Fits~ Overview –Options-2

• Allow Compaction（允许压缩）：使用此设置后，如有必要，AIDAPro将在辨识时压缩FIR系数。FIR辨识中涉及的一个核心数值计算是矩阵的转置，其维度取决于未知参数的数目。通过压实，AIDAPro引擎把FIR近似为更小的系数数据，使得压缩趋势更快。（请参阅第31页的Compaction（压缩））
• Initial Conditions at Steady State（稳态下的初始条件）：如果设置了此标签，则AIDAPro引擎将假定FIT涉及的所有变量在数据中断前时间段处于稳态。你可能因为数据段、限幅、坏的数据质量等原因导致数据出现中断。（请参阅第38页的Initial Conditions （初始条件）。）
Fit中适用每个输出的选项：
• Ramp Flag（斜坡标签）：选中复选框将把输出指定为斜坡变量。此标签将在FIR拟合和参数拟合窗口上设置Ramp Flags（斜坡标签）。
• Drifting Flag（漂移标签）：如果输出不稳定，且由于未测量的干扰影响而有漂移，请选中此框。若选中此框，则将使用预滤波器。默认值将从用于漂移的工作空间选项继承。（请参阅第36页的Data Prefiltering（数据预滤波）。）
• No Immediate Response（非立即响应）：被识别的模型是离散时间模型，并且由于采样通常具有“零阶保持”。这通常导致当改变独立变量后，输出不会立即响应。然而在某些情况下当改变输入后，可能会立即响应，输出响应会非常快。其中一些例子包括 – 快速响应流股或者具有即时比例反冲的过程控制器阀门等。关闭此标签表示存在立即响应，且在拟合时的增益估计通常会更好。（请参阅第39页的No Immediate Output Response（非立即输出响应）请注意此标签仅适用于FIR分析。
• Prefilter Horizon（预滤波器时域）：只有当漂移标签打开或选中时此时域才适用。预滤波器时域确定了预滤波器的特性。预滤波器应用于输入和输出数据集，因此同时影响了FIR和参数拟合。（请参阅第36页的Data Prefiltering（数据预滤波））

原文：
**• Allow Compaction: **When using this setting, AIDAPro compacts the FIR coefficients during identification if necessary. One of the core numerical computations involved in FIR identification involves the inversion of a matrix the size of which depends on the number of unknown parameters. With compaction the AIDAPro engine approximates the FIR by a smaller number of coefficients and the computations tend to be quicker. (Refer to “Compaction” on page 31.)
**• Initial Conditions at Steady State: **If this flag is set the AIDAPro engine assumes that all variables involved in the FIT were at steadystate during periods of time before breaks in the data. You may have breaks in data because of segments, slicing, or bad quality.(Refer to “Initial Conditions” on page 38.)
Options applicable for each output in Fit:
**• Ramp Flag: **Check box to designate an output as a ramp variable. This flag sets the Ramp Flags on the FIR fit and parametric fit windows.
**• Drifting Flag: **If the output is non-stationary and has a drift in it due to unmeasured disturbance effects, check this box. If checked, the prefilter is used. The default is inherited from the workspace options for drifting. (Refer to “Data Prefiltering” on page 36.)
**• No Immediate Response: **The models being identified are discrete time models and typically have a “zero-order-hold” due to sampling. This usually causes an output not to show an immediate response to changes in independents. However in certain situations where an output responds very quickly to an input change there may be an immediate effect. Some examples include – fast responding flows, or process controller valves that have an immediate proportional kick in them, etc. Turning this flag off implies there is an immediate response, and the gain estimate usually gets much better in the fit. (Refer to “No Immediate Output Response” on page 39.) Note that this flag applies only to the FIR analysis.
• Prefilter Horizon: This horizon applies only when the drifting flag is turned on or checked. The prefilter horizon determines the nature of the prefilter. The prefilter is applied to both the input and output datasets and therefore affects both the FIR and parametric fits. (Refer to “Data Prefiltering” on page 36.)

2016.11.14