of the present invention is an interval value, which has strong objectivity and can adapt to the uncertainty of the index value in the product design process; adopting the method of the decision support system can adapt to the dynamic change of product design requirements, when the external environment design When the demand changes, it can be processed in the corresponding module through the input and output interface according to the parameters affected by the change, which avoids recalculation, improves the effectiveness of the design scheme, and has high evaluation efficiency.

Description

A utility-based product design decision support system and method

technical field

The invention relates to a product design scheme evaluation method, in particular to a utility-based product design decision support system and method.

Background technique

The intelligent design system provides the possibility for the automation and intelligence of product design and development, and the evaluation and optimization of product design schemes are an important part of it. Designers input design requirements for the intelligent design system, and the system will generate multiple alternatives through its reasoning module. However, designers often only want to obtain the only comprehensive and optimal solution for subsequent simulation analysis and actual production. Therefore, a scheme evaluation method is needed to realize the evaluation and ranking of multiple alternative schemes to obtain the optimal scheme.

The scheme evaluation method is actually a decision-making process of product design scheme. The more popular design decision-making methods in the past, such as the program evaluation method (FDM-GRA) based on fuzzy decision-making diagram (FDM) and gray relational analysis (GRA), multi-granularity attribute index fuzzy measurement and additive Choquet integral model method, network analysis method , the design decision-making method of multi-level attribute index, the optimal decision-making method of design scheme considering the evaluation index weight and gray relational analysis. The above research attempts to apply the decision-making method to select the optimal product design scheme, focusing on the combination of the weight of each design attribute. The evaluation method of the design scheme relies on expert experience to score and evaluate, the scheme effectiveness is low, and it also ignores the The influence of uncertainty in the design scheme.

At present, in order to solve the uncertainties in the firearms design process, Monte Carlo simulation and interval number theory, robust optimization design based on asymptotic global agent model, and fuzzy classification comprehensive evaluation model of product schemes are mainly used.

The existing research results have achieved decision support for the product design process and the expression of design parameter uncertainty from different perspectives, but there is still a lack of a method that fully considers the uncertainty of design parameters and the dynamic changes in design requirements in the product design process. The decision support method of design uncertainty affects the validity of product design scheme.

Contents of the invention

In view of this, the present invention provides a utility-based product design decision support system and method, which can adapt to the design uncertainty in the product design process, improve the effectiveness of the design scheme, and have high evaluation efficiency.

A utility-based product design decision support system, the decision support system includes a knowledge base module, a single utility function module, a multi-utility function module and an overall expected utility module, and each module has an independent input and output interface;

The knowledge base module is used to store previous design schemes and evaluation indicators, and select alternatives according to n evaluation indicators related to product requirements;

The single utility function module receives the designer's preference attribute value for n evaluation indicators based on the utility theory input through the input and output interface, fits n single utility functions according to the preference attribute value, and outputs the corresponding function coefficient through the input and output interface value to the control panel, extract the alternatives from the knowledge base module, use the evaluation index interval of the alternatives as the integral interval, and use the single utility function to calculate n single utility function values of each alternative, and output them to the overall Expected utility module, n is a positive integer;

The multi-utility function module uses the attribute values of n-1 evaluation indicators input by the input and output interface to construct a non-preference attribute combination equation group, and outputs the weight ki of each evaluation index to the control panel, _i =1～n;

The overall expected utility value and ranking of each alternative are output on the control panel based on the weight k _i and the single utility function value by the overall expected utility module;

The decision support system further includes a post-processing analysis module, which is used to test the robustness of the top two alternatives in the overall expected utility value, and change the Δk, Δx, and ΔL of the alternatives through the input and output interfaces to solve again The overall expected utility value of the first two alternative design schemes is used to determine whether the ranking has changed. Δk refers to the change in weight, and Δk is added to _ki in the multi-utility function module, and the new weight is calculated and output again; Δx Refers to the change of the curvature of the single utility function, changing the fitting function coefficient value in the single utility function module, and recalculating the value of the single utility function correspondingly; The evaluation index range of the selected scheme, and the value of the single utility function is calculated again accordingly;

Increase the alternatives in the knowledge base module through the input and output interfaces, and change the ranking of all alternatives; increase the evaluation indicators in the knowledge base module through the input and output interfaces, add a single utility function in the single utility function module, and change each The weight _ki of an evaluation index and the ranking of all alternatives; changing the evaluation index interval of the alternatives through the input and output interface, and changing the single utility function value of the alternatives in the single utility function module.

