CN103149840A - Semanteme service combination method based on dynamic planning - Google Patents

Abstract

The invention provides a semanteme service combination method based on dynamic planning. The semanteme service combination method based on the dynamic planning comprises firstly, decomposing a manufacturing service process model through a workflow engine, extracting requirements of each active node, and then obtaining all services which satisfy the active node through a service find matching process, and at last, selecting an appropriate candidate service from the services to generate a service combined scheme according to a certain strategy, achieving mapping a specific semanteme onto a task, and using a manufacturing service process which is described in a workflow model to translate a generating process of the semanteme service combined scheme into a dynamic evolution multistage decision process by combining characteristics of a networked manufacturing mode, using a data semanteme relationship between services as restraint, using comprehensive quality of the combined scheme as an optimization target, and using a dynamic planning method to generate a combined scheme which is globally optimum. The semanteme service combination method based on the dynamic planning has the advantages of enabling confirmation of a globally optimum solution to be easy, saving computing amount, being beneficial for providing abundant combined scheme results for a user, and improving performability of the combined scheme and automation degree of a combining process.

Description

A kind of semantic service combined method based on dynamic programming

Technical field

The present invention relates to the combination technique field of networking manufacturing service, be specially a kind of semantic service combined method based on dynamic programming.

Background technology

Services Composition is the operation flow that builds in Networked Manufacturing, realizes making resource demand assigned effective way.Existing Services Composition technology can be summarized as two large classes.A kind of method that is based on workflow, with flowing the similar model of modeling method with classical works, composite services are described, realize simple, but require to know in advance function and purpose that in the concrete structure of flow process and flow process, each Operations Requirements is realized, mapping from the workflow business model to service environment can only be carried out static conversion, and the robotization of process flow operation and dynamic adjustment degree are not high.The another kind of artificial intelligence approach that is based on, utilize body to carry out semantic tagger and business procedure modeling to input/output argument, prerequisite and the result etc. of Web service, come to such an extent that realize the Auto-matching of Web service by formal reasoning, the composite sequence that obtains serving, but the complexity of the method is higher, and purely relying on computing machine, automatically to carry out being combined at present of Web service also not yet ripe.

Summary of the invention

The technical matters that solves

In order to overcome the shortcoming of above-mentioned the whole bag of tricks, the automaticity of Services Composition and preferred process under the raising Networked Manufacturing, the present invention combines workflow with the Semantic Web Services technical advantage, a kind of semantic service combined method based on dynamic programming is proposed, accurately, dynamically generate from the manufacturing resource that is encapsulated as semantic service and meet consumers' demand and the assembled scheme of optimal quality, be conducive to the reasonable disposition resource, improve its utilization ratio, have meaning for the application implementation of networking manufacturing mode.

Technical scheme

Not enough for actual effect, the dirigibility of avoiding the assembled scheme generating mode that prior art causes, realize the distribution according to need between manufacturing operation demand and semantic service, the present invention adopts the dynamic binding pattern.Do not bind concrete service in the manufacturing operations process modeling, when carrying out a manufacturing operation, at first by workflow engine, the manufacturing operations procedural model is decomposed, extract the demand of each active node, then cross the service discovery matching process and obtain all services of satisfying this activity, the certain strategy of last basis is selected suitable candidate service from each and is generated the Services Composition scheme, realizes concrete semantic service is mapped on task.

Match because each manufacturing activities has a plurality of candidate service, this chooses optimum semantic service assembled scheme problem with regard to having produced global optimization.Although enumerate calculating can be exhaustive institute might scheme, along with increasing of the expansion of problem scale, movable number or candidate service number, will produce shot array.Therefore, the present invention is converted into a dynamic evolution, multistage decision process with the generation of Services Composition scheme, be about to whole Services Composition process and be divided into the stage that several interknit, need to make the services selection decision-making in each stage, and after the service in a upper stage is determined, can exert an influence to next stage services selection decision-making.Target of the present invention is exactly in the decision region of each stage permission, selects an optimizing decision sequence, is issued to best of breed service quality in predetermined constraint.Therefore, the present invention utilizes dynamic programming method under the prerequisite that keeps data semantic dependence between service, solves the optimum Services Composition scheme of global service quality (Quality of Service, QoS).

