One thing that has bugged me about these radial designs is that the stator isn't well supported against rotational forces - which is where it's loaded.
Structures like this are strong in compression and tension, but not in bending or torsion. Forces should be along the length of at least one straight piece, not at right angles to a single support. Single supports - especially thin ones - should not be twisted. Joints that may be pushed in more than one direction should be supported in at least two directions in two-D, three directions in three-D, so the supports can divide the load such that they're only loaded in compression or tension. The members should be connected to form triangles and pyramids, not squares or arms - especially arms with bends.
So think of them as if they had springy flexible joints where they make a corner or are fastened to another member. How do they bend?
In these designs you have a star of thin radial strips, with allthread at right angles on their ends, supporting a disk that is TWISTED by the main load. OOPS!
- When under a torque load the strips are in torsion and act like springs, while the allthread rods act like levers. The twisting stator is pulled toward the downwind rotor while enormous leverage on the allthread rods tries to fracture the stator where they are attached to it.
- When loaded along the axis, the strips act like springs in bending, allowing the stator to move in either direction and contact one rotor or the other. Further, there will be some frequency where it resonates. Vibration at this frequency will "pump up" a mechanical oscillation until it bounces between the rotors and friction there takes out vibrational energy faster than it is added. If this frequency happens to match the "singing" at some RPM you're toast.
Seems to me there are a couple things that could be done to fix this.
First: Stiffen the supports: Weld on a support against along-axis bending and torsion, making them into a T or L cross-section. They can be made ever stronger by making the strip and/or the support roughly triangular - wider at the hub than near the attachment of the allthreads. Alternatively you could use three rods or angle-irons to form a thin pyramid, being sure that they don't all have their planes parallel if they're sheet metal.
Second: Stiffen the allthreads against the helical analog of the "parallelogram collapse" mode of a square of supports by running a strip from the base of each to a point near the stator's attachment, forming triangles. This brace will be even stiffer if it can't fold up - which can be done by using an "angle iron" shape rather than a strap. (Alternatively you could brace in two directions with strip so at least one of each set of braces will be loaded in tension and not subject to fold-up failure modes.)
