Controlled multivariate experiments

Controlled multivariate experiments are the most logical, the most scientific, and the most efficient way that scientists know to collect data. Controlled experiments are the scientific... 

All experiments were set np so as to be realistic scenarios with respect to multivariate statistical process control (MSPC). The pilot plant grannlator is operated exactly as the industrial scale counterpart, but most of the experiments inclnded mnch more severe variation than will usually be found in an otherwise stable industrial production sitnation of similar duration (e.g. it is normally not necessary to change nozzle types, or to change prodnct types within snch short intervals). 

There are broadly two uses of chemometrics that interest the process chemist. The first of these is simply data display. It is a truism that the human eye is the best analytical tool, and by displaying multivariate data in a way that can be easily assimilated by eye a number of diagnostic assessments can be made of the state of health of a process, or of reasons for its failure [ 153], a process known as MSPC [154—156]. The key concept in MSPC is the acknowledgement that variability in process quality can arise not just by variation in single process parameters such as temperature, but by subtle combinations of process parameters. This source of product variability would be missed by simple control charts for the individual process parameters. This is also the concept behind the use of experimental design during process development in order to identify such variability in the minimum number of experiments. 

Figure 6.3. Scatter plot of simulation data illustrating an experiment comparing the results of a control group with those of a treatment group. Because of the strong correlation between the two variables, the difference between the groups is readily detected by eye and by multivariate statistics, but not by Student s t-test applied to each variable separately.

Multivariate regression analysis plays an important role in modem process control analysis, particularly for quantitative UV-visible absorption spectrometry and near-IR reflectance analysis. It is conunon practice with these techniques to monitor absorbance, or reflectance, at several wavelengths and relate these individual measures to the concentration of some analyte. The results from a simple two-wavelength experiment serve to illustrate the details of multivariate regression and its application to multivariate calibration procedures. 

However, as noted, these tools are passive. There is no deliberate and specific control of the environment or critical process parameters. These observational tools cannot find and describe cause-and-effect relationships directly. The only way to find these relationships is to conduct a multivariate controlled experiment. 

In the language of statistics, this is a stratified experiment, that is, the samples are grouped according to some criteria. In this case, the criteria is consecutive production of the units. This is not, therefore, a random selection of samples as is required in many statistical methods. It also is not a control chart even though the plot may resemble one. It is a snapshot of the process taken at the time of the sampling. Incidentally, the multivari chart is not a new procedure, dating from the 1950s, but it has been incorporated into this system by Shainen. 

SVA, etc.) for multivariable control problems. In the absence of process models, one must resort to heuristic (rule-of-thumb) approaches. Although these approaches generally are based on prior experience, they also incorporate an understanding of the fundamental physics and chemistry that apply to all plants. In this chapter, several case studies are used to introduce important plantwide concepts. In the final chapter (Appendix H), we present a general strategy for designing plantwide control systems. 

---