过程中绝大多数非独立变量（被控变量）都是稳态变量，也就是说若一个独立变量发生变化，非独立变量最终将到达一个新的恒定（稳态）值。
然而，一些被控变量的质量或能量在持续累积或消耗。这些变量被称为积分变量，或斜坡变量。所有斜坡变量的共同特点是，当独立变量发生改变后，它们不会进入一个稳态。
斜坡变量与部分过程材料或能量的积累或消耗直接相关。最常见的一个斜坡响应例子是容器液位。如果一个罐中的进流量增加，无限大的罐中液位将增加（假设出流量保持恒定）。罐的出流量必须增加相同的量才能平衡液位。因此，对所有独立变量（进流量和出流量）液位都表现出斜坡响应。不变的是，响应特性是非独立变量相对于独立变量的属性。
其它变量，如温度或压力，也能表现出斜坡型响应。如用于流化床催化裂化的再生器运行于部分燃烧模式时，再生器的温度由催化剂中碳含量指示，其可以当做斜坡型。在加氢裂化反应器中，反应器压力是衡量氢消耗量的指标。如果补充的氢气流与被消耗的氢气不完全相等，压力将上升或下降；所以在这种情况下，压力是氢平衡的量度。
通常在简单的两产物精馏塔（特别是高纯度塔）中，中间塔盘温度展现出斜坡型的行为。尽管这一变量事实上是稳态变量，但由于过程稳态时间太长以至于在感兴趣的时域它表现出斜坡行为。我们现在把这种类型的变量称为伪斜坡。在DMCplus计算时它被视为一个斜坡，但不要求其对应用程序非常重要。
在实践中绝大多数斜坡变量都使用监管控制器来控制。通常我们通过设计一种多变量DMCplus控制器代替其控制斜坡。这样做的潜在好处是：
·PI液位控制器的动态被从过程中除去。这样往往会减少过程到稳态的时间，并且可以从DMCplus模型响应中移去PI控制器的动态。
·鉴于DMCplus控制器可以考虑系统中所有因变量的相互关系，其可以更好地控制液位。
·不同于监管控制器，DMCplus控制器可以使用组合操作变量来控制斜坡变量。因此该方法可以在操作变量的约束条件下对其子集进行操作，而不会牺牲控制性能。这些增加的灵活性常常能带来可观的经济效益，并且改善斜坡控制。
切断斜坡PI控制回路进行手动阶跃测试的潜在缺点是：
·若干扰非常大（与系统容量相比）而又未提供紧急保护，因变量有饱和的可能性（如溢出缓冲罐）。
·若干扰是突然的，抑或相对于DMCplus控制间隔是高频的，系统存在触犯控制界限的可能性。
·项目在工厂测试阶段时过程需要更密切的观察，并且可能需要更多的矫正动作，因此斜坡变量必须手动控制，同时，必须避免过度依赖人工反馈控制，使得数据不会被相关，以确保有效的模型辨识。
注：Aspen SmartStep具有自动阶跃测试时维持斜坡变量控制，同时提供高品质不相关的数据进行模型辨识的功能。
附原文：
Most dependent (controlled) variables in a process are steady-state variables; that is, if an independent variable is moved, the dependent variable eventually reaches a new constant (steady state) value.
However,some controlled variables measure accumulation or depletion of mass or energy. These variables are known as integrators, or ramp variables. The common feature of all ramp variables is that they do not come to steady state after a change in an independent variable.
Ramp variables are related directly to the accumulation or depletion of material or energy in sections of the process. The most common example of a ramp responseis the level in a vessel. If the flow entering a tank is increased, the level in the tank increases for an infinitely large tank (if the out-flow is held fixed). The flow exiting the tank must be increased by the same amount to"balance" the level. Thus, the level exhibits ramp responses to both independent variables (in-flow and out-flow). Invariably, the response characteristic is a property of the dependent variable as opposed to the independent variable.
Other variables, such as temperature or pressure, can also exhibit ramp-type responses.For a Fluidized Catalytic Cracker whose regenerator runs in partial burn mode, the regenerator temperatures are an indication of carbon level on the catalyst, and act as ramps. In a hydrocracker reactor, the reactor pressure is a measure of hydrogen consumption. If the make-up hydrogen flow is not exactly equal to the hydrogen being consumed, the pressure will either rise or fall; so in this case, the pressure is a measure of hydrogen balance.
Often,in simple two-product distillation towers (especially in high purity towers),an intermediate tray temperature exhibits ramp type behavior. Although this variable is really a steady-state variable, the steady-state time of the process is so long that over the time domain of interest, it acts like a ramp.This type of variable is now known as a pseudo-ramp. In the DMCplus calculation it is treated like a ramp, but is not required to be critical to the application.
In practice, most ramp variables are controlled using regulatory controllers. Itis often advantageous to replace these controls by designing a multi-variable DMCplus controller to control the ramp. The potential benefits of doing this are:
·The dynamics of a PI level controller are removed from the process. This will often reduce the timeto steady state for the process, and will remove the PI controller dynamics from the DMCplus model responses.
·The DMCplus controller can control the level better by considering the inter-relationship among all of the dependent variables in the system.
·Unlike the regulatory controller, the DMCplus controller can use a combination of manipulated variables to control a ramp variable. Thus, the process can be operated with a subset of the manipulated variables at their constraints, without sacrificing control performance. This added flexibility can often result in appreciable economic returns and improved ramp control.
The potential drawbacks of disconnecting ramp PI control loops for manual step testing are:
·If the disturbances are very large (compared to the system capacity), there is a possibility of saturating the dependent variable (e.g. overflowing a surge tank) if emergency overrides are not provided.
·If the disturbances are abrupt, or of high frequency relative to the DMCplus control interval, there is a possibility of violating control limits.
·During the plant test phase of the project, the ramp variable must be controlled manually, which requires much closer observation of the process; this potentially requires more corrective moves. At the same time, excessive reliance on manual feedback control must be avoided so the data will not be correlated to ensure valid model identification.
Note:Aspen SmartStep includes features for maintaining control of ramp variables during automated step testing, while providing high-quality uncorrelated data for model identification.
2015.9.27