Further, using the utility-based product design decision support system as claimed in claim 1, the decision support method is:

Step 1, select alternatives in the knowledge base module according to n evaluation indicators related to product requirements;

Step 2: Based on the utility theory, the designer inputs the preference attribute values of n evaluation indicators through the input and output interfaces in the single utility function module, fits n single utility functions according to the preference attribute values, and outputs the corresponding function coefficient values to the controller on the panel;

Step 3, extract the alternatives from the knowledge base module, use the evaluation index interval of the alternatives as the integral interval, use the single utility function to calculate and output n single utility function values of each alternative, where n is a positive integer; And input the attribute values of n-1 evaluation indicators through the input and output interface to construct a non-preference attribute combination equation group, and output the weight k _i of each evaluation index to the control panel, i=1~n;

Step 4, output the overall expected utility value and ranking of each alternative to the control panel based on the weight k _i and the single utility function value;

Step 5: Change Δk, Δx or ΔL of the top two alternatives, and recalculate the overall expected utility value and ranking of the alternatives to determine whether the rankings of the top two alternatives have changed.

Further, when it is necessary to introduce new alternatives, after step four, it further includes adding new alternatives and corresponding evaluation indicators in the knowledge base module described in step one, and the decision support system calculates the value of the new alternatives. Overall expected utility value and rerank all alternatives.

Further, when a new evaluation index needs to be introduced, after step four, it further includes adding a new evaluation index in the knowledge base module described in step one, and the designer inputs the value of the new evaluation index in the single utility function module based on the utility theory. Attribute preference value, fit the corresponding single utility function according to the attribute preference value, and output the corresponding function coefficient value; respectively input the upper and lower boundaries of the new evaluation index interval that synthesizes all alternatives, and use each alternative in the knowledge base module The evaluation index interval and single utility function of the scheme, and output all the single utility function values; recalculate the weight and output the overall expected utility value and ranking of each alternative based on the weight and single utility function value.

Further, when the interval of the evaluation index of the alternative is changed, the interval of the evaluation index is changed in the multi-utility function module of step three and the value of the single utility function is recalculated.

Further, in the second step, the attribute preference data of the n evaluation indicators obey the uniform distribution.

Beneficial effect:

1. The decision support system and method of the present invention obtain the overall expected utility value of each design scheme based on the utility theory, and the effectiveness of the design scheme is high; the evaluation index of the alternative design scheme is an interval value, which has strong objectivity and can adapt to product Uncertainty of index values in the design process; the present invention modularizes the solution process, and when the design requirements of the external environment change, it can be processed in the corresponding module through the input and output interfaces according to the parameters affected by the change, avoiding recalculation and improving evaluation efficiency.

2. The post-processing analysis module of the present invention tests the robustness of the design scheme by changing the design requirements and solving it again, which further improves the effectiveness of the design scheme.

3. Each module included in the decision support system of the present invention has an independent input and output interface on the control panel, so that it can be quickly reused in similar designs or new designs, has strong adaptability, and improves design efficiency.

Description of drawings

Fig. 1 is the flowchart of decision support of the present invention;

Fig. 2 is a block diagram of modularized flexible formwork design of the present invention;

Fig. 3 is the scheme knowledge base panel figure of the present invention;

Fig. 4 is attribute knowledge base of the present invention and single utility function module panel figure;

Fig. 5 is overall expected utility module panel diagram of the present invention;

Fig. 6 is a diagram of the post-processing analysis module of the present invention.

detailed description

The present invention will be described in detail below with reference to the accompanying drawings and examples.

