Performance Simulation Agent Operational Strategy

The Performance Simulation Agent will execute the simulation of a given configuration. To achieve this goal, two main operational states need to exist, namely the establishment of the simulation model that represents the configuration solution and the execution of the simulation, as defined in chapter 5. To that end it is proposed the use of the syntheses model presented in (Ferreira et al. [128]), extending it to cater for all performance characteristics. It recognises that the assembly process accuracy and repeatability, of an assembly system depends upon two aspects; the physical arrangement of different pieces of equipment and the logical sequence of operations which they need to jointly execute to achieve their common assembly objective. Furthermore, the same model can be extended to cater for identified performance characteristics. Figure 6.10 provides an overview of this model.

153 Figure 6.10- Token flow description for an example of MAS

The use of a Petri net based model has been adopted which allows the use of different token types for components and equipment that are propagated throughout the assembly process chain. Essentially, a token is being created for each component which is being assembled. Component tokens are merged into a product token when the assembly process is of “Fixating” type. However this does not take into account the modules that are responsible for the assembly processes. These are represented through module specific tokens which carry the repeatability properties of the equipment. These are merged into the component token when another module takes responsibility for the component or a “Fixating” type process occurs.

It is proposed that Performance Simulation Agents emulate this behaviour through the exchange of messages that contain structured information on equipment tokens and assembly process tokens. This information consists of updated objects for each of the performance characteristics, which represent the assembly process tokens, in conjunction with the last agent that effected the equipment token and the equipment ID. This way, once the equipment is different, the agent can simply give back the equipment token to the relevant agent. Therefore the agent is able to perform the two place holder roles defined in the model.

In addition, the agent is also responsible to perform the necessary operations for the transition that precedes it, this in effect ensures the emulation of the Petri Net model. Each Equipment Module Agent will deploy the required Performance Simulation

Agents for a given solution with the specific information on the assembly process

that is executing, plus the equipment responsible for it. In addition to this, the

Process Place Holder Equipment Place Holder Process Type: “Fixating” Attributes: Accuracy Values ... ... Component Token Equipment Token ... Final Product Token

154

Equipment Module Agents also need to provide the required connection for the

execution of the simulation, so that it provides the individual connections that each

Performance Simulation Agent needs to establish. This provides a straightforward

manner to establish the virtual configuration which enables the behaviour model of a given solution.

Once the behaviour model has been synthesised, it enables the simulation of the underlying system behaviour based on token passing approach. An unaltered Petri net, however, does not provide the desired behaviour characteristics and requires a more specific definition of how the tokens behave in the model through the established place holders and transitions.

The transitions are responsible for the management of the tokens, making sure tokens exiting the flow are incorporated into the component token. This is its basic behaviour if a token is exiting the flow this needs to be passed on to the component token, otherwise the tokens are simply passed to the next process place holder.

Figure 6.11 describes the operational behaviour of the transitions presented in the

model. This diagram details how the different performance characteristics are processed in the transitions in order to calculate the results.

Figure 6.11 - Transition Behaviour Algorithm

Begin

Wait for Message (Tolken) All Tokens? Equipment Token Leaving? Stack Up Equipment Token Errors to Component Errors Check Equipment Tokens Multiple? Update Cycle Time With Maximum Among Option

Add Cycle Time

Stack Up Power Values Send Equipment Back to Start Trigger Place Holder Process Yes No Yes Yes No No End

155 The process place holders are assembly process specific and thus they incorporate the process classification, as defined in chapter 4. Their behaviour is driven by the assembly processes classification which serves as input to establish how to affect the tokens, namely in the relation to the assembly system errors. Figure 6.12 describes the process place holder behaviour where it is clearly defined how this model should react to the different process types. The behaviour of the place holders will be broken into the four performance simulation characteristics which will require different actions. The repeatability and accuracy performance characteristics can be broken down into three types: one for Qualifying, Decision and Compensate processes which affect the equipment token compensate error matrix which will be incorporated into the component token once the equipment token leaves the system. The other type is the fixating processes which merge all tokens present into a new component type token (or final product). The final type is for any other process types which simply stack up the equipment token with the relevant error. It is important to note that only when the equipment token leaves the system will the stack up of component token errors occur. The power consumption has no compensation possibility and therefore it is simply stacking the value that all the assembly processes are consuming in the simulation, which is treated in the transitions algorithm. The cycle time uses a different approach for the merger of the cycle time values, if two tokens are to be merged, therefore coming from two different sources the value that is set is the highest, which again is treated in the transitions algorithm. Therefore the place holders operate only on the precision aspects as seen in Figure

6.12.

Figure 6.12 - Place Holder Behaviour Algorithm

Begin Qualifying Process? Compensate Process? Decision Process? No

Affect Eq. Token Compensate Matrix Fixating Process? No No Merge Components Tokens Errors and relevant Eq. Errors

Yes

Add relevant Eq. Errors when not already present in

Eq. Token No Yes

Affect Eq. Token Compensate

Matrix Yes

Affect Eq. Token Compensate Matrix Yes Send Message To Next Agent End

156 The synthesis algorithm for the precision characteristics is based on a state transition approach which is used to construct a 3D parametric model using 4x4 matrix transformations of all contribution factors and sources of errors leading up to and during the completion of a full assembly process. The algorithm distinguishes between processes that contribute to the error, those that do not and those that compensate errors from previous operations. Each Module/Skill can contribute in 6 Degrees of Freedom to the assembly error of the workstation (3 translations and 3 rotations). Each error is expressed by its variation (upper and lower bound) and the accuracy of the error value (3 or 6 sigma) which is provided in the equipment module description. The synthesis algorithm for the remaining characteristics is less complex since it is a simple value, so no matrix is required, yet the overall behaviour is the same as the synthesis algorithm for the precision characteristics.