Sensor response model

To define a feature extraction procedure it is necessary to consider that the output signal of a chemical sensor follows the variation of the concentration of gases at which it is exposed with a certain dynamics. The nontrivial handling of gas samples complicates the investigation of the dynamics of the sensor response. Generally, sensor response models based on the assumption of a very rapid concentration transition from two steady states results in exponential behaviour. 

Further sensor response modelling approaches are in development for other multi-gas sensor pairs. In summary, it is emphasized that accurate determination of plume gas ratios from co-deployed in situ electrochemical (and other) sensors requires a consideration of sensor response times. Non-identical sensor response times can result in scatter and bias in the derived gas ratios, particularly in cases where cross-sensitivities need to be removed. Reported plume gas ratios from multigas instruments may need to be revisited in this context, and the effect is likely also important in the monitoring of other environments, such as urban pollution. 

Over the last several years, the number of studies on application of artificial neural network for solving modeling problems in analytical chemistry and especially in optical fibre chemical sensor technology, has increase substantially69. The constructed sensors (e.g. the optical fibre pH sensor based on bromophenol blue immobilized in silica sol-gel film) are evaluated with respect to prediction of error of the artificial neural network, reproducibility, repeatability, photostability, response time and effect of ionic strength of the buffer solution on the sensor response. 

Guided mode calculations were also carried out to compare the sensor response of several waveguide systems. In these simulations a model molecular monolayer is represented by a 2-nm thick layer with a refractive index of n 1.5. The optical properties of this model layer are typical of a dense layer of organic molecules on a substrate1 41, and are a reasonable approximation for a streptavidin protein layer bound to a biotinylated surface, the experimental model system we use to characterize our sensors. The ambient upper cladding was assumed to be water with a refractive index of n 1.32. For all examples, the lower cladding was assumed to be Si02 with an index of n 1.44. In the simulations, the effective index of... 

The simplest way to estimate a supervised model is to consider that the descriptor of each class may be represented as a linear combination of the sensor responses. Considering N sensors and M classes the expressions can be written as ... 

In order to better investigate the relationship between sensor response and interaction mechanism it is useful to consider the way in which each volatile compound is expected to interact when in contact with a solid phase. These interactions can be modelled using the linear sorption energy relationship approach (LSER) [23]. 

The model sensitive layer, which will be used for gas sensor performance tests throughout this book, was Sn02 that has been doped with 0.2 wt % Pd. The minute Pd-content leads to a better sensitivity to carbon monoxide. The larger response is a consequence of the increased reaction rate. For the sensor arrays in Chap. 6, two additional materials have been prepared. Pure tin oxide shows a good sensor response... 

First, an estimate can be made of expected response time from Equation (11). Tube length is L = 10 cm and D = 0.175 cm2/sec. Using these values, t = 1 3 seconds was computed. Secondly, the steady state current was estimated for 1 ppm CO by means of Equation (12). Note that tube area A = 2.85 cm2, nF = 2 x 96,500 coul/mole and 1 ppm CO is equivalent to k.OQ x 10 11 moles/cm3. Using these numbers, a sensor response of 0.29 /ppm CO was calculated. The simplified diffusion model provided predictions of operating characteristics which were sufficiently promising to proceed with an experimental study. 

Antecedents of the treated topic can be traced to the building of response models for arrays of ISEs, which considers the case of crossresponse terms. This has been historically addressed by the application of different chemometric tools. The first attempt was by Otto and Thomas [40] in the 1980s, who employed an eight-sensor array and... 

Partial least square (PLS) regression model describes the dependences between two variables blocks, e.g. sensor responses and time variables. Let the X matrix represent the sensor responses and the Y matrix represent time, the X and Y matrices could be approximated to few orthogonal score vectors, respectively. These components are then rotated in order to get as good a prediction of y variables as possible [25], Linear discriminant analysis (LDA) is among the most used classification techniques. The method maximises the variance between... 

A completely different model is given by Richards and Parks (1971). It is based on modified von Kries coefficients. If we assume that the sensor response functions have the shape of delta functions, then it is possible to transform a given color of a patch taken under one illuminant to the color of the patch when viewed under a different illuminant by multiplying the colors using three constant factors for the three channels. These factors are known as von Kries coefficients, as described in Section 4.6. The von Kries coefficients are defined as... 

