HK1081646A - Processing system for remote chemical identification - Google Patents

Description

Processing system for remote chemical identification

no marking

no marking

no marking

Background

This is a formal application that claims priority from provisional application serial No. 60/382,435 filed on 5/22/2002.

Technical Field

The present invention relates to an apparatus and method for identifying unknown chemical compounds under field conditions. In particular, the present invention relates to apparatus and methods for identifying chemical compounds using remote passive infrared spectroscopy in combination with spectroscopic data.

There are often situations where emergency response personnel ("first response personnel") or military personnel are called to the scene of an accident or emergency for some type of chemical leak and these personnel are faced with a gaseous cloud or smoke whose chemical composition is unknown. This may occur when a rail tanker or highway tanker trailer carrying chemicals encounters a traffic accident, or when an unexpected chemical leak occurs at a chemical manufacturing plant. It is clear that knowing the composition of the smoke is important in deciding how to deal with emergency situations, what protective clothing is needed, and whether local inhabitants need to be evacuated. Methods of using remote passive Infrared (IR) spectrometers to attempt to identify chemical compounds in gaseous smoke are known in the art. As used herein, the term "passive" means that the spectrometer does not use a dedicated infrared photon source (as opposed to an "active" system, which uses photons from an optically optimized high temperature source). The term "remote" means that the sample gas of interest is outside the spectrometer, such as smoke at the site of an accident. Closed loop spectral analysis may include an extraction system that examines the sample in an absorption cell, or uses a known photon source and a known path length.

The general concept of IR spectral analysis in this case is illustrated in fig. 1. The passive spectrometer 101 will be positioned so that the gaseous smoke 103 is located between the spectrometer and some background object 105, which provides a source of IR energy. In the case where the background 105 is significantly warmer than the smoke 103, the IR energy emitted from the background 105 will pass through the smoke 103 and be recorded by the spectrometer 101. Since different chemical compounds tend to absorb IR energy at different wavelengths, measuring the relative intensity of different frequencies of IR energy received at the spectrometer will provide information for identifying the compounds in smoke 103. At a certain pointIn some cases, the smoke is warmer than its surroundings, and the compounds in the smoke may emit themselves (rather than absorb) at these IR frequencies. The absorbance is linearly related to the gas concentration by the Beer-Lambert relationship, or "Beer's law". When infrared radiation passes linearly through a gas sample of distance L, its initial intensity (I) is absorbed by the gas₀) Down to the level (I) measured at the spectrometer. Beer's law states that at each infrared frequency, the absorbance is defined by the following relationship:

equation (0)

The detailed functionality of conventional spectrometers is explained in references, such as U.S. patent No.5,982,486, which is incorporated herein by reference and need not be described in detail herein. It is sufficient to understand from fig. 2 that the spectrometer will initially create an "interferogram" (step 107) representing the spatial domain response of its detector to infrared radiation incident on the detector. Application of a fast Fourier transform to the interferogram will form a single sample beam (step 109) representing the total power incident on the infrared detector as a function of infrared frequency, typically in "wavenumbers" or in reciprocal centimeters (cm)^-1) Expressed in units. A graphical representation of the sample single beam spectrum can be seen in the lower trace of fig. 3.

Since passive IR spectroscopy uses a background source with IR energy, the true signal of interest, i.e. the relatively narrow absorption and/or emission band of the gas located between the spectrometer and the background, is superimposed on the smooth, broad emission spectrum of the background IR source. Therefore, it is necessary to develop some type of background spectrum. While the method of forming the background spectrum of the present invention is described below, the middle trace of fig. 3 illustrates the background single beam spectrum in graphical form, which is helpful in conceptually understanding the background of the present invention.

Once the sample single beam spectrum (step 111) and the background single beam spectrum (step 112) are determined, the sample absorption spectrum can be calculated. In the notation used hereinafter, the sample absorption spectrum (step 117) is based on a single-beam sample spectrum SB_i ^SAnd single beam background spectrum SB_i ^BIs defined as:

equation (1)

The upper trace of fig. 3 shows the sample absorption spectrum.

