The ability to accurately develop subject-specific, input causation models, for blood glucose concentration (BGC) for large input sets can have a significant impact on tightening control for insulin dependent diabetes. More specifically, for Type 1 diabetics (T1Ds), it can lead to an effective artificial pancreas (i.e., an automatic control system that delivers exogenous insulin) under extreme changes in critical disturbances. These disturbances include food consumption, activity variations, and physiological stress changes. Thus, this paper presents a free-living, outpatient, multiple-input, modeling method for BGC with strong causation attributes that is stable and guards against overfitting to provide an effective modeling approach for feedforward control (FFC). This approach is a Wiener block-oriented methodology, which has unique attributes for meeting critical requirements for effective, long-term, FFC.

The potential for successful automatic control of blood glucose concentration (BGC) has entered a new era because of recent technological advancements in insulin pumps and blood glucose sensors. However, a critical advancement necessary for full automation and long-term use is a control algorithm that can effectively maintain tight control of BGC under extreme variation of important disturbances such as activity, stress, and food consumption. Because feedforward control (FFC) models disturbances directly, it has the potential to eliminate the effects of disturbances completely. A Wiener-type feedforward control law is limited to the inclusion of only input (i.e., modeled disturbances and the manipulated variable) dynamics. Using a semicoupled modeling network that includes pseudo-blood insulin concentration, this work presents a more phenomenological FFC law that includes input dynamics, blood insulin and blood glucose dynamics, and blood glucose levels. Modeling results on 15 adults with type 1 diabetes mellitus for the proposed method are nearly identical to Wiener modeling results.

When modeling dynamic processes for several inputs with freely existing data, such as data collected with normal process operations, the ability to accurately model the output response for a given input change is impeded when inputs are cross-correlated (i.e., pairwise) as this adversely affects accurate estimation of the causative effects of inputs on the response variable. The causative effects of the inputs can be evaluated functionally and analytically via the Jacobian matrix which is done in this work for NARMAX and Wiener structures that are linear and nonlinear in model parameters. This analysis shows that the Wiener structure with physically based nonlinear parametrization is superior. This conclusion is also supported in this work by a modeling study on a real distillation column consisting of eight test runs over a period of three years.

Many input variables of chemical processes have a continuous-time stochastic (CTS) behavior. The nature of these variables is a persistent, time-correlated variation that manifests as process variation as the variables deviate in time from their nominal levels. This work introduces methodologies in process identification for improving the modeling of process outputs by exploiting CTS input modeling under cases where the input is measured and unmeasured. In the measured input case, the output variable is measured offline, infrequently, and at a varying sampling rate. A method is proposed for estimating CTS parameters from the measured input by exploiting statistical properties of its CTS model. The proposed approach is evaluated based on both output accuracy and predictive ability several steps ahead of the current input measurement. Two parameter estimation techniques are proposed when the input is unmeasured. The first is a derivative-free approach that uses sample moments and analytical expressions for population moments to estimate the CTS model parameters. The second exploits the CTS input model and uses the analytical solution of the dynamic model to estimate these parameters. The predictive ability of the latter approach is evaluated in the same way as the measured input case. All of the data in this work were artificially generated under the probabilistic CTS model. ABSTRACT: Many input variables of chemical processes have a continuous-time stochastic (CTS) behavior. The nature of these variables is a persistent, time-correlated variation that manifests as process variation as the variables deviate in time from their nominal levels. This work introduces methodologies in process identification for improving the modeling of process outputs by exploiting CTS input modeling under cases where the input is measured and unmeasured. In the measured input case, the output variable is measured offline, infrequently, and at a varying sampling rate. A method is proposed for estimating CTS parameters from the measured input by exploiting statistical properties of its CTS model. The proposed approach is evaluated based on both output accuracy and predictive ability several steps ahead of the current input measurement. Two parameter estimation techniques are proposed when the input is unmeasured. The first is a derivative-free approach that uses sample moments and analytical expressions for population moments to estimate the CTS model parameters. The second exploits the CTS input model and uses the analytical solution of the dynamic model to estimate these parameters. The predictive ability of the latter approach is evaluated in the same way as the measured input case. All of the data in this work were artificially generated under the probabilistic CTS model.