The invention provides a utility-based product design decision support system. The decision support system includes a knowledge base module, a single utility function module, a multi-utility function module and an overall expected utility module. Each module has an independent input and output interface. The control panel is presented accordingly in a modular manner. As shown in Figure 2, the design process of modular flexible formwork mainly consists of five steps, including: template creation, template customization, design of a single module, overall template design, and example verification. First analyze the design requirements and functions of each part of the entire design decision-making process, split the decision-making process into several complete and independent modules with standardized input and output interfaces, and split the entire decision-making process based on utility theory into scheme knowledge base, Attribute knowledge base, single utility function module, multi-utility function module, overall expected utility module, post-processing analysis module, in the process of module design, parameters need to be extracted, and data processing is performed, and finally the module is instantiated.

The knowledge base module includes a scheme knowledge base and an attribute knowledge base. The scheme knowledge base is used to store the design scheme of the product, and the attribute knowledge base is used to store the corresponding evaluation indicators. The contents of the scheme knowledge base and the attribute knowledge base are related to each other. Alternatives are selected according to n evaluation indicators related to product demand.

The single utility function module receives the attribute preference values of n evaluation indicators input by the designer based on the utility theory through the input and output interface, and fits n single utility functions according to the attribute preference values, and outputs the corresponding function coefficient values to On the control panel, extract the alternatives from the knowledge base module, use the evaluation index interval of the alternatives as the integral interval, and use the single utility function to calculate n single utility function values for each alternative, and output them to the overall expected utility module, n is a positive integer.

The multi-utility function module uses the attribute values of n-1 evaluation indicators input through the input and output interface to construct a non-preference combination equation group, and outputs the weight k _i of each evaluation index to the control panel, where i=1~n.

The overall expected utility module outputs the overall expected utility value and ranking of each alternative to the control panel based on the weight _ki and the single utility function value.

The decision support system further includes a post-processing analysis module, which is used to test the robustness of the top two alternatives in the overall expected utility value, and change the Δk, Δx, and ΔL of the alternatives through the input and output interfaces to re-solve the The overall expected utility value of the first two alternative design schemes is used to judge whether the ranking has changed. Δk refers to the change of weight, and Δk is added to _ki in the multi-utility function module, and the new weight is calculated again; Δx refers to the single The change of the curvature of the utility function changes the fitting function coefficient value in the single utility function module, and recalculates the value of the single utility function correspondingly; The interval of the evaluation index, correspondingly recalculate the value of the single utility function.

Increase the alternatives in the knowledge base module through the input and output interfaces, and change the ranking of all alternatives; increase the evaluation indicators in the knowledge base module through the input and output interfaces, add a single utility function in the single utility function module, and change each The weight _ki of an evaluation index and the ranking of all alternatives; changing the evaluation index interval of the alternatives through the input and output interface, and changing the single utility function value of the alternatives in the single utility function module.

The specific method is shown in Figure 1:

Step A, STEP A1: Pre-construct the product design scheme knowledge base, which stores previous design schemes;

STEP A2: Build an attribute knowledge base, store evaluation indicators, and provide knowledge support for design decision-making. The sources include experimental data, design manuals and other materials, as well as attribute characteristics reflected in the actual design process.

STEP A3: Select alternatives according to n evaluation indicators related to product demand.

Step B,

STEP B1: Designers answer the lottery questions (lottery questions) and enter them into the single utility function module to determine the five different levels of preference attribute values for the alternatives, which are utility U=(0, 0.25, 0.5, 0.75, 1) attribute value at time.

STEP B2: Fit n single utility functions according to the preference attribute values, and output the corresponding function coefficient values; divide the attributes of product alternatives into small-looking attributes, big-looking attributes, and big-looking attributes, and give single-utility functions respectively The fitting process:

1) Look at the big attributes. The expected large attribute represents the attribute that the utility of an attribute in the design scheme increases monotonically with the increase of the attribute value. For the Wangda attribute, first determine the attribute values of five different levels of preference And x ₁ , through curve fitting into an exponential utility function, the default utility function type in this paper is u(x):

In the formula: the values of a _L , b _L , c _L and d _L respectively correspond to the coefficients of the exponential utility after curve fitting. 2) Looking at small attributes. The expected small attribute represents the attribute that the utility of an attribute in the firearm design scheme decreases monotonically with the increase of the attribute value. For the Wangxiao attribute, first determine the attribute values of five different levels of preference As well as x ₁ , as shown in Figure 3, an exponential utility function u(x) is formed by curve fitting:

In the formula: the values of a _R , b _R , c _R and d _R respectively correspond to the coefficients of the exponential utility after curve fitting.

