The ship-repair enterprises belong to a special industry which concentrates
management, production, technology and service. The process of ship-repair has
highly complexity because of the constantly changed task and different ship
type and the short repair period. In general, the ship-repair belongs to individual
production and includes numerous projects. It is difficult to plan for ship-repair
enterprises (Taiyi et al., 2003).
In recent 30 years, particularly the last 10 years, the ship maintenance industry
in the China has been developed rapidly and the capability of ship maintenance
also has been greatly increased. But there are still greatly difference between
the developed countries and China (Chen and Ye, 2009).
In this paper, the Multi-Agent System (MAS) has been adopted. The production management modelling of ship-repair based on MAS has been built. It can realize the integration of data stream, business flow and fund flow.
MULTI-AGENT SYSTEM STRUCTURE
Agent is a new computing model in the artificial intelligence field and can
continually perceive the change outside and itself and also can give a corresponding
action by itself. Agent can be looked as a black box and can perceive the environment
via sensors and act on the environment via effectors. A complex system maybe
includes one or more interactional agent systems. This system is named Multi-Agent
System (MAS). MAS denotes that one complex problem can be solved via cooperation,
harmonization and negotiation of two or more agents together. MAS is considered
as the key technique to lower production cost, realize decentralized production
and self-adaptation and resolve complicated process. And it is also a new methodology,
which runs through each modern advanced manufacturing field, from the dynamic
alliance, distributed intelligent manufacture system, enterprises integration,
Enterprise Resource Planning (ERP) to field control (Zhang
et al., 2009).
MODELLING BASED ON MAS
Function and structure of each agent: The production management modeling of ship-repair included Order Agent (OA), Manager Agent (MA), Production Agent (PA), Quotation Agent (QA) and Inspection Agent (IA) and the Production Agent was built also based on MAS, which includes Task Agent (TA), Out-worker Agent (WA) and Resource Agent (RA). Each agent defines its own rule according to its own task and characteristic and then put it in rule database which was packaged in agents.
Manager agent (MA): MA, which is core of the whole system, is built to manage, harmonize and control the whole system. It can get the information from other agents and resolve the conflict among agents. MA must record the registration information, identity (ID), address and function of each agent. When add a new agent or delete an old one, MA will update the system in accordance with the information. In addition, MA can apperceive other agents at all times. MA has the ability of judgment, decision making and get the optimal compound mode. The structure of MA includes information management, shipowner management, order management, production management and feedback information.
Order agent (OA): OA receives the repair orders and saves the information
to the system.
|| Flow chart of OA
|| The modelling of PA based on MAS
The structure of OA includes the order number, type, shipowner name, repairing
item, contract amount, repairing period, etc. The flow chart of OA (Fig.
Inspection agent (IA): The inspection of ship-repair includes three-level inspection, self-inspection, workshop-inspection and shipowner-inspection. IA creates inspection report and transmits the inspection report to MA after finishing the inspection. The report includes the information about order number, engineering name, quality requirement, index, results and remarks and so on.
Quotation agent (QA): When the shipowner or its representative sends the inquiry list to the sales department of the ship-repair enterprise, the sales department must form a quotation according to the repairing items of the inquiry list and send it to the shipowner as soon as possible. QA is responsible for forming the detailed quotation. The structure of QA includes quotation number, inquiry list number, inquiry list content, detailed price, total price, delivery date, requirement, index and remarks. When QA completes the quotation, it saves the quotation and transmits the quotation to MA.
Production agent (PA): PA can get the tasks and available resources according to the engineering order. PA disassembles the engineering task received by contract into work order and then transmits it to TA. At the same time, PA records the information to the task table. The system inquires the unoccupied equipments via RA and the assign the task to suitable equipment. The structure of PA includes information management, task management, resource management, outer worker management.
PA was built also based on MAS. Each agent also defines its own rule according to respective tasks and characteristics and keeps it in the rule database packaged in agents. PA implements general scheduling and harmonizes the conflict between RA, WA and TA (Fig. 2).
