CogMaSh

• Management of knowledge and information

• Interdisciplinarity
• Virtual product development

• Methods
• Tools

Modern manufacturing systems such as flexible or reconfigurable manufacturing systems (FMS/RMS) still do not offer the degree of automation and flexibility necessary to meet todays market demands. This has several reasons: Static global and local control systems, pre-programmed process plans and inflexible hardware make it impossible to react to dynamic changes or fabricate non-predefined parts without excessive human intervention. In contrast, human machine shops handle unexpected events regularly, improvise solutions, plan on-the-fly and reach a high quality level of parts but for a high cost.

To combine the economic efficiency of a manufacturing system and the flexibility of a human-operated machine shop an autonomous manufacturing system is necessary, that has technical cognitive capabilities such as embedded knowledge, learning, reasoning and planning capabilities. The goal of the project is to develop such a "Cognitive Machine Shop" (CogMaSh). The system should be able to autonomously produce a wide range of non-predefined parts, adapt to unexpected events as well as learn and exploit new capabilities.

In this joint project we focus on creating an autonomous design-to-fabrication system looking at the areas of design synthesis, workpiece selection, machining planning and hardware reconfiguration. Our aim is to develop new methods and tools that will be tested and validated on the CogMaSh demonstrator FMS, an integrated Flexible Manufacturing System which includes substantial parts and capabilities of an industrial production system.

To autonomously identify possible workpieces for a new part a workpiece selection system based on semantic web technologies is developed. The system is able to find and rank feasible workpieces for a given job taking geometric and material data into account.

To address machining planning a shape grammar based method is being developed. The grammar is capable of representing machining processes and plans based on tool models and machine capabilities. Advanced search methods are being used to guide the tool path creation. The goal of the method is to output NC code for the fabrication of non pre-defined parts while being able to react to changes in the manufacturing environment, like the unavailability of tools, and reason about the created plans online.

Looking at hardware that supports the flexibility of an autonomous system an automatically reconfigurable fixture device is developed. Further, creation of a failure-safe agent-based control system capable of online operation and handling the reconfiguration processes, an ontology for computational representation of the reconfiguration knowledge and for reasoning about feasible configurations as well as synthesis methods for the design of new modular components are being investigated.

As an initial implementation, autonomous production of different example parts was achieved including computationally generating alternative part geometries, choosing workpieces, creating the required CNC code and adapting the fixture device to the workpiece shape.

For autonomous workpiece selection an ontology and according interfaces have been developed. The ontology takes the workpiece stock, the part's geometric model, material data and additional user input into account and outputs a ranked list of feasible workpieces. This computational knowledge representation and allows for autonomous identification of feasible workpieces and for reasoning about alternatives.

Further, a shape grammar based method for maching planning has been developed and implemented as a software prototype. The 3D workpiece model and the 3D part model are used to identify the removal volume. The shape grammar is used to represent the tools and machining process capabilities. Search techniques guide the tool path generation. The created plan is directly converted into NC code which can be transferred and executed on the milling machine of the CogMaSh FMS.

To automatically adapt the fixture device to workpieces of different geoemetries an autonomously reconfigurable vise-type fixture was developed. Further, a PLC-based software-agent control system capable of online operation was created. To computationally represent the configuration knowledge and reason about feasible configurations based on the geometric data of the workpiece an ontology was implemented. A prototype of the device was successfully integrated in the milling machine of the CogMaSh FMS. Using the developed methods the fixture device can be reconfigured autonomously.

In general, these results show the viability of using technical cognitive methods like planning and reasoning to increase the flexibility and degree of automation of manufacturing systems. Potential product applications include customized consumer products, engineered-to-order parts and all scenarios where efficient production of small batch sizes is necessary.