Technical scheme of the present invention is:

Described a kind of semantic service combined method based on dynamic programming is characterized in that: adopt following steps to form:

Step 1: the assembled scheme that will include n-1 manufacturing activities is divided into n manufacturing activities node v according to manufacture process _i, i={0 wherein ..., n}, v ₀Represent start node, v _nRepresent end node; According to the service discovery matching process, obtain the candidate service that satisfies its functional requirement corresponding to each manufacturing activities node, wherein candidate service

Expression manufacturing activities node v _iJ corresponding candidate service; The candidate service that satisfies its functional requirement that each manufacturing activities node is corresponding forms the original state set of this manufacturing activities node, wherein for original state set U _i, and

Expression manufacturing activities node v _iM candidate service that satisfies its functional requirement arranged;

Step 2: the data dependence relation between the candidate service of the candidate service of computational manufacturing active node and follow-up adjacent manufacturing activities node, and effective follow-up set of service corresponding to the candidate service of definite manufacturing activities node: wherein for manufacturing activities node v _iJ candidate service

The employing following steps obtain

Effective follow-up set of service:

Step 2.1: get manufacturing activities node v _i+1Corresponding original state set U _i+1In element

Calculate and describe The concept c of output parameter ₁With description

The concept c of input parameter ₂Between semantic similarity Sem (c ₁, c ₂):

$\mathrm{Sem}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2)=\frac{|P\left(c_1\right)\&cap;P\left(c_2\right)|}{|P\left(c_1\right)\&cap;P\left(c_2\right)|+\mathrm{\&alpha;}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2)|P\left(c_1\right)/P\left(c_2\right)|+(1-\mathrm{\&alpha;}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2))|P\left(c_2\right)/P\left(c_1\right)|}$

Wherein in function P (c) expression body to the attribute definition of concept c, ∩ ,/, ‖ is respectively intersection of sets, difference and gesture computing, $\mathrm{\&alpha;}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2)=\left\{\begin{array}{cc}\frac{\mathrm{depth}\left(c_1\right)}{\mathrm{depth}\left(c_1\right)+\mathrm{depth}\left(c_2\right)}& \mathrm{depth}\left(C_1\right)>\mathrm{depth}\left(C_2\right)\\ 1-\frac{\mathrm{depth}\left(c_1\right)}{\mathrm{depth}\left(c_1\right)+\mathrm{depth}\left(c_2\right)}& \mathrm{depth}\left(C_1\right)\&le;\mathrm{depth}\left(C_2\right)\end{array}\right.,$ Function depth (c) calculates concept c residing degree of depth in body;

Step 2.2: the method traversal manufacturing activities node v that adopts step 2.1 _i+1Corresponding original state set U _i+1In all elements, described

The concept of output parameter and description U _i+1In semantic relevancy between the concept of each element input parameter, get U _i+1In the semantic relevancy element that is not less than setting threshold form

Effective follow-up set of service $X_i\left(s_i^{\left(j\right)}\right);$

Step 3: the original quality parameter matrix of setting up each manufacturing activities node: for m candidate service arranged

Manufacturing activities node v _i, v _iOriginal quality parameter matrix Q (v _i) be:

$Q\left(v_i\right)={[q{s}_i^{\left(1\right)},......,q\left(s_i^{\left(m\right)}\right)]}^T={\left[\begin{array}{c}{q}_{i1}^{\left(1\right)},q_{i2}^{\left(1\right)},......,q_{\mathrm{iw}}^{\left(1\right)}\\ q_{i1}^{\left(2\right)},q_{i2}^{\left(2\right)},......,q_{\mathrm{iw}}^{\left(2\right)}\\ ........................\\ q_{i1}^{\left(m\right)},q_{i2}^{\left(m\right)},......,q_{\mathrm{iw}}^{\left(m\right)}\end{array}\right]}_{m\&times;w}$

Wherein

The expression candidate service

The original quality parameter

W represents the mass parameter number;

Step 4: to the original quality parameter matrix normalization of each manufacturing activities node: for manufacturing activities node v _iOriginal quality parameter matrix Q (v _i), get

Be respectively matrix Q (v _i) in maximal value and the minimum value of k row, k=1 ..., w; When k row original quality parameter is the parameter value direct ratio type mass parameter that more high quality-of-service is better, with k row original quality parameter according to

${\mathrm{ib}}_{\mathrm{ik}}^{\left(j\right)}=\left\{\begin{array}{cc}\frac{q_{\mathrm{ik}}^{\left(j\right)}-q_k^{\min }}{q_k^{\max }-q_k^{\min }}& q_k^{\max }\&NotEqual;q_k^{\min }\\ 1& q_k^{\max }=q_k^{\min }\end{array}\right.$