Resource agent (RA): It is built on client. After registration, RA begins to wait for inviting public bidding. When RA receives inviting public bidding document, it will evaluate the task according to production state and ability. Then computes a feasible bidding value and fills in the bidding document and sends it to PA (Fig. 3).
When RA receives the winning bidding message, the task will be scheduled on corresponding machine. After finished the task, RA will waiting for the next bidding. Also RA monitors and controls the machining process real time. If the normal state are changed, for example, there are something wrong with certain machine or add new machines, RA immediately reports it to PA and PA will reschedule. The dynamic scheduling is finished. The structure of RA includes equipment name, equipment information and equipment condition.
When the new equipment is purchased, the corresponding information will be added to RA. When the new task arrives, RA will pass the unoccupied resources information to TA and PA.
Task agent (TA): TA answers for executing the whole production plan.
It creates automatically and will be vanished automatically after finished one
task. It can finish the distribution and execution of the plan via negotiation
of each client and server.
|| Flow chart of RA
|| Flow chart of TA
An incepting buffer and a sending buffer are packaged in each agent for communicating
to other agents. The working flow is shown in Fig. 4. The
structure of TA includes task information, contract content, and delivery date.
When the new task arrives, TA will receive the unoccupied equipments and outworkers information from RA and WA and scheduling these tasks and transmit the scheduling to PA.
Out-worker agent (WA): There is a characteristic in the labour force management in almost all the ship-repair enterprises. Most workers are out workers. That is to say, they belong to certain engineering brigade. When the repairing task are so much or so urgent or some other condition that the existing hired workers can not be competent for the repairing work, the production management department will hire extra engineering brigade. WA is responsible for managing the engineering brigades and its workers. Instantaneously record the engineering brigade and its workers information, contact number telephone number, the starting and ending time, amount of work, working hours, reward, punishment, specialties and its equipments and so on. The structure of WA includes register, brigade name, brigade number, worker number, telephone number, material, index and remarks.
When the new task arrives, WA will pass the available workers information to TA and PA.
The modeling based on MAS: This modeling, based on multi-agent and the
negotiation of Contract Net Protocol (CNP), adopts Client/Server structure for
communication. It realizes communication via the communication port of the numerical
control system. RA and equipments are connected by Controller Area Network (CAN)
bus (Ren et al., 2009) and transmits data two
ways between numerical control equipments.
The working flow is the following: After OA received the repairing orders
from shipowners, it begins to estimate if the enterprise has the capability
to meet the requirements. For the unserviceable task, sales department gives
the shipowners practicable suggestions. For the competent task, OA receives
it and changes it into engineering order and transmits the order to MA; MA issues
information to PA and IA. IA will give the inspection flow and requirement.
PA starts up and transmits the task to TA, RA and WA. In this paper, TA, RA
and WA ensure the distribution of each task on each resource according to the
inviting public bidding-bidding mechanism of CNP (Smith,
1980). TA receives the information of unoccupied equipments and available
workers from RA and WA after RA and WA evaluates throughput itself according
by the task from PA. PA determines the equipment winning the bidding by the
biding information and sends the winning bidding information to TA and RA and
thus the whole inviting public bidding-bidding process can be completed (Fig.
|| The Modeling based on MAS
This new modeling realized the information transmits and shares instantly via MA, OA, PA, IA and QA. Each agent can work in parallel and communicate and collaborate with each other. So, shorten the time of delivery and improve the working flow and production efficiency. Each department shares the information and thus can reduce the misunderstanding. It has been found that it can improve the working flow and production efficiency and shorten the time of delivery.
This research was supported by the State Natural Sciences Foundation Monumental Projects of China (No. 61034003) and the National Science and Technology Planning Preparatory Project of the Twelfth Five-year Plan of China (No. SQ2011GX03E00708).