Once an absorption spectrum for the sample ("sample absorption spectrum") is created, it can be compared to known absorption spectra for different chemical compounds ("reference spectra"), which are represented in the form of sample absorption spectra. The reference spectrum of a compound can be considered as a graphical representation of the extent of absorption of the chemical compound at these frequencies of absorbing IR radiation. The reference spectrum is, in a descriptive sense, a "graphic stamp" of the chemical compound. There are a number of methods for determining how well a portion of the sample absorption spectrum matches the reference spectrum and thus how well it is to determine how well the reference chemical (i.e., the compound represented by the reference spectrum) is present in the sample gas. One method of comparing the sample absorption spectrum to one or more reference spectra is to use a conventional least squares analysis.

The use of Conventional Least Squares (CLS) analysis is known in the art and is documented in published literature for spectroscopic analysis, such as "Improved Sensitivity of acquired Spectroscopy by the D.M. Haaland and R.G. esterlingquares Methods，” Appl.Spectrosc.34(5): 539-548 (1980); haaland, R.G.Esterling, "Application of New Least-Square methods for Quantitative innovative Analysis of Multi-component Samples," (see FIGS.) "Appl.Spectrosc.36(6): 665-673 (1982); haaland, R.G.Esterling and D.A.Vopicka "Multivariate blast-Square Methods Applied to the quantitative Spectral Analysis of Multi component Samples"Appl. Spectrosc.39(1): 73-84 (1985); hamilton's of W.CStatistics in Physical ScienceRonald Press co., New York, 1964, chapter iv, and references therein; and U.S. patent No.5,982,486, all of which are incorporated herein by reference. Although a complete mathematical description of CLS is disclosed in the above references, a simple description, particularly with respect to matrix operations, will provide useful context.

CLS analysis is commonly used to infer solutions to overdetermined systems with linear equations; the equations of this system are typically expressed as matrix equations having the form:

DX + E equation (2)

Wherein:

A	＝	set of N measurement data, represented by a row vector
			X	＝	Set of M parameters to be evaluated, represented by a column vector
D	＝	"design matrix", having N rows and M columns, describing a linear mathematical relationship between measurement data A and parameter X
			E	＝	The error of the linear model for each measurement data is represented by a row vector having a length N

Since it involves comparing a sample spectrum to one or more reference spectra, the matrix:

will represent the sample spectrum, each element A of A_i ^SIs shown in the frequency range v₁，v₂，...v_N]Wave number v of_iThe following intensity values. In the design matrix:

equation (3)

Each column will represent a reference spectrum, while column A_i1 ^RThe intensity values of the reference spectrum as absorption spectrum over the same frequency range are represented (using the first column as an example).

In the case where N > M (i.e., when the number of measurement data exceeds the number of parameters to be estimated), the system of equations described in equation 2 is referred to as "overdetermined". In this case, which applies to all CLS applications described herein, there is no unique solution to equation 2. In this case, however, it is possible to form estimates of the parameter X and to describe the accuracy of these estimates. The estimates and characterization may be based on any selected set of mathematical criteria and constraints. The widely used criterion is aimed at forming an estimate of the parameter X which minimizes the "weighted sum of residual squares" for the model of equation 2; the sum may (or may not) be defined in such a way that it describes the measurement data a_iThe quality of (2) is changed. This estimate is broadly referred to as the result of a "least squares" technique, and only this estimate is described in the work herein. The term "conventional least squares" refers to a least squares technique based entirely on the linear model described in equation 2; other least squares techniques, such as those known as "partial least squares" analysis, typically use additional data processing and result in more complex methods for estimating the desired parameter X.

Formally, "traditional least squares" estimation is based onThe assumption is that the error vector E possesses a joint distribution with a variance-covariance matrix M of zero mean and rank N_f. Many CLS estimates also include a further assumption, namely the known matrix M_fAt (non-negative) scale factor σ²In, i.e.

Equation (4)

In fact, equation 2 includes the assumption that the measurement data A_iIs known. The relative quality of the data may be determined by assigning each A_iSpecifying non-negative "weights" P_iiIs quantized, wherein the weight matrix P of the diagonal is (usually) the inverse of N, i.e.

