The Design Space

In the analysis of artifacts above, we observed that the mask of every binary, uniform, stationary scheme can be expressed as the product of a number of a = (1 + z)/2yfz factors and a further factor called the kernel. Then we saw that the kernel itself can be expressed as the sum of a number of terms, each of which is just a constant times an even power of a. 

Note that = 1 is necessary in order for the overall scheme to have weighted mean stencils. 

The space of binary uniform stationary schemes is therefore understandable in terms of two components one contains all primal schemes, with an odd number of entries in the mask, which is a linear combination of odd degree B-splines the other all dual schemes, with an even number of entries in the mask, which is a linear combination of even degree B-splines. In principle each is only a countably infinite-dimensional space, and in practice each is only a finite dimensional space because we shall not wish to include B-splines above some maximum degree in order to keep the support limited. 

Each component may therefore be viewed as having barycentric coordinates, and any choice of a specific set of barycentric coordinates is the design of a scheme. 

The coordinate values themselves become the coefficients q giving the terms in the kernel, and so translating from them to the mask itself is fairly trivial arithmetic. 

---

Table 2 shows the design layout in terms of actual factor values and viscosity results from each experiment at room temperature. A set of candidate points in the design space is selected using the D-optimal criterion. In this work, 25 candidate points have been selected. 

Note, however, there are two critical limitations to these "predicting" procedures. First, the mathematical models must adequately fit the data. Correlation coefficients (R ), adjusted for degrees of freedom, of 0.8 or better are considered necessary for reliable prediction when using factorial designs. Second, no predictions outside the design space can be made confidently, because no data are available to warn of unexpectedly abrupt changes in direction of the response surface. The areas covered by Figures 8 and 9 officially violate this latter limitation, but because more detailed... 

Methods employing low sample volumes (< pL) have been explored, which are aimed at the high-throughput and cost-effective exploration of the design space for chromatographic separations, especially related to bioanalytical techniques focused on large biomolecules analyses (proteins and peptides). 

Rosas J.G. Blanco M. Gonzalez J.M. Alcala M. Quality by design approach of a pharmaceutical gel manufacturing process, part 1 Determination of the design space. Journal of Pharmaceutical Sciences, 2011, 100 (10), 4432-4441. 

Figure 4.7 is the mean predicted response at a constant x calculated over the design space ofz, pxzis a predicted value ofy at a certain combination of x and z... 

Predictions are made on grid points in the design space. These predictions can in turn be used to predict resolutions. 

The construction of an experimental design for this separation problem is complicated because both mixture and process variables are present. The former variables, which describe the composition of a mixture in terms of fractions, usually result in design spaces which are a subspace of a simplex (e.g. of a triangle or a tetrahedron). Process variables, on the other hand, are really independent. The design space is often a square or a cube. In this paper there are four mixture variables and two process variables. The design space is therefore a part of a tetrahedron in the mixture space, and a square in the process variables space. 

The construction of models by multiple regression is, in the design space used, somewhat more complicated than usual. This is due to the presence of both mixture and process variables. In this section a short discussion on the consequences of the construction of models with both kind of variables is given. 

The design space was scanned with a stepsize of 1% to find the maximum of the minimum resolution. It appeared that the maximum value of the minimum resolution (2.1) occurred at composition DEA=0, MeOH=0.18, CHCl3=0, EtAc=0.82. The K values of the solutes at this composition are given in Table 6.6. 

In this section first some preliminary study is made of the robustness of some mixture compositions. After that a systematical examination of the design space is performed. 

In Figure 6.5 the effect of temperature and relative humidity on the minimum resolution between the spots is showii for solvent composition DEA=0.217, MeOH=0.267, CHCl3=0.258, EtAc=0.258 (the centre of the design space). It is clear that there are different elution orders, 9 in total. This is mainly caused by quinine. At 25 °C quinine is the second worst eluting solute at 40% RH and best eluting at 80% RH. At intermediate relative humidities it crosses many other spots. The final interpretation is that this solvent composition does not result in a robust TLC separation. 

Predictive models make it possible to perform true process scale-up, which consists of the use of a predictive model to find quantitative criteria for establishing process similarity across scales. The model is also used to determine the changes in both the design space and the target function across scales, and to predict optimum conditions of manufacturing facilities yet to be built. 

As described within the Q8 guideline, a design space is the multidimensional combination and interaction of input variables and process parameters that have been demonstrated to provide assurance of quahty. So long as process control is maintained within the bounds of the design space, operating parameters can be adjusted to improve product quality or manufacturing efficiency. Based on the... 

Product quality attributes can be accurately and reliably predicted over the design space established for materials used, process parameters, manufacturing, environmental, and other conditions. 

Relevant critical-to-quahty product attributes should be considered by the design space model (e.g., content uniformity, bioavailability, stability). 

Borrowing from a famous quote by Albert Einstein, the (design space) model should be as complex as necessary (for accurate prediction), but no less. 

In the same way that in vitro-in vivo correlation (IVIVC) is required to be granted a biowaiver for implementation of postapproval changes, the ability of the design space model to predict the quality of finished goods must be validated prior to implementation. 

If the accuracy of the design space model cannot be established a priori with statistical significance within portions of the parameter hyperspace, operation in such regimes should initiate supplementary quality assurance (inspection) activities until the design space model can be updated and revalidated. 

If unacceptable product quality is observed during operation within a region of the design space expected to yield acceptable quality, the design space should be considered unsuitable for process control (due to drift or the appearance of new factors in the parameter space) until the missing factor(s) can be identified and incorporated into the model and the model is revalidated. 

Explore the design space. Assume that the reactor dimensions and temperatures are fixed. However, the flow rates and the flow direction need to be established. 

---