Robust Design and Effective Control of Mechanical Systems

我提议的研究计划是我在之前的探索基金中进行的工作的自然延伸。该计划现在更加专注，强调设计的健壮性和控制的有效性。 我们遵循G.Taguchi(1980年代)的概念，根据信噪比来定义稳健设计(RD)。实施这一概念的困难在于对噪声和信号的可信识别。在概念层面，我们引入了复杂性的概念，并以机构设计为例进行了说明(“复杂性”强烈依赖于上下文)，特别强调了较低的六个运动学副(回转、棱柱、螺杆、圆柱、平面和球面)。在可获得元件或组件的数学模型的实施例和详细设计级别，信号被识别为定义设计环境的参数，例如来自供应商线路的电力的频率和电压，设计者对其没有影响。噪声被定义为供应商参数中不可预测的变化。我们将这些过程建模为具有平均值和标准差的随机高斯过程。 我们计划在机器人的背景下使用我们的研发范例。具体地说，我们针对两种机器人：1)用于快速拾取和放置操作的机器人，这些机器人存在于广泛的工业流程中，从制造具有平面几何设计的手持消费品(例如手机)到食品包装。这些是四自由度(四自由度)机器人；和2)六自由度并联机器人，用于摇动操作，定义为高频(&gt；10赫兹)(高频)和小幅(SA)(“小”w.r.t.试验机的尺寸)。它的应用也很广泛，从简单的流体混合到惯性参数识别(IPI)，这是通信卫星姿态控制中的一个主要问题，自80年代以来，加拿大在这方面一直处于领先地位。我们的建议包括开发一种新的IPI方法，再次使用田口的稳健性范例：最大化信噪比。不仅如此。我们还计划开发更简单、低成本的IPI机器，通过对试件进行HF、SA平面位移。我们声称，在将样品呈现给具有不同姿势的平面三自由度机器人测试机时，将有可能识别刚性物体的所有10个惯性参数：质量、质心的三个坐标和惯性张量的六个独立分量。 对于我们设计和制作的机械系统的有效控制，这项工作是实用的多于理论的。我们计划继续开发控制策略及其算法，以便在“小”位移下实现10赫兹及以上的快速操作。挑战是巨大的，因为存在激发机器运动部件的较低自然模式的风险。

The research program I am proposing is a natural extension of the work I have conducted in previous Discovery Grants. The program is now more focused, stressing robustness in design and effectiveness in control. We define robust design (RD) in following G. Taguchi's concept (1980s), in terms of the signal-to-noise ratio. The difficulty in implementing this concept lies in the plausible identification of noise and of signal. At the conceptual level we introduced the concept of complexity, and illustrated it for the case of mechanism design ("complexity" is strongly dependent on the context), with special emphasis on the six lower kinematic pairs (revolute, prismatic, screw, cylindrical, planar and spherical). At the embodiment and the detailed-design levels, at which a mathematical model of elements or assemblies is available, the signal is identified as the parameters defining the design environment, for example, the frequency and voltage of the electric power coming from the supplier's line, over which the designer has no influence. Noise is defined as the unpredictable variations in the supplier's parameters. These we model as stochastic Gaussian processes with available values of mean and standard deviation. We plan to use our RD paradigm in the context of robotics. Specifically, we target 2 kinds of robots: 1) Those for fast pick-and-place operations, which are present in a broad range of industrial processes, from the fabrication of hand-held consumer goods designed with a flat geometry (e.g., cell phones) to the packaging of food. These are four-degree-of-freedom (four-dof) robots; and 2) six-dof parallel robots for shaking operations, defined by high frequency (> 10 Hz) (HF) and small amplitude (SA) ("small" w.r.t. the dimensions of the testing machine). Applications thereof are also broad, from simple fluid-mixing to inertia-parameter identification (IPI), a major issue in the attitude control of telecommunications satellites, in which Canada has been a leader since the eighties. Our proposal includes the development of a novel methodology for IPI, using, again, Taguchi's robustness paradigm: maximize the signal-to-noise ratio. Not only this. We also plan to develop simpler, low-cost machines of IPI, by subjecting the test specimen to HF, SA planar displacements. Upon presenting the specimen to the testing machine, a planar 3-dof robot, with different poses, it will be possible, so we claim, to identify all 10 inertia parameters of the rigid object: mass, three coordinates of the centre of mass, and six independent components of the inertia tensor. On the effective control of the mechanical systems we design and prototype, the work is more applied than theoretical. We plan to continue developing control strategies and their algorithms for fast operations, of 10 Hz and beyond, under "small" displacements. The challenges are enormous, because of the risk of exciting the lower natural modes of the moving parts of the machine.