3) Look at the property of the eye. The objective attribute refers to the attribute whose utility is at or close to the target attribute value, the better it is. For wangmu attributes based on specific target values, the input to the curve fit includes the left and right attribute values. For attributes with the goal of attribute maximization, the attributes to the left of the target value are required as input, while for attributes with the goal of attribute minimization, the attributes to the right are required as input. First determine the attribute values of 9 different levels of preference And x ₁ , through curve fitting into an exponential utility function u(x):

In the formula: the values of a, b, c and d respectively correspond to the coefficients in the exponential utility function u(x)=a+bx+ce ^dx after the curve fitting, and the subscript indicates that it is on the left side of the target value or Right side, L means left, R means right.

STEP B3: Input the upper and lower bounds of the same evaluation index range of all alternatives through the input and output interface, and use the evaluation index interval and single utility function of each alternative in the knowledge base module to output the value of each alternative n single utility function values, n is a positive integer.

Using the "Wangmu" single-attribute utility function, and assuming that the value of the specific option attribute obeys a uniform distribution, the probability density function is:

f(x)＝1/(x _u -x _l ) (4)

In the formula: x _l , x _u are the upper and lower boundaries of the same evaluation index interval for all alternatives.

According to the fitted single utility function (1) (2) (3) and formula (4) to solve the expected utility value of a single attribute, the calculation formula of the expected utility value of a single attribute is as follows:

E(u)＝∫u(x)·f(x)dx (5)

In the formula: u(x) represents the single utility function formula, and f(x) represents the attribute value probability density function. Substituting the evaluation index interval of each alternative to output n single utility function values respectively.

Step C: Construct the non-preference attribute combination equations by inputting the attribute values of n-1 evaluation indexes, and output the weights of the evaluation indexes to the control panel;

STEP C1: The weight k _i is determined by solving a system of n linear equations. Based on the superimposed utility function form and the non-preference attribute combination equations, multiple linear equations with the same utility are established, and the equations composed of these equations are used to solve the weight k _i .

STEP C2: Combining multi-utility non-preference attributes to form a system of equations. There are n-1 equations in the equation system established by the designer through the combination of non-preferred attribute values, and the remaining one represents that the sum of proportional constants is 1,

The specific multi-utility function construction process is shown in formula (6).

In the formula: as well as Indicates the value of attribute i at different levels. Represents the worst level (its utility is 0); Represents except other than two different levels, they satisfy (E.g The corresponding utility is 0.55 and has a utility of 0.45); is the attribute value specified by the decision maker, which must make the utility of two different multi-attribute value combinations equal, so that the decision maker has no preference for these two combinations, that is,

STEP C3: Enter n-1 attributes Output weight k _i .

Step D: Output the overall expected utility value and ranking of each alternative based on the weight k _i and the single utility function value and store them in the multi-utility function module.

STEP D1: Using the single-attribute expected utility value formula (5) to calculate the overall expected utility value, so that the designer can quickly convert it into the overall expected utility of the option after certain utility function parameters and option attribute parameters are given.

In the formula: k _i represents the proportional constant of attribute i, and E(u _i (A _i )) represents the expected utility value of attribute i.

Use the formula (8) to calculate the overall expected utility value of each scheme and sort the overall expected utility value of each scheme to provide a basis for the design decision of the scheme. The option with the highest overall expected utility value is called the "most valuable" option, and the option with the second highest overall expected utility value is called the "second" option.

