Modelling of cross organizational manufacturing resource service chain based on service supply-demand dynamic matching network

In the cloud manufacturing systems, both manufacturing tasks and manufacturing services are in a dynamic environment. How could cloud manufacturing platform optimizes manufacturing cloud services based on QoS, matching an optimal service composition for manufacturing tasks has become an urgent problem at present. In view of this problem, we study the matching of manufacturing tasks and manufacturing services from the perspective of complex network theory. On the basis of manufacturing task network and manufacturing service network, a dynamic matching network theory model of manufacturing task-service is constructed. And then, we take a dynamic assessment of QoS. Finally, we use load and dynamic QoS as the optimization objectivities, transform the optimal manufacturing service composition problem into the shortest path problem, and the dynamic scheduling of manufacturing services is realized.


Introduction
Because of the diversification of customer needs, the manufacturing industry is increasingly trying to integrate products and service components together to quickly respond to customer needs, shorten product development cycles and maintain agility by creating more complete solutions. However, shifting from a product-centric perspective to a solutionoriented perspective is challenging. [1] manufacturing tasks and manufacturing services are in a dynamic and changing environment. Because of the dynamic nature of manufacturing resources, manufacturing services are in a dynamic environment, including the addition of new services, the evolution or extinction of old services. Exceptions may also occur to the services being invoked. [2] In view of dynamic and changeable operating environment, there are many different researchers study the issue of supply demand matching from different angles. Moon etc. building an integer programming model to match the supply and demand of manufacturing resource services to ensure the optimality of QoS. [3] Zeng ets put forward the idea of Global re-planning to cope with the anomalies in service composition. [4] However, re-planning will bring additional costs. For this reason, Dai et al. proposed a self-healing method for service composition, which reduces the number of re selection through performance prediction. [5] Dotoli etc. design an integrated E-supply chain for agile and environmentally conscious manufacturing. [6] Most of these methods cannot be well applied to manufacturing services. Firstly, there is no systematic approach to considering changes in demand and services. Secondly, compared with the traditional Web service environment, the complexity and uncertainty of cloud manufacturing environment are more prominent. Thirdly, the variety of manufacturing resources is more complex and diverse. As the result, the existing researches on supply and demand matching and optimal allocation of manufacturing resources (for example, the problem of matching for simple demand or primitive task [7] and matching for complex demand or compound task [8]) are still far from the service-oriented practice.
The dynamic changes of manufacturing tasks and manufacturing services pose challenges to the management of cloud manufacturing systems. At present, there is a lack of in-depth research on the supply-demand dynamic matching method in the cloud manufacturing environment, and no mature scheme has been formed. We will use the theory of complex network to construct the service-demand dynamic matching network model. The model will provide a theoretical basis for the dynamic management of manufacturing tasks and manufacturing services.

Problem description
In service-oriented manufacturing systems, the supply is the set of various social available manufacturing resources and capabilities in the form of services.Nodes load of service manufacturing, which are denoted as L, referring to the number of manufacturing tasks currently processed by manufacturing services. Quality of Service (QoS)，which refers to the non-functional attributes of manufacturing services. QoS is an important basis for the service composition.
In the traditional manufacturing service composition, we need to pick up one suitable service from the service classes of 1 C , 2 C and 3 C to build the service composition for a manufacturing task sequence 1 2 3 T T T   ,. Each service class contains several services which having same or similar function but different in QoS, as shown in fig.1. Based on the potential call relationship between services, we can choose different service composition to carry out the task, e.g. 1 4 8 Assuming that the user's requirement changes in t , so we need to adjust the manufacturing task 2 T that is in the state of execution, which causing the service 4 S unable to meet the requirement and the service needs to be reselected. It can be seen that traditional service composition methods need to replace multiple services simultaneously, and it is difficult to adapt to changes in dynamic environment, resulting in Qos degradation.