P＝N^-1Equation (5)

And thus, according to equation 5, have

Equation (6)

The type of CLS analysis commonly used in the prior art will be designated as "unweighted CLS" to facilitate differentiation from the "weighted CLS" analysis described in the detailed description of the invention below. Both types of CLS analysis use the assumptions indicated above in conjunction with equations 4, 5 and 6.

In unweighted CLS analysis, all measurement data are assumed to be of the same quality; i.e. in the unweighted case, both matrices P and N are identical to the identity matrix I. For any estimated parameter set X, the residual V is defined as:

v ≡ A-D X equation (7)

And "weighted sum of residual squares" is defined as:

V²≡V^tPV＝(A-D X)^tp (A-D X) equation (8)

Where the superscript "t" denotes the transpose of the matrix.

The following estimates of the parameter X exist, are unique, and let V²Minimum ("weighted sum of residual squares"):

X＝(D^tPD)^-1D^tPA equation (9)

Equation 9 describes basic CLS parameter estimation, which is useful and accurate in many applications. It should be noted, however, that all CLS analysis also provides a useful statistical measure of uncertainty in parameter estimates.

In particular, the CLS analysis provides a "marginal standard deviation" (MSD) for each parameter estimate. Wherein the CLS estimate of the variance-covariance matrix is

M＝V²(D^tPD)^-1Equation (10)

With an estimate X of each parameter_jThe associated Marginal Standard Deviation (MSD) is

Equation (11)

MSD is sometimes referred to as "1 σ uncertainty" in the relevant parameters. X_jAnd Δ_jIs typically used as the parameter estimate X for the CLS analysis_jThe quality of (c).

Weighted CLS analysis, which forms part of the invention discussed below, wherein at least one P_iiUnlike the other values in the matrix P, i.e. assuming at least one datum A_iHaving a better or worse than the other A_iThe quality of (c).

In addition to the spectroscopic aspects of the present invention, the present invention also relates to novel methods for processing spectroscopic data and identifying chemical compounds based on the location of chemical leaks and the conditions (e.g., color, odor, etc.) observed at the chemical leak site. This non-spectroscopic method of identifying possible chemical compounds can be used in combination with or independently of spectroscopic methods. When used in combination with spectroscopic methods, non-spectroscopic methods will be used to identify or assist in identifying an initial list of chemical compounds, the reference spectrum of which will be selected for comparison with the undetermined sample spectrum.

Background

One embodiment of the present invention provides a system for remotely identifying chemical compounds. The system includes a passive infrared spectrometer, a locator, a range finder, and a user interface. The system further includes a database having data representing reference spectra of the chemical species, data associating observable properties with certain chemical compounds, data associating locations with certain chemical compounds. The computer processor communicates with the spectrometer, the locator, the rangefinder, the user interface, and the database. The system also includes software for comparing data from the spectrometer, the localizer and the rangefinder to data in a database in order to identify the possible presence and concentration of one or more chemical compounds.

A second embodiment of the invention is a method of determining the temperature of a background object against which a sample spectrum is recorded. The method comprises the following steps: a. providing a predetermined relationship between a parabolic center frequency and a background temperature, wherein the parabolic center frequency is the frequency of a single beam spectrum of a reference background at a known temperature; b. providing a sample spectrum recorded against a background of unknown temperature; c. determining a parabolic curve which is most matched with the spectrum; d. determining a center frequency of a sample parabola of the best fitting parabolic curve; e. comparing the sample parabolic center frequency to a predetermined relationship of center frequency and background temperature; estimating the temperature of the background based on the comparison result.

A third embodiment of the invention is a method of generating a background spectrum for spectral analysis. The method comprises the following steps: a. providing a sample spectrum and an estimated temperature of a background object; b. selecting at least two known temperature spectra, representing background temperatures above and below the estimated temperature, from a set of known temperature spectra for known background temperatures; comparing the sample spectrum to a known temperature spectrum to facilitate determination of a sample background spectrum.

