Hybrid Control in Manufacturing - The Producer Scheduler Consumer (PSC) Problem

By
A H Ahmad Kamil
Information Systems Engineering II
Imperial College of Science, Technology and Medicine

Abstract

The article is about hybrid approach in controlling the Producer-Scheduler-Control (PSC) problem of a manufacturing factory. The scheduler is the main element and it is controlled using multiple-control-agent. The model for that factory is discussed which is a collection of work units and also their network connections. A single experiment with two work units is demonstrated and how this can be implemented using a University of Maryland-developed operating systems called Maruti.

Introduction
Scheduler in the Manufacturing PSC Problem
The Factory Model
Experiment With Two Work units
Implementation Using Maruti Operating System
Conclusion
Reference

1. Introduction

The term Hybrid Systems are the systems described by composition of logical and composition models and used to describe large and varied class of systems. In addition, hybrid systems also display interactions between continuous(or continuum) and discrete(or logical) control. Control programs that integrate these logical and continuum models work well for small scale systems. Unfortunately, for large scale systems, this integration has two main implications namely

increase in cost and
decrease in reliability.

The fundamental result of a hybrid control program simultaneously complies with logical and continuous-time constraints. The hybrid control system is then used in a manufacturing factory in solving the Producer-Scheduler-Consumer (PSC) problem. The producer needs to adjust the number of output according to the demands from the consumer or if any problems along the production line occur. Hybrid control is applied to the scheduler that plays the vital role in ensuring the smoothness of this adjustment.

2. Scheduler in the Manufacturing PSC Problem

2.1 The Constraints

A manufacturing factory has many work units that actually doing the process of a particular production. Since a hybrid control system is applied to control the scheduler of a production process, integrating scheduling process of the manufacturing production and execution of manufacturing process within a work unit of a factory is required. Unfortunately, there are two main constraints due to this integration. These are

Qualitative characteristics and logical constraints.Schedule that was produced by the scheduler are done by evaluating throughput, tardiness (how slow the systems is), work in progress and other production factors which are associated with the quality of the product based on the appearance, density, texture, and thickness. These factors will create a logical, or event-based, production process switching function that alter the production quantity.
Quantitative characteristics and continuum constraints: It is basically the problem in deciding the next thing to be done. The most important point here is that the logical decisions discussed earlier are captured as events occuring at instants in time and must be compatible with the laws of physics governing the continuous operation of physical components of the factory such as the conveyors, motors etc.

Furthermore, it is also very important that the synchronisation of logical events and continuous processes between producers and consumers is required in implementing the planning, scheduling and synchronisation of the factory. The smoothness of this synchronisation determines the return on investment of the components for that factory. This return on investment can be increased by improving appropriate utilities that effecting the particular synchronisation. Figure of merit as in figure 1 govern the scheduler process.

Figure 1

2.2 Raw Materials Transformation

Operations performed at a number of processing points transform the raw materials (or stocks) into finished goods. The scheduler plays a vital role in assigning this stocks into a sequence of processing points where operations that have been fixed earlier in its plan are undertaken. A typical work unit in the factory would consists of one or more processing points.

2.3 Important factors for the Scheduler

A few factors must be taken care by the scheduler;

2.4 Rescheduling: The Cause and Effect

In general, on the other hand, schedules are hardly used to completion due to few problems such as a breakdown of a machine, stocks failed to arrive and etc. Hence a change in schedule is required. Anyway, one should realise that many work units have been started and currently in operation before doing any rescheduling. In addition, restarting the whole work is costly and hence the scheduler must produce a new schedule as soon as possible and take the new constraints into considerations.

While rescheduling the whole plan, degrading of the factory's perfomance is inevitable since work units need to be reconfigured. Thus the factory need a distributed control system so that each work unit can make its ownself decision. Generally, factory plans are based on nominal manufacturing scenario but execution of that particular plan occurs based on the actual decisions and hence deviation from the original plan is possible. Inconsistencies in logical and continuum systems leads to failure in executing the current plan. A new plan therefore must be reactively created that still complies with the logical and continuum constraints. Multiple-agent declarative control supports off-line analysis and design and on-line generation of planning, scheduling and control software that acts as a "glue" between between comercially available planning software and commercially available control systems.

3. The Factory Model

A factory can be modelled as a network of work units. Each work unit is considered as a physical plant which is a component of the factory. An agent controller is associated with each unit. The attributes of a work unit are

Configurations:

Sensors:

Actuators:

Operations:

Constraints:

4. An Experiment with Two Work Unit

A reactive planning, scheduling and control system for a simulated shop floor consisting of two work units with an agent controlling each unit has been experimented to know whether the system can handle both continuum and logical constraints. Two work units are illustrated in figure 2.

Figure 2

The experiment used three processes namely producer controller agent, consumer controller agent and the simulation itself. The main function of each controller agent at a particular interval is to prove that the system follows the status logically and satisfies the the constraints imposed by other agents. This simulation also has the capabilities of generating external events so that the controlling agents are forced to react to unexpected changes. Some changes in inventory or failures are two examples of these events. The proccessing rate from of the producer and the consumer is tuned in adapting the above changes.

5. Implementation using Maruti Operating Systems

Maruti Operating Systems, developed at the University of Maryland, is one example where implementing manufacturing control is possible. The main reason why this system used is that it has the capability to support changes in the "calendar" for scheduling of process during execution. Changing of schedules during execution is vital to adapt any unexpected changes as mentioned earlier. The schedule of executions is mantained in calendars and Maruti supports the management of resources using calendars and also supports calendar entries can be changed during execution, i.e. a reactive operation. Fault handling and monitoring and detection of any improper operation are another examples of Maruti applications.

6. Conclusions

Hybrid control is used in the scheduler since the fundamental result complies with both logical and continuum constraints for each work unit in the factory. Hence the path from the producer to the consumer is effectively control which is undoubtly vital for a manufacturing factory.

7. References

IEEE Control Systems,