Carry out normalization; When k row original quality parameter is parameter value more during the poorer inverse ratio type mass parameter of high quality-of-service, with k row original quality parameter according to

${\mathrm{db}}_{\mathrm{ik}}^{\left(j\right)}=\left\{\begin{array}{cc}\frac{q_k^{\max }-q_{\mathrm{ik}}^{\left(j\right)}}{q_k^{\max }-q_k^{\min }}& q_k^{\max }\&NotEqual;q_k^{\min }\\ 1& q_k^{\max }=q_k^{\min }\end{array}\right.$

Carry out normalization;

Step 5: the overall quality value of calculating each candidate service of each manufacturing activities node: for manufacturing activities node v _i, according to the mass parameter matrix after normalization, according to

$\mathrm{QoS}\left(s_i^{\left(j\right)}\right)=\frac{\underset{k=1}{\overset{t}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_{k}i{b}_{\mathrm{ik}}^{\left(j\right)}}{\underset{k=1}{\overset{w-t}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_{k}d{b}_{\mathrm{ik}}^{\left(j\right)}}$

Computational manufacturing active node v _iThe overall quality value of each candidate service, ω wherein _k∈ [0,1] and T represents the quantity of direct ratio type mass parameter;

Step 6: with f _i(u _i) be optimization aim, set up Dynamic Programming Equation

$\left\{\begin{array}{c}{f}_i\left(u_i\right)=\underset{s_{i+1}^{\left(a\right)}\&Element;X_i\left(u_i\right)}{\max }\{\mathrm{QoS}\left(s_{i+1}^{\left(a\right)}\right)+f_{i+1}\left(u_{i+1}\right)\\ u_{i+1}=s_{i+1}^{\left(a\right)}\\ f_n\left(u_n\right)=0\\ i=n-1,...,1\end{array}\right.$

U wherein _iExpression state set U _iElement; Adopt recursive algorithm that Dynamic Programming Equation is found the solution, obtain the Services Composition scheme of global optimum.

Beneficial effect

The present invention is in conjunction with networking manufacturing mode characteristics, the manufacturing operations process that adopts Work flow model to describe, the generative process of semantic service assembled scheme is converted into a dynamic evolution, multistage decision process, closing with data semantic between service is constraint, take the assembled scheme overall quality as optimization aim, utilize dynamic programming method to generate the assembled scheme of global optimum.The method is easy to determine globally optimal solution, has saved calculated amount, helps simultaneously to have improved the enforceability of assembled scheme and the automaticity of anabolic process for the user provides abundanter assembled scheme result.

Description of drawings

Fig. 1: the generative process of Services Composition scheme.

Fig. 2: the generative process of Services Composition scheme in the present embodiment.

Embodiment

Below in conjunction with specific embodiment, the present invention is described:

The present embodiment is take the machining blade part as example, and this blade processing is divided into processing damping platform (v ₁), blade (v ₂), blade root (v ₃) three process, v ₀And v ₄Be beginning and end node, as shown in Figure 2.The user is identical for the degree of concern of each mass parameter, requires various deviation control in 10%, finds the solution the optimal service assembled scheme that satisfies this manufacturing operation requirement.

The semantic service combined method based on dynamic programming in the present embodiment adopts following steps:

Step 1: in the present embodiment, the processing assembled scheme of machining blade part includes 3 manufacturing activities, is divided into processing damping platform (v ₁), blade (v ₂), blade root (v ₃) three process, be divided into 4 manufacturing activities node v according to manufacture process _i, i={0 wherein ..., 4}, v ₀Represent start node, v ₄Represent end node; According to the service discovery matching process, every procedure exists a plurality of candidate service examples to meet the demands, wherein candidate service

Expression manufacturing activities node v _iJ corresponding candidate service; The candidate service that satisfies its functional requirement that each manufacturing activities node is corresponding forms the original state set of this manufacturing activities node, and the original state set of five manufacturing activities nodes in the present embodiment is respectively

${\mathrm{<figure-callout\; id="1"\; label="U"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">U</figure-callout>}}_1=\{s_1^{\left(1\right)},s_1^{\left(2\right)},s_1^{\left(3\right)}\},$ ${\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">U</figure-callout>}}_2=\{s_2^{\left(1\right)},s_2^{\left(2\right)},s_2^{\left(3\right)},s_2^{\left(4\right)}\},$ ${\mathrm{<figure-callout\; id="3"\; label="U"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">U</figure-callout>}}_3=\{s_3^{\left(1\right)},s_3^{\left(2\right)}\},{\mathrm{<figure-callout\; id="4"\; label="U"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">U</figure-callout>}}_4=\left\{s_4^{\left(1\right)}\right\}.$

Step 2: for guaranteeing the correctness of composite services behavior, must satisfy data dependence relation between the service that is combined, namely in the composite services sequence, each forerunner's service that is positioned at the sequence front should produce the desired input of follow-up service.The present invention will make resource and encapsulate by the semantic service form, and the concept in the output in service interface, input parameter and domain body is mapped.

