06-12-2012, 04:28 PM
INDUSTRIAL WORKFLOWS RECOGNITION BY COMPUTER VISION AND ARTIFICIAL INTELLIGENCE TECHNOLOGIES
INDUSTRIAL WORKFLOWS.doc (Size: 816.5 KB / Downloads: 27)
ABSTRACT
The introduction of the Gigabit-Ethernet for Machine Vision opened a new road for workflows recognition in industrial and SMEs environments. The network centric approach to running a production level vision system are coming on stream, which provide distributed computing across industrial developments. Humans and machines interaction can be exploited and surveyed using artificial intelligence algorithms applied on signals from multiple standalone agents on camera networks and giving to the end user, a full view of abnormal or alarm events conditions during the workflows execution. This article is presenting the architecture of such a system, where the designer of the workflows, plus the engineer of the workflow surveillance system should consider, in order either to simulate the industrial process offline, or during the real time workflow execution in the SMEs or in the industrial environment. The related methodology is based on Java technologies, which are presented and latest innovations from the multi agents and workflow processes composition.
INTRODUCTION
A workflow process (or workflow for short) is a specific type of business process, a way of organizing work and resources. Workflows are
Commonly found within large administrative organizations such as banks, insurance companies, governmental agencies, and industrial production environments. Carrying out the tasks
of a workflow in a particular order is required to handle one type of case. Examples of ases are mortgage applications, customer complaints, claims for unemployment benefits, and production of especially SMEs industrial products.
In the last case, a workflow is used in handling the productivity and the sequence of events in order to construct an end product. In these environments humans and machines are cooperating sometimes standalone and sometimes in a sequence that the one part (human) should follow in strictly specific time limits and task description, in order the automated part (robot) to continue its activities in this industrial environment. For this reason, there are a lot of software tools that try to simulate SMEs companies beforehand of the starting of a production process. Humans should know exactly and need training in order to perform like robots.
WORKFLOWS
Workflow modelling supports simulation and analysis of workflows and helps the SMEs to specify a new workflow design very fast and with no cost before the system implementation. Software tools draw and play the manufacturing processes in a manner that human and robots could be simulated and also the automation procedures, that inform either the employees or the systems to start/stop and continue/pause the activities in the industry, can visualize their results through the usage of multiple cameras controllers, giving a full view of the workflow environment.
DESIGN
The execution of a business process passes through several stages in producing products. It often is convenient to distinguish the state of a business process execution. For example, to inspect whether a deadline will be met in producing a certain item, its current state of completion is relevant. Distinguishing an execution state is often done by referring to the operations that are already executed, the parts that still need to be constructed, or other milestones that are reached during the execution. As there may be many concurrent executions of a business process, we can refer to the state of a complete business process as the collection of states of its individual executions. Events that start a process are commonly referred to as triggers. In many approaches and definitions of business processes it is indeed very common to decompose a business process into smaller parts. One way of decomposition is to distinguish subprocesses, also known as subflows.
MODELLING
Benefits of the Petri net modelling technique are: their formal semantics their graphical notation support for complex process constructions in particular concurrency, and the availability of many analysis techniques . Petri nets in their basic form, however, lack the expressive power to model a complete workflow model, covering all the workflow components in detail. It should be noticed that this type of Petri net has the structure of a workflow net it is in general not sound (see Fig.1). On the basis of the solutions definition, we can define the notion of fulfilment. A fulfilling workflow model implements all solutions of a confined product data model in a correct way.
THROUGHPUT TIME
One of the most important performance indicators in industry is the throughput time. The throughput time of a specific case is the total amount of time spent from the moment that the case is initiated, until the moment it is completed. The throughput time of a case is in general a combination of service, queue, and wait times.
CONCLUSIONS
This paper is presenting an analysis of the workflow specifications that a designer in an SME or an industry should take into account in order to design and develop a system that manipulates humans and machines workflows. Such systems are not so easy to be designed and especially when the processes that take place are random or not so detailed constrained, in order to be followed exactly by a human. These difficult random cases that are part of a workflows process can be surveyed by a system
automatically, using artificial intelligence technologies and multi agent based systems applied on computer vision technologies, under workflows simulation programs or in real life industrial and SMEs environments. Either the simulation environment, or the actual surveyed industrial system by these technologies, benefits and can provide useful information to the end user (related alarms for abnormal conditions or exceptions in the workflow), and also to the workflow designer by bringing affordable and effective factory simulations and analysis factors, and minimizing or eliminating the risk of system failure.