A fourth embodiment of the invention is a method of generating a temperature compensated absorption spectrum. The method comprises the following steps: a. providing a sample spectrum and an estimated temperature of a background object; b. selecting at least two known temperature spectra, representing background temperatures above and below the estimated temperature, from a set of known temperature spectra for known background temperatures; c. comparing the sample spectrum to a known temperature spectrum to facilitate determination of a sample background spectrum; calculating an absorption spectrum from the sample spectrum and the background spectrum.

A fifth embodiment of the present invention is to identify the chemical species represented in the absorption spectrum. The method comprises the following steps: a. providing a sample absorption spectrum; combining the absorption spectrum with CO₂And H₂The spectrum of O is compared and compared to at least one chemical reference spectrum to identify the chemical represented in the absorption spectrum.

A sixth embodiment of the invention is a computer system for identifying a chemical using observable features. The system includes a chemical database that correlates predefined characteristics and attributes with a plurality of chemical compounds, and a user interface that receives input of observable characteristics. The system further includes a processor running software that directs the processor to identify chemical compounds in the database that correspond to the observable feature.

A seventh embodiment of the invention is a computer system for identifying chemical compounds based on the observed position of the compound. The system includes a chemical/location database correlating at least one chemical compound with a map of the location of the compound dispensed, and a location input for inputting the location of the system into the system. A processor running software that directs the processor to identify chemical compounds in the database that correspond to the systematic locations.

Drawings

FIG. 1 is a conceptual illustration of a spectrometer, a chemical compound cloud, and a background temperature source.

Fig. 2 illustrates graphically the prior art signal processing steps for obtaining an interferogram and converting it into an absorption spectrum.

Fig. 3 shows three different spectral curves relating to the present invention.

Fig. 4 is a conceptual hardware block diagram of the system of the present invention.

FIG. 5 is a flow chart of the "Chemical Properties wizard" relating to the present invention.

Fig. 6 is a schematic diagram of a chemical property database of the present invention.

FIG. 7 is a flow chart of the Facility Wizard for the present invention.

Fig. 8 is an example of a container profile used in the present invention.

FIG. 9 is a screen shot of the NFPA ticket selection subroutine used in the present invention.

Fig. 10 is a flowchart of the "railroad car Wizard" (Railcar Wizard) of the present invention.

FIG. 11 is a screen shot of a DOT ticket selection subroutine used in the present invention.

Fig. 12 is a screen shot of a tank trailer profile selection subroutine used in the present invention.

FIG. 13 is a flow chart illustrating the steps of obtaining a spectral reading.

FIG. 14 is a flow chart illustrating the steps of calibrating a spectrometer.

FIG. 15 is a flow chart illustrating the steps of estimating an initial background temperature.

Fig. 16 is a flowchart illustrating a step of weight iteration CLS.

FIG. 17 is a flow chart illustrating the steps of estimating the final background temperature.

FIG. 18 is a flow chart illustrating the steps of generating a background and absorption spectrum.

FIG. 19 shows a series of black body temperature curves.

Fig. 20 shows a parabola fitted to two black body curves.

FIG. 21 shows a plot of parabolic peak position versus black body temperature.

Fig. 22 is a flowchart illustrating the steps of the CLS search of the present invention.

Fig. 23a and 23b are analysis range diagrams.

FIG. 24 shows a floor map for analyzing region selection and levels.

Fig. 25 is a flowchart illustrating the steps of the galcanic spectral search.

FIG. 26 is a flow chart illustrating the steps of another chemical species discrimination spectral search.

Fig. 27 is a flowchart illustrating steps of adjusting the CLS process.

Fig. 28 is a graph illustrating contrast scale factor versus temperature contrast.

Fig. 29 is a flow chart illustrating the steps of a sample acquisition subroutine used in the present invention.

Detailed Description

1. Non-spectroscopic identification of compounds

As described above, the present invention relates not only to novel apparatus and methods for processing spectroscopic data, but also to novel methods of identifying chemical compounds based on "non-spectroscopic" characteristics, such as the location of a chemical leak or the conditions (e.g., color, odor, etc.) observed in the field of a chemical leak.