Data dependence relation between the candidate service of this step computational manufacturing active node and the candidate service of follow-up adjacent manufacturing activities node, and effective follow-up set of service corresponding to the candidate service of definite manufacturing activities node:

Wherein for manufacturing activities node v ₁The 1st candidate service

The employing following steps obtain

Effective follow-up set of service:

Get manufacturing activities node v ₂Corresponding original state set U ₂In element The concept c of each element input parameter is described ₂Be respectively:

Describe

The concept c of output parameter ₁Be CNC milling machine.Calculate respectively and describe

The concept c of output parameter ₁With description U ₂In the concept c of each element input parameter ₂Between semantic similarity Sem (c ₁, c ₂):

$\mathrm{Sem}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2)=\frac{|P\left(c_1\right)\&cap;P\left(c_2\right)|}{|P\left(c_1\right)\&cap;P\left(c_2\right)|+\mathrm{\&alpha;}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2)|P\left(c_1\right)/P\left(c_2\right)|+(1-\mathrm{\&alpha;}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2))|P\left(c_2\right)/P\left(c_1\right)|}$

wherein in function P (c) expression body to the attribute definition of concept c, attribute definition to CNC milling machine and these two concepts of machining center in making resource ontology is as follows: P (CNC milling machine)={ specifications and models, the worktable width, the speed of mainshaft, rapid traverse speed, tool-changing speed, interlock speed, main motor current, the main shaft peak torque, repetitive positioning accuracy, main shaft master high radial circle is beated }, P (machining center)={ specifications and models, the worktable width, the speed of mainshaft, number of motion axes, main motor current, the main shaft peak torque, repetitive positioning accuracy, milling plane degree }.∩ ,/, ‖ is respectively intersection of sets, difference and gesture computing, for example | P (C ₁) ∩ P (c ₂) | expression P (c ₁) and P (c ₂) number of total element, $\mathrm{\&alpha;}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2)=\left\{\begin{array}{cc}\frac{\mathrm{depth}\left(c_1\right)}{\mathrm{depth}\left(c_1\right)+\mathrm{depth}\left(c_2\right)}& \mathrm{depth}\left(C_1\right)>\mathrm{depth}\left(C_2\right)\\ 1-\frac{\mathrm{depth}\left(c_1\right)}{\mathrm{depth}\left(c_1\right)+\mathrm{depth}\left(c_2\right)}& \mathrm{depth}\left(C_1\right)\&le;\mathrm{depth}\left(C_2\right)\end{array}\right.,$ α (c ₁, c ₂) reflected the Asymmetric ef-fect that two concepts positions in the body level produces semantically similar evaluation.Function depth (c) calculates concept c residing degree of depth in body.0≤Sem (c ₁, c ₂)≤1,1 expression concept c ₁, c ₂In full accord, 0 expression is fully uncorrelated.Calculate successively

With U ₂In the data dependence relation degree of each element be $\mathrm{Sem}(s_1^{\left(1\right)}\&CenterDot;{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,s_2^{\left(1\right)}\&CenterDot;c_2)=1,$ $\mathrm{Sem}(s_1^{\left(1\right)}\&CenterDot;{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,s_2^{\left(2\right)}\&CenterDot;c_2)=0.78,$ $\mathrm{Sem}(s_1^{\left(1\right)}\&CenterDot;{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,s_2^{\left(3\right)}\&CenterDot;c_2)=1,$ $\mathrm{Sem}(s_1^{\left(1\right)}\&CenterDot;{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,s_2^{\left(4\right)}\&CenterDot;c_2)=0.78.$ In 10%, the threshold value that namely requires the data dependence relation degree is 0.9 due to the various deviation control of customer requirements, so

Be not

Effective follow-up service, do not satisfy data dependence relation,

Effective follow-up set of service is

As shown in Figure 2, if there is the limit to be connected between service node, satisfy data dependence relation between expression both, can carry out combination operation, otherwise there is no line.In the assembled scheme solution procedure, only keep real effectively segmentation scheme on each stage like this, the segmentation scheme that does not meet the demands will be abandoned as inferior solution, thereby in time cut off the path that can not expand to optimal case, accelerate to find the solution speed.