The decision support system further includes a post-processing analysis module, which is used to test the robustness of the top two design schemes with overall expected utility values. Select the top two options: the "most valuable" solution and the "second" solution, and then test the overall expected utility of these two options for the solutions, attributes and design parameters caused by changes in certain design requirements or program innovations, etc. Whether the dynamically changing response is enough to affect its raw ranking. It can be seen from the calculation process of expected utility that possible parameter changes can only occur in two formulas, namely formula (5) and formula (8). The parameter changes in these two formulas are represented by Δk, Δx, and ΔL, respectively. Among them, Δk refers to the change of weight, and Δk is added to _ki in the multi-utility function module respectively, and the new weight is calculated and output again; Δx refers to the change of the curvature of the single-utility function, and the value of the function coefficient fitted in the single-utility function module is changed. Correspondingly recalculate the value of the single utility function; ΔL refers to the length change of the evaluation index interval of the alternative, change the evaluation index interval of the alternative in the single utility function module, and recalculate the value of the single utility function accordingly.

The specific implementation method of post-processing analysis is to increase or decrease Δk, Δx and ΔL. In this paper, the default change values of Δk, Δx and ΔL are 5%, and then recalculate the E(u) and E(U) of the top two options, and analyze the changes of these two options E(U) in post-processing Displayed in the control panel of the module.

Furthermore, when the design requirements of the external environment change, it can be processed in the corresponding module according to the parameters affected by the change,

(1) Changes in the number of programs

After step D, it further includes adding new alternatives and corresponding evaluation indicators in the knowledge base module of step A, and the decision support system calculates the overall expected utility value of the new alternatives and re-ranks all alternatives.

(2) Changes in program evaluation indicators

After step D, it further includes adding a new evaluation index in the knowledge base module of step A. Based on the utility theory, the designer inputs the attribute preference data for the new evaluation index in the single utility function module, and fits the corresponding Single utility function, output the corresponding function coefficient value; respectively input the upper and lower boundaries of the new evaluation index interval of all alternatives, use the evaluation index interval and single utility function of each alternative in the knowledge base module, output all Single utility function value; recalculates weights and outputs the overall expected utility value and rank for each alternative based on weights and single utility function value.

(3) Changes in fixed parameters

The change of fixed parameters means that the value of some parameters (these parameters are fixed parameters in the original decision) has changed in the new decision. For example, in a new decision-making, due to further experiments, the upper and lower bounds of a certain scheme attribute value (such as the strength of the material) have changed and become more accurate; or the importance of a certain attribute (such as wear resistance) has changed. Changes need to be prioritized. At this time, in the multi-utility function module of step C, the interval of the evaluation index is changed and the value of the single utility function is recalculated.

The following examples are used to illustrate, according to the above-mentioned utility-based selection decision-making theory, the above-mentioned method is verified by taking the gun barrel as an example, and a decision-making support system for firearm design considering uncertainty is designed.

The barrel is one of the most basic parts of firearms products, and its main function is to give the bullet a certain direction and initial velocity. The barrel design is in the design stage of the gun body plan, and most of the design parameters have been determined in the previous internal and external ballistic design and bullet design process.

Table 1 shows the input tactical technical requirements indicators. Due to the high temperature of the gunpowder, the effect of high-pressure gunpowder gas, and the mechanical friction with the warhead when the gun barrel is shooting, in order to achieve the above goals, the barrel material is required to have a higher tensile strength (not less than 50 kg/ ^mm2 ), sufficient impact toughness (not less than 5 kg-m/ ^cm2 ), high yield point (not less than 50 kg/ ^mm2 ), good machinability and sufficient wear resistance and ablation resistance. At present, the barrels from heavy machine guns to pistols generally use 50BA or 50AE steel. Anti-aircraft machine guns and some heavy machine guns use 30SiMnMoVA or 30CrNi ₂ MoVA steel, and other materials are also used as barrels. However, the choice of barrel materials mostly depends on Designer experience design. According to these design requirements, six evaluation indicators are selected from the attribute knowledge base, and four alternatives are selected from the program knowledge base according to the evaluation indicators, as shown in Table 2. Among the attributes of the six evaluation indexes, three are quantitative attributes, namely, tensile strength, yield point, and impact value, and the other three are qualitative attributes, namely, processability, wear resistance, and ablation resistance. The designer's expectations for these properties include both "looking at" properties, such as the expected target value of tensile strength is 95 kg/ ^mm2 , and non-"looking at" properties, such as processing performance, the higher the value, the better . The value of all attributes of each option is uncertain, and it is a range determined by the upper and lower bounds. For example, the value range of tensile strength of option 50BA is 44-69 kg/ ^mm2 . Assume that all attribute values are subject to uniform distribution. Designers need to choose from the alternatives the option that best meets the aforementioned goals.

