Software Sensors

On-Line Monitoring From Laboratory-Transposed Methods to Software Sensors... 

Software sensors and related methods - This last group is considered because of the complexity of wastewater composition and of treatment process control. As all relevant parameters are not directly measurable, as will be seen hereafter, the use of more or less complex mathematical models for the calculation (estimation) of some of them is sometimes proposed. Software sensing is thus based on methods that allow calculation of the value of a parameter from the measurement of one or more other parameters, the measurement principle of which is completely different from an existing standard/reference method, or has no direct relation. Statistical correlative methods can also be considered in this group. Some examples will be presented in the following section. 

As seen previously for some specific applications such as wastewater treatment plants, software sensors can be envisaged to provide on-line estimation of non-measurable variables, model parameters or to overcome measurement delays [81-83]. Software sensors have been developed mainly for monitoring bioprocesses because the control system design of bioreactors is not straightforward due to [84] significant model uncertainty, lack of reliable on-line sensors, the non-linear and time-varying nature of the system or slow response of the process. 

Table 11 Software sensor main types for wastewater treatment applications... 

V. Alcaraz-Gonzalez, J. Harmand, A. Rapaport, J.P. Steyer, V. Gonzalez-Alvarez, and C. Pelayo-Ortfz. Software sensors for uncertain wastewater treatment processes A new approach based on interval observers. Wat. Res., 36(10) 2515-2524, 2002. 

O. Bernard, D. Dochain, A. Genovesi, A. Punal, D. Perez-Alvarino, J.P. Steyer, and J. Lema. Software sensor design for an anaerobic wastewater treatment plant. In IFAC-Intemational Workshop on Decision and Control in Waste Bio-Processing, Narbonne, Prance, 1998. 

The presented approach has been applied on the process illustrated in Figure 8. In this case, several physical redundancy relations are available but the same method can be successfully applied with a standard instrumentation fulfilled by software sensors [17]. r, T2, rs, re, rr and ril are based on physical redundancy, r4 and rs on functional redundancy, rg, rg and riO are derived from dynamical modeling. As expressed in Table 2, there is not one single residual that is representative of a specific fault. A diagnosticability analysis of the signature table shows that among the 256 different states of the system ... 

G.A. Montague, A.J. Morris and M.T. Tham, Enhancing bioprocess operability with generic software sensors, J. Biotechnol, 25, 183-201 (1992). 

A software sensor for on-line determination of substrate was developed based on a model for fed-batch alcoholic fermentation process and on-line measured signals of ethanol, biomass, and feed flow. The ethanol and biomass signals were obtained using a colorimetric biosensor and an optical sensor developed in previous works that permitted determination of ethanol at a concentration of 0-40 g/L and biomass of 0-60 g/L. The volume in the fermentor could be continuously calculated using the total measured signal of the feed flow. The results obtained show that the model used is adequate for the proposed software sensor and determines continuously the substrate concentration with efficiency and security during the fermentation process. 

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. 

Court (2), Eberhard (3), and Tyagi et al. (4) have reported some applications of computers and software sensors for fermentation control in experimental research in data acquisition of bioreactors. Neural network models were used to interprete sensor signals in the control of an alcohol fed-batch fermentation (5) and in the detection of the individual components of a gas mixture and to measure the concentration of both gases (6). 

This article presents the design and implementation of a software sensor for the continuous determination of substrate concentration based on a simple model of a fed-batch fermentation process and the available signals of two other sensors—one for on-line biomass determination (7) and the other for on-line ethanol determination (8)—developed in previous works. The software sensor proposed provides a continuous signal that can be used in a control loop to manipulate the substrate feed flow in order to maintain almost constant substrate concentration and obtain an excellent level of productivity and yield during all of the process, as shown in experimental control strategy studies in previous works (9). 

Figure 3 shows the experimental results of tests 1 and 2 obtained for the substrate concentration calculated using the software sensor and the filtered sensor measurements for ethanol and biomass. The differences between the two experiments are the initial biomass concentration and concentration of substrate in the feed, as shown in Table 1. 

Software sensors are virtual sensors which calculate the desired variable or parameter from related physical measurements [58]. In other words, there must always be a model available that relates reliably the measured variable with the target variable or parameter. Normally, measured variables are easily measurable effects that are caused and influenced by the target. The most prominent software sensor is the respiratory quotient (RQ-value) which characterizes the physiological state of a culture. However, its determination can be tricky (see Sect. 5). 