Step 3: the overall quality of calculation services, as the index of stage decision-making.In order to guarantee that each stage can make optimizing decision, need to set up the stage target function that is used for weighing selected decision policy quality.At functional similarity and satisfying under the prerequisite of data dependence relation, service quality becomes the important indicator of weighing the decision-making quality of doing, and the stage target function is designated as

Because the target direction of QoS parameter, dimension, span etc. are all inconsistent, simply original quality numerical value are made weighted sum and can not correctly reflect relative superior or inferior between the Services Composition scheme, so need to carry out standardization processing to QoS parameter.Here adopt the min-max normalization method, select [0,1] interval that original semantic service qualitative data is carried out standardization processing, realize objectivity and the comparability of qualitative data.

Candidate service is described

All original quality parameters

Can be expressed as

W represents the QoS parameter number, manufacturing activities node v _iCandidate service set U _iIn the mass parameter of all services consisted of the original quality parameter matrix.By this structure service quality matrix-style, the mass parameter in a plurality of spaces is unified in a computation model, when increasing a new mass parameter, namely increase new row from matrix, make interpolation and the deletion of mass parameter not affect follow-up solution procedure, satisfied the extensibility requirement of QoS parameter.

This example adopts 5 parameters such as response time, cost, reliability, availability and satisfaction to come service quality measurement, sets up the original quality parameter matrix of each manufacturing activities node, here with manufacturing activities node v ₂Be example, have 4 candidate service to satisfy its functional requirement, their qos parameter is as shown in table 1, constructs accordingly manufacturing activities node v ₂Original quality parameter matrix Q (v ₂):

Table 1 v ₂The candidate service mass parameter of manufacturing activities node

$Q\left(v_2\right)=\left[\begin{array}{ccccc}10& 500& 0.9& 1& 7\\ 12& 100& 0.6& 0.8& 4\\ 5& 200& 0.8& 0.6& 8\\ 20& 300& 1& 0.9& 9\end{array}\right]$

Step 4: to the original quality parameter matrix normalization of each manufacturing activities node: still with manufacturing activities node v ₂Be example, get

Be respectively matrix Q (v ₂) in maximal value and the minimum value of k row, k=1 ..., w.

When k row original quality parameter is the parameter value direct ratio type mass parameter that more high quality-of-service is better, with k row original quality parameter according to

${\mathrm{ib}}_{\mathrm{ik}}^{\left(j\right)}=\left\{\begin{array}{cc}\frac{q_{\mathrm{ik}}^{\left(j\right)}-q_k^{\min }}{q_k^{\max }-q_k^{\min }}& q_k^{\max }\&NotEqual;q_k^{\min }\\ 1& q_k^{\max }=q_k^{\min }\end{array}\right.$

Carry out normalization; When k row original quality parameter is parameter value more during the poorer inverse ratio type mass parameter of high quality-of-service, with k row original quality parameter according to

${\mathrm{db}}_{\mathrm{ik}}^{\left(j\right)}=\left\{\begin{array}{cc}\frac{q_k^{\max }-q_{\mathrm{ik}}^{\left(j\right)}}{q_k^{\max }-q_k^{\min }}& q_k^{\max }\&NotEqual;q_k^{\min }\\ 1& q_k^{\max }=q_k^{\min }\end{array}\right.$

Carry out normalization;

Because reliability, availability and satisfaction are direct ratio type mass parameter, response time, cost are inverse ratio type mass parameter, so the manufacturing activities node v after normalization ₂Mass parameter matrix B (v ₂):

$B\left(v_2\right)=\left[\begin{array}{ccccc}2/3& 0& 0.75& 1& 0.6\\ 1/3& 1& 0& 0.5& 0\\ 1& 0.75& 0.5& 0& 0.8\\ 0& 0.5& 1& 0.75& 1\end{array}\right]$

Step 5: the overall quality value of calculating each candidate service of each manufacturing activities node: for manufacturing activities node v _i, according to the mass parameter matrix after normalization, according to

$\mathrm{QoS}\left(s_i^{\left(j\right)}\right)=\frac{\underset{k=1}{\overset{t}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_{k}i{b}_{\mathrm{ik}}^{\left(j\right)}}{\underset{k=1}{\overset{w-t}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_{k}d{b}_{\mathrm{ik}}^{\left(j\right)}}$

Computational manufacturing active node v _iThe overall quality value of each candidate service, ω wherein _k∈ [0,1] and

Represent k mass parameter shared proportion on effectiveness of quality, reflected user preference, t represents the quantity of direct ratio type mass parameter.