Table 1 Input Tactical Technical Requirements Indicators

Table 2 Barrel Material Alternatives

As shown in Figure 3, it shows the program knowledge base in the computer support system. When there are new materials or new processing techniques that need to be compared to assist decision makers in their selection, they can be quickly configured on the control panel.

The control panel shown in Figure 4 is the attribute knowledge base of each alternative scheme and the calculation module diagram of the single utility function, which effectively supports the solution of the single utility function of each attribute on the panel. The control panel shown in Figure 5 is composed of a series of modules such as multi-attribute utility value calculation module, scheme overall expected utility calculation module and sensitivity analysis module.

The operation steps and process of the decision support system for the overall firearms design scheme are as follows:

Step 1: Select evaluation indicators and screen out alternatives according to the design requirements of rapid barrel design. Table 2 shows the schemes selected based on the above tactical and technical demand indicators, and the number of candidate schemes is n=4.

Step 3: Identify decision maker preferences. Table 3 shows the preference attribute values of different levels of firearm design schemes obtained by decision makers by answering the gambling questions.

Step 4: Fit a single utility function. Table 4 is the coefficient values of a, b, c, and d obtained after fitting the attribute values in Table 3 into the exponential utility function of the attribute.

Step 5: Find the expected utility value of the single utility function. Table 5 is the value of each single utility function of the firearm design alternatives calculated by combining the exponential utility function obtained in Table 4 and the evaluation index intervals of the alternatives.

Step 6: Determine the weights. Table 6 shows the weight of the single utility function value obtained through the superimposed utility function and the non-preference combination formula, that is, the weight of the evaluation index.

Step 7: Solve for the overall expected utility. Table 7 combines the expected utility values of single attributes in Table 5 and the weights in Table 6, according to the formula The obtained overall expected utility value of each firearm design scheme. Thus, the "most valuable" firearm design scheme is 30SiMnMoVA, and its heat treatment process is: 870°C high temperature quenching/650°C high temperature tempering.

Table 3 Attribute values of utility at different levels

Table 4 Coefficient table of single utility function of each attribute

Table 5 Expected utility value of each evaluation index

Table 6 Weight

Table 7 The overall expected utility of each program

Further, when it is necessary to introduce new alternatives, such as this is a high thermal conductivity lightweight barrel made of a new type of composite material, where for the steel inner tube, carbon fiber reinforced epoxy resin is wound on the outside, The surface of this carbon fiber-reinforced epoxy resin is coated with a layer of metal nickel; while the steel outer tube is made of carbon fiber-reinforced composite material with high specific strength and high specific modulus mechanical properties, which can effectively improve the shooting accuracy of the barrel while ensuring light weight , but its processing performance is relatively poor. The details of the six attributes of this new alternative are shown in Table 8. Due to the introduction of the new scheme, the original four-alternative-plan selection problem has been transformed into five-alternative-plan selection problem, and the designer needs to re-model the problem and make a choice. Since most of the decision-making processes such as attribute description, single-attribute utility function, and multi-attribute utility function are archived in the original configurable modular template instance, new decisions only need to make some necessary modifications on the basis of the original instance That's it. It is necessary to add the new alternatives described in Table 8 to the original template instance, and then update the ranking of the alternatives, the specific steps are as follows:

1) Basic information of the specified scheme: input information in the knowledge base module to instantiate a new alternative scheme;

2) Specify the attribute range of the scheme: input the n evaluation index intervals of the new alternative scheme and the distribution obeyed through the input and output interface.

3) Updating proposal rankings: When all necessary input information is set, the rankings of all proposals will be updated automatically. The new ranking shows that the program "New Composite Materials" has an expected utility of 1.635387, ranking first.