Generally, software sensors are typical solutions of so-called inverse problems. A so-called forward problem is one in which the parameters and starting conditions of a system, and the kinetic or other equations which govern its behavior, are known. In a complex biological system, in particular, the things which are normally easiest to measure are the variables, not the parameters. In the case of metabolism, the usual parameters of interest are the enzymatic rate and affinity constants, which are difficult to measure accurately in vitro and virtually impossible in vivo [93,118,275,384]. Yet to describe, understand, and simulate the system of interest we need knowledge of the parameters. In other words, one must go backwards from variables such as fluxes and metabolite concentrations, which are relatively easy to measure, to the parameters. Such problems, in which the inputs are the variables and the outputs the parameters, are known as system identification problems or as so-called inverse problems. 

The undelayed evaluation of state of a culture by using software sensors and computers, based on the quantitative analytical information provided by hardware sensors and intelligent analytical subsystems, constitutes an excellent basis for targeted process control. Experts - either human or computer - have the data and the deterministic knowledge to trace observed behavior back to the physical, chemical and physiological roots thereby gaining a qualitative improvement of bioprocess control, a quantum leap process control can act on the causes of effects rather than just cure symptoms. A simple standard operating procedure [398] has proven useful, namely ... 

Because of their predictive capabilities, models are also essential tools in modem biochemical engineering for the design of processes and the optimization of media and reactor operational parameters in batch or continuous operation. They can also serve in the development of software sensors to estimate on-line the time variation of the medium composition. 

The use of an infrared sensor (Aquasant Messtechnik AG, Bubendorf, Switzerland) may be a solution for many apphcations, particularly in combination with a control of viability (Merten et al, 1987). Other alternatives are sensors measuring conductance/capacitance (ABER Instruments, Cefnllan, Aberystwyth, UK) and software sensors (Pelletier et al, 1994). For further reading see De Gouys et al (1996). A recently pubhshed method based on real-time imaging opens up new possibihties by real cell counting (Ozturk et al, 1997). 

Pelletier, F, Fonteix C, Lourenco-Da-Silva A, Marc A Engasser JM (1994) Software sensors for the monitoring of perfusion cultures evolution of the hybridoma density and the medium composition from glucose concentration measurements. Cytotechnology 15 291-299. 

The software sensor for (pi is an observer-based estimator with the following structure ... 

Some of the properties that cannot be measured on-line can be inferred fi"om the measurements of other variables by means of state estimation methods and software sensors... 

Kiviharju, K., Salonen, K., MoUanen, U., Eerikainen, T. (2008). Biomass measurement online the performance of in situ measurements and software sensors. Journal of Industrial Microbiology and Biotechnology, 35, 657-665. 

Barresi, A. A Velardi, S. A., Pisano, R., Rasetto, V Vallan, A Galan, M., 2009c. In-line control of the lyophilization process. A gentle PAT approach using software sensors, fnt. J. Refiig. 32 1003-1014. 

Furthermore, state estimation techniques such as the boot-strapping method developed by BenAmor et al. [14] allow one to account for fouling in real time. The method assumes that the overall heat transfer coefficient (multiplied by the surface) changes very little in the course of the time required to measure the reactor temperature. Thus, the reactor temperature is used to calculate first a heat generation rate over the course of 1-2 s, assuming constant heat transfer conditions, then a heat transfer coefficient assuming constant rate for a very short time. If fouling occurs over a reasonable timescale, it can be accounted for by the software sensor in-line. 

The safety barrier analysis principle requires the analysis to start at the potential hazardous energy concentration e.g. rocket motor, igniter, firing mechanism. Causal paths are traced backward from this, moving through the active hardware. Control electronics, C software, sensors operator actions, to the environment in which the C tystems works. At each stage, the safety barriers, which prevent accidents, are noted [6] (see fig. 1). 

As part of this process, we need to extend our view of the system we are seeking to assure. In the early days, the focus was on the technical system (hardware, software, sensor and actuators). In the future we need to think about the larger socio-technical system that includes the management, people and processes that interact with the technical system. 

---