In the present embodiment, the user is identical to the attention rate of each mass parameter, i.e. ω ₁=ω ₂=ω ₃=ω ₄=ω ₅=0.2.Manufacturing activities node v ₂Be example, according to B (v ₂) in data, obtain each the service overall quality value be respectively:

$\mathrm{QoS}\left(s_2^{\left(2\right)}\right)=0.38,$ $\mathrm{QoS}\left(s_2^{\left(3\right)}\right)=0.74,$ $\mathrm{Qos}\left(s_2^{\left(4\right)}\right)=5.5,$ As shown in Figure 2, service node numerical value that s identifies is this service colligate mass value.

Step 6: with f _i(u _i) be optimization aim, set up Dynamic Programming Equation

$\left\{\begin{array}{c}{f}_i\left(u_i\right)=\underset{s_{i+1}^{\left(a\right)}\&Element;X_i\left(u_i\right)}{\max }\{\mathrm{QoS}\left(s_{i+1}^{\left(a\right)}\right)+f_{i+1}\left(u_{i+1}\right)\\ u_{i+1}=s_{i+1}^{\left(a\right)}\\ f_n\left(u_n\right)=0\\ i=n-1,...,1\end{array}\right.$

U wherein _iExpression state set U _iElement; Adopt recursive algorithm that Dynamic Programming Equation is found the solution, obtain the Services Composition scheme of global optimum.

In the present embodiment:

1. when i=3, state set

Work as state

Perhaps

The time, its follow-up set of service

So set up recursion equation be:

$\left\{\begin{array}{c}{f}_3\left(u_3\right)=\underset{s_4^{\left(1\right)}\&Element;X_3\left(u_3\right)}{\max }\{\mathrm{QoS}\left(s_4^{\left(1\right)}\right)+f_4\left(u_4\right)\}\\ f_4\left(u_4\right)=0\end{array}\right.$

Have

Obtain f by recursion equation ₃(u ₃)=0, optimum follow-up service

2. when i=2, state set ${\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">U</figure-callout>}}_2=\{s_2^{\left(1\right)},s_2^{\left(2\right)},s_2^{\left(3\right)},s_2^{\left(4\right)}\}:$

Work as state

The time, effective follow-up set of service is

Obtain according to recursion equation:

$f_2\left(s_2^{\left(1\right)}\right)=\max \left\{\begin{array}{c}\mathrm{QoS}\left(s_3^{\left(1\right)}\right)+f_3\left(u_3\right)\\ \mathrm{QoS}\left(s_3^{\left(2\right)}\right)+f_3\left(u_3\right)\end{array}\right\}=\max \left\{\begin{array}{c}0.92+0\\ 0.89+0\end{array}\right\}=0.92,$ Optimum follow-up service $x_2^{*}\left(s_2^{\left(1\right)}\right)=s_3^{\left(1\right)};$

In like manner, $f_2\left(s_2^{\left(2\right)}\right)=0.92,$ Optimum follow-up service $x_2^{*}\left(s_2^{\left(2\right)}\right)=s_3^{\left(1\right)};$

$f_2\left(s_2^{\left(3\right)}\right)=0.92,$ Optimum follow-up service $x_2^{*}\left(s_2^{\left(3\right)}\right)=s_3^{\left(1\right)};$

$f_2\left(s_2^{\left(4\right)}\right)=0.89,$ Optimum follow-up service $x_2^{*}\left(s_2^{\left(4\right)}\right)=s_3^{\left(2\right)}.$

3. work as i=1, state set ${\mathrm{<figure-callout\; id="1"\; label="U"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">U</figure-callout>}}_1=\{s_1^{\left(1\right)},s_1^{\left(2\right)},s_1^{\left(3\right)}\}:$

Work as state

The time, effective follow-up set of service is

Obtain according to recursion equation:

$f_1\left(s_1^{\left(1\right)}\right)=\max \left\{\begin{array}{c}\mathrm{QoS}\left(s_2^{\left(1\right)}\right)+f_2\left(s_2^{\left(1\right)}\right)\\ \mathrm{QoS}\left(s_2^{\left(3\right)}\right)+f_2\left(s_2^{\left(3\right)}\right)\end{array}\right\}=\max \left\{\begin{array}{c}3.53+0.92\\ 0.74+0.92\end{array}\right\}=4.45,$ Optimum follow-up service $x_1^{*}\left(s_1^{\left(1\right)}\right)=s_2^{\left(1\right)};$

In like manner, $f_1\left(s_1^{\left(2\right)}\right)=1.66,$ Optimum follow-up service $x_1^{*}\left(s_1^{\left(2\right)}\right)=s_2^{\left(3\right)};$

$f_1\left(s_1^{\left(3\right)}\right)=0.39,$ Optimum follow-up service $x_1^{*}\left(s_1^{\left(3\right)}\right)=s_2^{\left(4\right)};$

According to state transition equation and above computation process, employing is back tracking method sequentially, can try to achieve the optimizing decision sequence of this example:

$\left[\begin{array}{cccc}{\mathrm{<figure-callout\; id="1"\; label="u"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">u</figure-callout>}}_1=s_1^{\left(2\right)}& \&RightArrow;{\mathrm{<figure-callout\; id="2"\; label="u"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">u</figure-callout>}}_2=s_1^{\left(2\right)}& \&RightArrow;{\mathrm{<figure-callout\; id="3"\; label="u"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">u</figure-callout>}}_3=s_2^{\left(3\right)}& \&RightArrow;{\mathrm{<figure-callout\; id="4"\; label="u"\; filenames="HDA00002808189500012.png"\; state="\{\{state\}\}">u</figure-callout>}}_4=s_3^{\left(1\right)}\\ \&DownArrow;& \&DownArrow;& \&DownArrow;& \&DownArrow;\\ x_1^{*}=s_1^{\left(2\right)}& x_2^{*}=s_2^{\left(3\right)}& x_3^{*}=s_3^{\left(1\right)}& x_4^{*}=s_4^{\left(1\right)}\end{array}\right]$

Hence one can see that, and the optimal service assembled scheme that satisfies this mission requirements is

Therefore after determining the optimal service assembled scheme, namely determine the assignment to each movable QoS in operation flow, can also calculate other crucial mass parameter values such as deadline overall tasks time, cost of this scheme.

Claims (1)

1. semantic service combined method based on dynamic programming is characterized in that: adopt following steps:

Step 1: the assembled scheme that will include n-1 manufacturing activities is divided into n manufacturing activities node v according to manufacture process _i, i={0 wherein ..., n}, v ₀Represent start node, v _nRepresent end node; According to the service discovery matching process, obtain the candidate service that satisfies its functional requirement corresponding to each manufacturing activities node, wherein candidate service

Expression manufacturing activities node v _iJ corresponding candidate service; The candidate service that satisfies its functional requirement that each manufacturing activities node is corresponding forms the original state set of this manufacturing activities node, wherein for original state set U _i, and Expression manufacturing activities node v _iM candidate service that satisfies its functional requirement arranged;

Step 2: the data dependence relation between the candidate service of the candidate service of computational manufacturing active node and follow-up adjacent manufacturing activities node, and effective follow-up set of service corresponding to the candidate service of definite manufacturing activities node: wherein for manufacturing activities node v _iJ candidate service The employing following steps obtain

Effective follow-up set of service:

Step 2.1: get manufacturing activities node v _i+1Corresponding original state set U _i+1In element

Calculate and describe

The concept c of output parameter ₁With description

The concept c of input parameter ₂Between semantic similarity Sem (c ₁, c ₂):

$\mathrm{Sem}(c_1,c_2)=\frac{|P\left(c_1\right)\&cap;P\left(c_2\right)|}{|P\left(c_1\right)\&cap;P\left(c_2\right)|+\mathrm{\&alpha;}(c_1,c_2)|P\left(c_1\right)/P\left(c_2\right)|+(1-\mathrm{\&alpha;}(c_1,c_2))|P\left(c_2\right)/P\left(c_1\right)|}$

Wherein in function P (c) expression body to the attribute definition of concept c, ∩ ,/, ‖ is respectively intersection of sets, difference and gesture computing, $\mathrm{\&alpha;}(c_1,c_2)=\left\{\begin{array}{cc}\frac{\mathrm{depth}\left(c_1\right)}{\mathrm{depth}\left(c_1\right)+\mathrm{depth}\left(c_2\right)}& \mathrm{depth}\left(C_1\right)>\mathrm{depth}\left(C_2\right)\\ 1-\frac{\mathrm{depth}\left(c_1\right)}{\mathrm{depth}\left(c_1\right)+\mathrm{depth}\left(c_2\right)}& \mathrm{depth}\left(C_1\right)\&le;\mathrm{depth}\left(C_2\right)\end{array}\right.,$ Function depth (c) calculates concept c residing degree of depth in body;