TISSUE AND MICROVASCULAR DIFFERENCES BETWEEN NORMOGLYCEMIC AND DIABETIC INDIVIDUALS THAT MAY BE RELEVANT TO SENSOR RESPONSE AND THE NEED FOR A DYNAMIC IMAGING MODEL... 

The effect of this subtle difference in device function can be seen when the measured signal in the presence of biofouling is modeled. As a model patient, we considered the transient response of an individual with basal insulin provided after each of the three daily meals. Blood glucose dynamics predicted by Sorensen was corrected for diffusion to subcutaneous tissue using the mass transport model of Schmidtke et al.24 25 Figure 11.1 shows a model comparison between the sensor response of an electrochemical sensor and an optical sensor with an assumed... 

The process model can be obtained by different forms, and in bioprocesses mass balance equations canprovide much information. However, in order to have efficient process models and software sensors, a previous adjustment of the model is necessary using on-line data collected from a plant under different operational conditions. This databank is important to guarantee that the model remains calibrated and represents the plant adequately. Some requisites are indispensable for the experimental implementation of models in software sensors response speed to disturbances in the system and appropriate inference of primary variables of interest during key points of the process. 

Perturbation mechanisms for the various acoustic devices were discussed in general terms in Chapter 3. In this chapter, these mechanisms are reviewed specifically in the context of chemical and biochemical analysis. Performance criteria are discussed, and the fundamental coating-analyte interactions giving rise to sensor responses are presented as a basis for classification. Relevant physical and chemical models of these interactions are described, and examples of analytical applications employing each type of interaction are given to illustrate their advantages and limitations. While references have been included to illustrate specific points, this chapter is not intended to comprise an exhaustive review of the literature, particularly for TSM resonators, for which the number of references is far too great to be fully reviewed here. For more detailed information on the diversity of sensor applications, the reader is referred to the many review articles that have been published on these topics [2-8,13-15]. 

The relative importance of the mass-loading and viscoelastic contributions to the observed acoustic sensor response is an issue that has yet to be resolved. Capitalizing on these effects to improve chemical selectivity and detection sensitivity requires further characterization of sensor response, in terms of both velocity and attenuation changes, in addition to more accurate models describing how coating-analyte interactions affect relevant film properties. 

The relationship between the collected data was described by plotting scores of relevant principal components of the PCA model vs. each other as shown in Fig. 5.11. These plots demonstrated that responses of the sensor film to different vapors were well-separated in the PCA space. By plotting PC 1 vs. PC 2, some drift effects were pronounced (Fig. 5.11a). However, the drift effect was removed by plotting PC 1 vs. PC 3 (Fig. 5.11b). The remaining response scatter in regions 1 and 2 of Fig. 5.11b was due to nonequilibrated sensor responses during the kinetic experiments. 

Once a decision of the chemical functionality or host structure is made and a sensing film is included in a sensor device, the next goal would be to model the sensor response of the film in the device. Sensor response to an analyte is a complex function of the partitioning of the target analytes based on the interactions within the film as well as the transport properties of the analyte in the sensor. The sensor responses for polymer-based sensors have been modeled by various approaches using (1) first principles techniques such as Hansen solubilities, (2) multivariate techniques such as QSAR to correlate sensor response with molecular descriptors, and (3) simulations and empirical formulations used to calculate the partition coefficient, such as linear solvation energy relationships, to provide a measure of selectivity and sensitivity of the material under consideration. 

Muehlbauer et al. (1989) attached a Bi/Sb thermopile directly to a membrane containing immobilized GOD and catalase. The basic sensor was assembled by vacuum deposition of the metals. The dynamic behavior of the thermoelectric glucose sensor was modeled. The results properly reflected the sensor response. 

Non-linear sensor responses can be modeled using linear equations with the same assumptions as the ordinary least-squares approach. Using a polynomial to fit curves is better in most cases than using transformations to linearize the data. For example, the following equation can be used to estimate a curvilinear calibration curve ... 

The more common approach for sensor array calibration uses the direct regression model where the independent variables are the analyte concentrations and the dependent variables are the sensor responses. 

---