Data processing: artificial intelligence – Neural network – Learning task
Reexamination Certificate
2000-01-18
2002-08-27
Davis, George B. (Department: 2121)
Data processing: artificial intelligence
Neural network
Learning task
C706S015000
Reexamination Certificate
active
06442536
ABSTRACT:
FIELD OF THE INVENTION
This invention is generally related to predicting the flammability limits of chemical mixtures. More specifically, this invention relates to the use of neural networks as a model to predict the flammability limits of complex chemical mixtures.
BACKGROUND OF THE INVENTION
A vapor mixture containing combustible gaseous compounds (“fuels”) and an oxidant (typically oxygen or oxygen containing gas) may be flammable if the mixture composition and conditions are such to sustain a flame upon ignition of the mixture. Often, such vapor mixture may contain an inert gas forming a tertiary system. The ignition of a flammable mixture results in propagation of a flame to the surrounding unburned fuel-oxidant-inert mixture, a rapid rise in pressure, and the potential for severe damage to equipment and/or injury or death to humans. Therefore, an understanding of the flammability characteristics of a ternary system containing a fuel(s), oxidant(s), and an inert(s) is essential to the prevention and/or mitigation of industrial explosions. In particular, the safe design, operation, and/or optimization of industrial process, which handle potentially flammable mixtures, rely on the knowledge of these flammability limits.
The composition of a mixture under a given set of conditions which are necessary to achieve a sustained flame define the flammability limit of a vapor mixture. Therefore, those factors, which influence reaction, heat, and mass transfer during combustion of the fuel-oxidant mixture, will impact the values of the flammability limits. Because a flammable mixture must contain a fuel and an oxidant, and may also contain an inert, defined as species which does not typically participate in the combustion reaction (i.e., N
2
, Ar, He, CO
2
, H
2
O), the flammable envelope and its boundary (the flammability limits) are typically illustrated using ternary flammability diagrams. A typical ternary flammability diagram for a mixture containing a single fuel, a single oxidant, and an inert, at a given temperature and pressure is illustrated in FIG.
1
. The flammable mixtures lie within the “flammability envelope”, which is bounded above by the upper flammability limit (UFL), upper explosive limit (UEL), or the maximum oxygen concentration (MOC) and below by the lower flammability limit (LFL) or lower explosive limit (LEL). These two boundaries meet at a point referred to as the limiting oxygen value (LOV), the oxygen concentration below which no mixture of fuel, oxidant and inert is flammable.
Although the prior art addresses the measurement and prediction of flammability limits, it focused predominantly on simpler mixtures such as single fuels in air with single or two-component inert systems at or near normal temperature and pressure (25C and 1 atm). However, many industrial applications, including chemical processes, inerting, storage and transportation of flammable compounds, and many others, handle vapor mixtures containing an oxidant, oxygen, multiple inerts, and multiple fuels at elevated temperatures and pressure have not been believed to be addressed in the prior art. These mixtures will be referred to as “complex mixtures.”
Typically, the measurement of flammability limits is then required to appropriately design safe systems and processes handling potentially flammable mixtures. However, the characterization of the flammability envelope of a complex mixture through experimental tests can be quite difficult, time consuming, and expensive. An ability to predict the flammability limits of complex mixtures would serve as a very valuable tool to numerous industries. Such a tool would also help when exploring novel processes and process conditions. With a better understanding of the mixture flammability properties, one can pinpoint the conditions of interest, thereby minimizing the time and cost associated with extensive flammability measurements.
However, to develop a predictive model from first principles is a very formidable task. Neural networks, however, can offer a means of modeling these complex, non-linear relationships without detailed knowledge of the fundamental relationships, including thermodynamics, kinetics, heat, and mass transfer, which dictate the flammability behavior of these complex mixtures. This invention describes a novel approach for predicting the flammability limits of complex mixtures using neural networks.
An abundance of flammability limit data and techniques for predicting flammability limits exists in the prior art. A review of the prior art does reveal a number of predictive models which attempt to address the issue of predicting flammability limits. The authors of the relevant prior art have taken a number of approaches, including Le Chatelier's principle, constant adiabatic flame temperature (CAFT), linear and non-linear regression, group contribution techniques, and neural networks to tackle this problem.
Le Chatelier's principle is a traditional and simple approach used often to predicts LFL and UFL's of fuel mixtures in air based on the flammability limits of each fuel and the fuel mixture composition. Another approach is based on the observation that lower paraffins exhibit constant adiabatic flame temperatures at the limits of flammability. A number of empirical and semi-empirical models for predicting the flammability limits, temperature effects, pressure effects also exist in the prior art. Although neural networks have been used extensively to model the non-linear relationships which exist between a certain set of inputs and outputs, neural network based techniques for predicting flammability limits is limited.
For purposes of this invention, Flammability limits shall mean the point in which a flame initiated from an adequate ignition just fails to propagate throughout the fuel/oxidant mixture.
There is believed to be no teaching in the prior art for using neural networks based techniques to predict the upper and lower flammability limits of complex mixtures. It is, therefore, desirable in the art to provide for a method of predicting the flammability limits of complex mixtures.
SUMMARY OF THE INVENTION
This application is directed to a method for predicting the flammability limits of a complex mixture using critical variables of structural groups comprising the steps of training data from the critical variables of each structural groups, the critical variables comprising compositional and thermochemical data from each of the structural groups to produce a neural network model; testing the trained data from the neural network model; and validating the trained and tested data from the neural network to accurately predict the flammability limit of an analogous complex mixture having similar structural groups.
REFERENCES:
patent: 5327357 (1994-07-01), Feinstein et al.
T. Suzuki and M. Ishida, “Neural Network Techniques Applied to Predict Flammability Limits of Organic Compounds”,Fire and Materials, vol. 19, 179-189 (1990).
T. Suzuki and K. Koide, “Short Communication: Correlation between Upper Flammability Limits and Thermochemical Properties of Organic Compounds”,Fire and Materials, vol. 18, 393-397 (1994).
T. Suzuki, “Note: Empirical Relationship Between Lower Flammabilitiy Limits and Standard Enthalpies of Combustion of Organic Compounds”,Fir and Materials, vol. 18, 333-336 (1994).
I. Wierzba, S.O. Bade Shrestha and G.A. Karim, “An Approach for Predictiing the Flammabilitiy Limits of Fuel/Diluent Mixtures in Air”,Journal of the Institute of Energy, Sep., 69, 122-130 (1996).
J.G. Hansel, J.W. Mitchell and H.C. Klotz, Predicting and Controlling Flammabilitiy of Multiple Fuel and Multiple Inert Mixtures,Plant/Operations Progress, vol. 11, No. 4, 213-217 (1992).
T.K. Subramanian and J.V. Cangelosi, “Predict Safe Oxygen in Combustible Gases”,Chemical Engineering, Dec., 108-113 (1989).
W.H. Seaton, “Group Contribution Method for Predicting the Lower and Upper Flammable Limits of Vapors in Air”,Journal of Hazardous Materials, 27, 169-185 (1991).
Akhras Amer Wahid
Wagner Matthew Lincoln
Davis George B.
Lau Bernard
Praxair Technology Inc.
LandOfFree
Method for predicting flammability limits of complex mixtures does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for predicting flammability limits of complex mixtures, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for predicting flammability limits of complex mixtures will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2880965