Step 2.2: the method traversal manufacturing activities node v that adopts step 2.1 _i+1Corresponding original state set U _i+1In all elements, described

The concept of output parameter and description U _i+1In semantic relevancy between the concept of each element input parameter, get U _i+1In the semantic relevancy element that is not less than setting threshold form

Effective follow-up set of service

Step 3: the original quality parameter matrix of setting up each manufacturing activities node: for m candidate service arranged

Manufacturing activities node v _i, v _iOriginal quality parameter matrix Q (v _i) be:

$Q\left(v_i\right)={[q{s}_i^{\left(1\right)},......,q\left(s_i^{\left(m\right)}\right)]}^T={\left[\begin{array}{c}{q}_{i1}^{\left(1\right)},q_{i2}^{\left(1\right)},......,q_{\mathrm{iw}}^{\left(1\right)}\\ q_{i1}^{\left(2\right)},q_{i2}^{\left(2\right)},......,q_{\mathrm{iw}}^{\left(2\right)}\\ ........................\\ q_{i1}^{\left(m\right)},q_{i2}^{\left(m\right)},......,q_{\mathrm{iw}}^{\left(m\right)}\end{array}\right]}_{m\&times;w}$

Wherein

The expression candidate service

The original quality parameter

W represents the mass parameter number;

Step 4: to the original quality parameter matrix normalization of each manufacturing activities node: for manufacturing activities node v _iOriginal quality parameter matrix Q (v _i), get

Be respectively matrix Q (v _i) in maximal value and the minimum value of k row, k=1 ..., w; When k row original quality parameter is the parameter value direct ratio type mass parameter that more high quality-of-service is better, with k row original quality parameter according to

${\mathrm{ib}}_{\mathrm{ik}}^{\left(j\right)}=\left\{\begin{array}{cc}\frac{q_{\mathrm{ik}}^{\left(j\right)}-q_k^{\min }}{q_k^{\max }-q_k^{\min }}& q_k^{\max }\&NotEqual;q_k^{\min }\\ 1& q_k^{\max }=q_k^{\min }\end{array}\right.$

Carry out normalization; When k row original quality parameter is parameter value more during the poorer inverse ratio type mass parameter of high quality-of-service, with k row original quality parameter according to

${\mathrm{db}}_{\mathrm{ik}}^{\left(j\right)}=\left\{\begin{array}{cc}\frac{q_k^{\max }-q_{\mathrm{ik}}^{\left(j\right)}}{q_k^{\max }-q_k^{\min }}& q_k^{\max }\&NotEqual;q_k^{\min }\\ 1& q_k^{\max }=q_k^{\min }\end{array}\right.$

Carry out normalization;

Step 5: the overall quality value of calculating each candidate service of each manufacturing activities node: for manufacturing activities node v _i, according to the mass parameter matrix after normalization, according to

$\mathrm{QoS}\left(s_i^{\left(j\right)}\right)=\frac{\underset{k=1}{\overset{t}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_{k}i{b}_{\mathrm{ik}}^{\left(j\right)}}{\underset{k=1}{\overset{w-t}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_{k}d{b}_{\mathrm{ik}}^{\left(j\right)}}$

Computational manufacturing active node v _iThe overall quality value of each candidate service, ω wherein _k∈ [0,1] and

T represents the quantity of direct ratio type mass parameter;

Step 6: with f _i(u _i) be optimization aim, set up Dynamic Programming Equation

$\left\{\begin{array}{c}{f}_i\left(u_i\right)=\underset{s_{i+1}^{\left(a\right)}\&Element;X_i\left(u_i\right)}{\max }\{\mathrm{QoS}\left(s_{i+1}^{\left(a\right)}\right)+f_{i+1}\left(u_{i+1}\right)\\ u_{i+1}=s_{i+1}^{\left(a\right)}\\ f_n\left(u_n\right)=0\\ i=n-1,...,1\end{array}\right.$

U wherein _iExpression state set U _iElement; Adopt recursive algorithm that Dynamic Programming Equation is found the solution, obtain the Services Composition scheme of global optimum.

Images