Table 8 Property information of new composite materials

Furthermore, when a new evaluation index needs to be introduced, assuming a new attribute—elongation rate, it needs to be considered in the decision-making problems of the aforementioned five alternatives. The reason why elongation is introduced is that it characterizes the ratio of the stretched length to the original length of the gun material when it is broken under tension, which is very important for the quality of the barrel material. The information related to this attribute configuration is divided into four aspects, as shown in Table 9:

(1) Add new evaluation indicators in the knowledge base module;

(2) To establish the different levels of attribute preference values required by the elongation utility function, including the preferences on the left and right sides of the target value;

(3) Determine the input information required for the multi-attribute utility function weight _ki , that is, the values of n-1 evaluation indicators in the non-preferred multi-attribute value combination;

(4) The value range of the elongation of the five alternatives. After introducing new attributes, the six-attribute selection decision problem is transformed into a seven-attribute selection decision problem, and designers need to consider the elongation information in Table 9 and remodel the problem to make new decisions. Most of the knowledge in previous examples can be reused, just make some necessary modifications. The modification required is to instantiate a new attribute and re-rank the five alternatives, as follows:

1) specify the basic information of the attribute: enter the relevant information of the elongation rate in the attribute knowledge base of the knowledge base module to instantiate a new attribute;

2) Input the attribute preference values of different levels of elongation;

3) Configure multi-attribute utility function weight evaluation related information: specify the attribute value of each attribute in the non-preference attribute combination equation group.

4) Configure the value range of the elongation of each scheme: the upper and lower bounds of the elongation of all alternative schemes are combined as the boundary for solving the probability density function.

5) Recalculate the weight and output the overall expected utility value and ranking of each alternative based on the weight and single utility function value: the new ranking shows that the top two schemes are "new composite material" and 30SiMnMoVA respectively, corresponding to The expected utilities are 1.87532 and 1.78562, respectively.

Table 9 Configuration information of attribute elongation

The decision support system further includes a post-processing analysis module, which considers some possible parameter changes and tests the robustness of the No. 1 solution "New Composite Material" in the case of parameter changes, so as to enhance the confidence of designers in choosing it for firearm design. The expected utilities of the first-ranked scheme "new composite material" and the "second" scheme 30SiMnMoVA are very close, being 1.87532 and 1.78562 respectively, with a difference of only 0.0897. It will be tested in the post-processing analysis module whether the ranking of the two will change due to changes in some parameters, that is, whether the ranking is sensitive to parameter changes.

The present invention will take ΔL as an example to illustrate the post-processing analysis process. The original value range of the tensile strength of the scheme "new composite material" is 90-100 kg/mm ² , by gradually narrowing this range (that is, gradually reducing the ambiguity of the design parameters) to test the response of the overall expected utility of the scheme and the The ranking comparison between the scheme and the scheme 30SiMnMoVA is shown in Figure 6. In the figure, the degree of change is set to 5%, and the number of changes is set to 7 times. From the automatically generated image, it can be seen that when the value range of tensile strength is reduced to 80% of the original value, the expected utility value of the scheme 30SiMnMoVA rises from the original 1.78562 to 1.82782; The expected utility value of 30SiMnMoVA is still maintained at the level of 1.82782 as shown in Figure 6. From this, it can be concluded that the scheme ranking is not sensitive to the reduction of the range of tensile strength values, and it is safe to choose the first-ranked scheme "new composite material".

To sum up, it can be obtained from Figure 6 that the "most valuable" solution for the material requirements based on this tactical technical requirement is "new composite materials". Based on the above facts, it can be seen that the designer can reduce the time for referring to a large amount of knowledge, save a lot of design iteration process, and enhance the effectiveness of decision-making when the designer conducts subsequent design work on the basis of selecting the material. All in all, the decision support system for firearms design scheme considering uncertainty provides a relatively flexible scheme design method for designers. Designers can choose the "most valuable" design scheme based on tactical and technical requirements and the designer's own preferences. According to the characteristics of the firearms design process, the introduction of configurable modular templates makes the modules easier to modify and reduces data redundancy.

