Our aim with the prototype machine is to produce a carefully considered design which will be a convincing demonstration of the merits of the hexapod architecture. We hope that it will lead to the appearance in the marketplace of a new generation of affordable computer controlled machines. The design has evolved over the past year to provide optimised solutions to the different technical requirements and problems.
One of the first actions was the construction of a full scale mechanical model, from cardboard tube and wood. The model has been rebuilt twice to reflect changes of design and dimension. It gives a much better qualitative understanding of the kinematics of the hexapod mechanism, and the impact of the machine, than can be obtained from a scale model or a 2D computer representation. It has aided joint design and provided additional confirmation of the correctness of the computer model.
The detailed mechanical design of the prototype machine has relied on information provided by the computer model and the full scale mechanical model. However, there are a number of important initial choices to be made before detailed design can begin. These are influenced by experience, feasibility, cost, knowledge of the application, and so on. In the course of the work university departments, suppliers and exhibitions have been visited to gather information and discuss the project.
Working Volume: The target is 1.2m x 1.2m x 0.3m. This is a very useful size which will cover many applications. Two such machines mounted side by side in a tandem arrangement will have a working length of 2.4m.
Target Materials: Wood, plastics, non-ferrous metals, etc. We are not targetting ferrous metals, though the hexapod architecture is well suited to them. This is a question of cutting forces and the stiffness required to limit cutter displacement.
Machining Speeds: For wood, 50mm/s is usually too fast when cutting. 20mm/s or lower is the norm. Lower speeds would be required for metals.
Traversing Speeds: With a typical component, most of the time on the machine is spent cutting. There is a limited advantage in very high traversing speeds which have to be paid for in various ways. We will be happy with the 150mm/s the prototype design will provide.
Spindle Speed and Power: This will depend on the application. Say 2kW at 20,000rpm maximum. Speed control required. A background issue at the present stage of the design.
Precision: For wood, +- 0.05mm would be good, +- 0.025 excellent. Typically, short dimensions need to be more precise than long dimensions, eg hole sizes. The computer model shows resolution errors typically 1 or 2 times the actuator resolution, and very rarely up to 5 times. The design actuator resolution of 0.0125 should produce the lower target value with small movements. With large movements screw and joint inaccuracies will increase the error.
Rigidity: With high quality machining at slowish speeds in wood, cutter forces are not high. The representative cutter force used in the computer model is 50N (5kgF) in each of the x and y directions, resultant about 70N. The model gives typical mechanism displacements under 0.01mm, effectively rigid. The displacement of axially loaded frame members under this order of load is very small. The top joint support requires careful stiffening.
Singularities: Mechanisms of the hexapod type suffer from mathematical singularities (instabilities) which must be avoided. This would be the case for example if a leg pair plane became parallel to the platform plane. In that situation the mechanism has no stiffness against forces normal to the platform. The software model shows singularities as regions with high leg forces and cutter displacements.
Thermal Behaviour: To keep errors due to thermal movement to a minimum, steel (and not aluminium) should be used for all the critical parts: mechanism, support frame, and bed. Necessary sources of heat, such as leg motors, should be well separated from the critical parts, and the heat removed by ventilation or some other means. Unnecessary sources of heat, such as a vacuum pump, should not be mounted within the machine. The aim should be to keep the structure in a reasonably constant thermal environment.
Work Holding: The primary method will be vacuum, a well proven technique. Very small parts need a secondary method, such as an adhesive. The dessication effect of vacuum on jig materials must be considered, as this can lead to jig distortion.
Portability: Less than 5% of the machine volume is structure; the rest is necessary space. We don't want to pay to send this space when a machine is delivered! Unusually for a machine tool, the concept here is of a kit which can be delivered in pieces, to be assembled and calibrated by the customer.
Actuator Type: The obvious choice at the present is a ball screw mechanism, standard today for cnc machines. Truly low cost hexapod machines will require a breakthrough in this area. For now, ball screws are a good solution, combining precision positioning with rigidity. Backlash must be completely eliminated; the computer model shows how sensitive the mechanism is to it. The final choice is 20mm diameter 5mm pitch ground screws with preloaded nuts.