Models of manufacturing service network (MSN)
Manufacturing service network (MSN) is the virtual network topology structure composed of a large number of service nodes and a variety of edges. Atomic manufacturing services (AMS) refer to fine-grained manufacturing services that can be invoked independently and are functionally indivisible, and is recorded as S . Manufacturing services can be either mapped to manufacturing resources with a single function, but also can be mapped to manufacturing resources with complex functions. The former is called fine-grained manufacturing service, that is, atomic manufacturing service. The latter, which we call coarse-grained manufacturing services, is denoted as C S . We divided coarse-grained services C S into composite services made up of multiple atomic manufacturing services.
Thus, there are only atomic manufacturing services in MSN. The manufacturing service dependency edge (MSDE) refers to the weighted directed edges based on the potential call relationship between atomic manufacturing services, which is denoted by

Models of manufacturing task network (MTN)
Similar to MSN, manufacturing task network (MTN) is also a virtual network topology structure composed of a large number of manufacturing task nodes and a variety of correlation edges among those task nodes. The cloud manufacturing platform can be decomposed into a series of atomic manufacturing tasks (AMT) with sequential relations through a certain strategy. AMT is an indivisible fine-grained manufacturing task, which can be executed by calling an atomic manufacturing service, and is recorded as T .

Model description
Based on the models of MSN and MTN described in sections 3.1 and 3.2, the key point to construct the model of dynamic matching network (DMM), is to work out the existence of inter-network dynamic matching edge (INDME) which connect service nodes and task nodes. Inter-network dynamic matching edge indicates the dynamic matching relationship between atomic manufacturing task T and atomic manufacturing service, which are defined as formula . When 1 TS ij w  , it is indicates that the manufacturing task i T and the manufacturing service i S are in a dynamic binding state. There is a dynamic matching edge between i T and i S . When the manufacturing task i T is completed, the matching edge disappears. The time varying characteristics of the network are realized.
Dynamic matching network is a double-layer network formed by the dynamic matching relationship between manufacturing task network and manufacturing service network. It is recorded as

Dynamic scheduling of manufacturing services
QOS is an important evaluation standard for dynamic scheduling of manufacturing services. Due to the diversity of QoS indicators of manufacturing services, such as response time, cost, reliability, and so on, there are great differences between units and ranges. Therefore, it is usually necessary to normalize the QoS value, as shown in formula (3).
in which, k w is the weight of each attribute value of the QoS, which indicate the degree of preference for each attribute of a user. max Dynamic scheduling of manufacturing services has two optimization objectives. Firstly, from the perspective of users, cloud manufacturing platform needs to provide users with QoS optimal manufacturing services or service composition. Secondly, from the perspective of cloud manufacturing platform management, in a multi-user environment, how to realized the load balancing of various manufacturing service nodes through reasonable scheduling of resources and optimized the load queue is the other objectives. Suppose that the number of manufacturing service community that need to be invoked in the current manufacturing process is N , each manufacturing service community contains services number are n . We can model the dynamic scheduling of manufacturing services as formula (4 Next, we will use shortest path routing strategy of complex network to solve the above models. The globally optimal service composition can be transformed to search for a least cost service composition path from the manufacturing service network, as shown in algorithm 1.

Conclusions
As the common issue to be addressed in advanced manufacturing systems, the supplydemand matching of manufacturing service in service-oriented manufacturing system is becoming to the decisive factor for its application and operation. Especially under the current socialization and cloud environment, there are different kinds of supply-demand matching problems considering the uncertainty and dynamics in service-oriented manufacturing systems. The theory of complex network provides an alternative method to analyse and address these problems. From the perspective of complex network, this paper studies the dynamic management of manufacturing tasks and manufacturing services in cloud manufacturing systems. A dynamic matching network theory model has been built. The model takes the load and dynamic QoS as the optimization objectives, and transforms the optimal manufacturing service composition problem into the shortest path search in the manufacturing service network, so as to realize the dynamic scheduling of the manufacturing service.
Further research will improve the reliability of QoS's other indicators, such as the relationship between reliability and load, to improve the generality of the dynamic QoS model. Furthermore, we will study the impact of statistical properties of the service network on the service composition, such as the degree of nodes, and so on.