To sum up, the above are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (6)

1. A utility-based product design decision support system, characterized in that, the decision support system includes a knowledge base module, a single utility function module, a multiutility function module and an overall expected utility module, and each module has independent input Output Interface; The knowledge base module is used to store previous design schemes and evaluation indicators, and select alternatives according to n evaluation indicators related to product requirements; The single utility function module receives the designer's preference attribute value for n evaluation indicators based on the utility theory input through the input and output interface, fits n single utility functions according to the preference attribute value, and outputs the corresponding function coefficient through the input and output interface value to the control panel, extract the alternatives from the knowledge base module, use the evaluation index interval of the alternatives as the integral interval, and use the single utility function to calculate n single utility function values of each alternative, and output them to the overall Expected utility module, n is a positive integer; The multi-utility function module uses the attribute values of n-1 evaluation indicators input by the input and output interface to construct a non-preference attribute combination equation group, and outputs the weight ki of each evaluation index to the control panel, _i =1～n; The overall expected utility value and ranking of each alternative are output on the control panel based on the weight k _i and the single utility function value by the overall expected utility module; The decision support system further includes a post-processing analysis module, which is used to test the robustness of the top two alternatives in the overall expected utility value, and change the Δk, Δx, and ΔL of the alternatives through the input and output interfaces to solve again The overall expected utility value of the first two alternative design schemes is used to determine whether the ranking has changed. Δk refers to the change in weight, and Δk is added to _ki in the multi-utility function module, and the new weight is calculated and output again; Δx Refers to the change of the curvature of the single utility function, changing the fitting function coefficient value in the single utility function module, and recalculating the value of the single utility function correspondingly; The evaluation index range of the selected scheme, and the value of the single utility function is calculated again accordingly; Increase the alternatives in the knowledge base module through the input and output interfaces, and change the ranking of all alternatives; increase the evaluation indicators in the knowledge base module through the input and output interfaces, add a single utility function in the single utility function module, and change each The weight _ki of an evaluation index and the ranking of all alternatives; changing the evaluation index interval of the alternatives through the input and output interface, and changing the single utility function value of the alternatives in the single utility function module. 2. a kind of product design decision support method based on utility, it is characterized in that, adopt the product design decision support system based on utility as claimed in claim 1, described decision support method is: Step 1, select alternatives in the knowledge base module according to n evaluation indicators related to product requirements; Step 2: Based on the utility theory, the designer inputs the preference attribute values of n evaluation indicators through the input and output interfaces in the single utility function module, fits n single utility functions according to the preference attribute values, and outputs the corresponding function coefficient values to the controller on the panel; Step 3, extract the alternatives from the knowledge base module, use the evaluation index interval of the alternatives as the integral interval, use the single utility function to calculate and output n single utility function values of each alternative, where n is a positive integer; And input the attribute values of n-1 evaluation indicators through the input and output interface to construct a non-preference attribute combination equation group, and output the weight k _i of each evaluation index to the control panel, i=1~n; Step 4, output the overall expected utility value and ranking of each alternative to the control panel based on the weight k _i and the single utility function value; Step 5: Change Δk, Δx or ΔL of the top two alternatives, and recalculate the overall expected utility value and ranking of the alternatives to determine whether the rankings of the top two alternatives have changed. 3. The utility-based product design decision support method as claimed in claim 2, wherein when new alternatives need to be introduced, step 4 further includes adding new alternatives in the knowledge base module described in step 1. The decision support system calculates the overall expected utility value of the new alternatives and re-ranks all the alternatives. 4. The utility-based product design decision support method as claimed in claim 2, wherein when a new evaluation index needs to be introduced, the step 4 further includes adding a new evaluation index in the knowledge base module described in the step 1 , the designer inputs the attribute preference value of the new evaluation index in the single utility function module based on the utility theory, fits the corresponding single utility function according to the attribute preference value, and outputs the corresponding function coefficient value; respectively inputs and synthesizes all alternatives The upper and lower boundaries of the new evaluation index interval, using the evaluation index interval and single utility function of each alternative in the knowledge base module, output all the single utility function values; recalculate the weight and output each The overall expected utility value and ranking of each alternative. 5. The utility-based product design decision support method as claimed in claim 2, characterized in that, when the evaluation index interval of the alternative is changed, the interval of the evaluation index is changed in the multi-utility function module of step 3 and recalculated Single utility function value. 6. The utility-based product design decision support method according to claim 2, characterized in that, in the second step, the attribute preference data of the n evaluation indicators obey a uniform distribution.