Actuator Configuration: There are a number of related issues here: whether the actuator is telescopic or passes through the top joint; how the actuator and motor are connected; how resonance problems are avoided. We want to confine the mechanism to the zone between the joints to avoid complexity, and make the closed leg length as short as possible to keep down the height of the machine. This leads directly to the choice of a telescopic arrangement with the motor offset from the screw. We minimise resonance problems by using a telescopic tube arrangement, and by supporting the free end of the screw. The solution is very similar to that used in commercial linear actuators.
Actuator Motor: A stepper motor with rotary encoder feedback is our preferred solution. The system cost including drive is significantly lower than a dc servo system, and the control strategy is a lot simpler. Our fast integer control algorithm can be implemented in silicon at modest cost for commercial hexapod control systems. We are using slightly modified 853 motors from our partner Stebon Ltd. In 400 step/rev mode the screw resolution is 0.0125mm. Design maximum speed is 25 rps.
Actuator Belt Drive: We follow the suggestion of TU Dresden and use a precision timing belt drive with the motor mounted parallel to the screw. This gives the advantages of a slightly compliant connection, thermal separation, and ease of dismantling. After discussions with Gates we are using their PGGT 5MR belts. Expected backlash is of the order of 1/4 motor step. A high belt preload is required for high stiffness, and this influences the screw and motor bearing arrangements.
Universal Joints: This is the most difficult area of the detailed design. Two legs share an axis at the lower 3-axis joint which needs as wide a rotation range as possible for maximum platform manoevrability. The computer model shows the sensitivity of the mechanism to backlash, so plain bearings are not a good solution. We follow the approach of the ball screw, and use a preloaded angular contact joint design. Because of the required small envelope, the two axes of the ball joint need special bearings. We have visited a specialist bearing manfuacturer to discuss the design and see the techniques used. We plan to use ceramic balls for the smaller raceways; this will give electrical isolation and allow collisions between legs, platform and frame to be detected.
Layout: A natural support structure will echo the triangulation of the mechanism, benefitting from the high stiffness/weight ratio of members which are only axially loaded. Our first approach to the design started with the equilateral triangle of the mechanism top joints, and tried to fit the square bed into a triangular plan framework. This doesn't work well, and it is much better to let the bed determine the form. The top joint support platform can easily be stiffened with sheet material. Typical leg forces are low.
Connections: The ball and cone joint looks ideal, and allows easy assembly and disassembly. Each frame tube has inside it a rod with a left-hand and right-hand screw at alternate ends. The screws engage in balls which are drilled and tapped at the appropriate angles. The rod can float longitudinally, but is prevented from rotating where it passes through the cones. By turning the tube and cones the rod is turned, and the screws enter the balls tightening the assembly. The longitudinal stiffness is determined by the combined areas of the tube and the rod. The repetition should allow the parts to be made cheaply in production.
Construction: The bed will be made of light sheet steel with interlocking internal stiffening in a welded construction. The top surface must be easily machinable, and we envisage glueing a 10mm layer of a plastic material to the steel. This will be machined by the machine itself as part of the calibration and setup operations. Suitable plastic materials, such as cloth reinforced phenolics which are often used in this context, are hygroscopic and change dimension with humidity. In particular they dry out under vacuum, which is the working environment of the bed top. The effect of this on the flatness of the bed must be considered.
Support: The bed will carried on a triangulated structure echoing the main frame, connected to the main frame ball joints. There is ample opportunity for secondary stiffening if that is required.
Type: This remains undecided, to be considered after the prototype machine
is assembled and tested. As a temporary measure a portable router can easily
be fitted, as envisaged in the exploratory phase proposal. Key requirements
are:
(1) to contain the volume occupied by the spindle and drive;
(2) to avoid significant heat transfer to the platform;
(3) to give a wide range of cutter axis orientation;
(4) to provide an inexpensive tool changing system.