To perform the spectral discrimination method and non-spectral discrimination method of the present invention, the present invention uses a unique combination of hardware to form a system for remotely discriminating compounds present in a chemical leak. Fig. 4 is a block diagram illustrating the hardware used in the remote authentication system of the present invention. In general, the remote chemical identification system 1 will include a spectrometer 3, a locator 7, a range finder 5, a user interface 10, a computer processor 4, a weather station 12, a camera 9 and a database 13 of chemical reference spectra. Although not shown in fig. 4, it should be understood that the spectral search engine will run on the processor 4. An example of the spectral search engine is described below with reference to FIGS. 15-27. In a preferred embodiment, spectrometer 2 will comprise a passive infrared spectrometer, such as the spectrometer sold under the trademark "Illuminator" by MIDAC Corporation of Costa Mesa, CA. The locator 7 will be a conventional Global Positioning System (GPS), such as the device sold under the trade mark "Svee 8 Plus" by Trimble corporation of Sunnyvale, CA. The rangefinder 5 will typically be a Laser-based rangefinder such as that sold under the trade mark "Impulse XL" by Laser Technology of Englewood, CA. The camera may be of any conventional type, and in the preferred embodiment is sold as "VK-C77U" by Hitachi corporation. The compass 14 may be any type of compass that provides an electronic signal output, such as the HMR3000 sold by Honeywell corporation. A Weather Station is a device capable of measuring Weather conditions such as wind speed, ambient temperature and humidity, and one suitable example is the "1-Wire Weather Station" sold by Texas Weather Instruments of Dallas, TX. In a preferred embodiment, the user interface 10 includes a conventional touch screen 11 and an alphanumeric keypad. The processor 4 should be at least a 1GHz processor and be equipped with at least 512MB of RAM and a large capacity memory, such as a hard disk drive. In a preferred embodiment, the spectrometer 3, the locator 7, the range finder 5, the user interface 10, the computer processor 4, the weather station 12 and the camera 9 are organized to be mounted in a housing (not shown) that is typically adapted to be mounted on a vehicle instead of in a fixed position, or on a support structure (e.g., a tripod) for a separate application. The database 11 may reside on a hard disk drive or may be remotely located and accessed through, for example, a wireless data link or an optical fiber link.

The database 13 will include, among other information, a library of chemical reference spectra. As described above, a reference spectrum of a chemical compound provides the absorbance of the compound at a particular wavelength at which the compound absorbs or emits IR radiation. Databases of reference spectra are well known in the art and are available from companies such as Thermo practical of Salem, NH. However, chemical compound identification based on spectroscopic data forms only one aspect of the present invention. In a preferred embodiment, the database will include not only chemical spectral data, but also data relating observable properties to specific chemical compounds and/or data relating locations to specific chemical compounds. The term "database" is not limited to data on a separate storage device. For example, the database may include portions of information on a local storage device (e.g., hard disk) and portions of information remotely located (e.g., remote server). Alternatively, the database may reside entirely on a storage device remote from the other components of the system.

The present disclosure first details a method for identifying chemical compounds based on certain attributes, conditions, or characteristics observed at a chemical leak site. The method is typically implemented in a computer system that includes a processor, a user interface that can execute software programs, and a database that associates observable properties with specific chemical compounds. Referring to fig. 4, the processor and user interface may be of the type described above. The database will include a plurality of chemical compounds and for each compound a series of observable attributes or characteristics are correlated therewith. These properties are usually related to the compound in the gaseous state, but in some cases, to the compound in the solid or liquid state. In the specific embodiments described herein, these attributes may include whether the compound produces a visible vapor, whether the vapor sinks or rises, the color of the vapor, any odors associated with the compound, and any color and odor changes upon exposure of the compound to water (hydrolysis). An additional attribute that may be included in the data is what secondary compounds are brought about by incomplete combustion of the suspected compound (the "incomplete combustion products" or PIC). Which attributes various compounds exhibit are well known in the art and may be found in references such as the "Chemical Safety use micro Guide" (Pocket Guide to Chemical Hazards) published by the national institute for Occupational Safety and Health, NIOSH), the "Chemical Safety Data card" (MSDS) commonly published by Chemical manufacturers, and the "Chemical hazard reaction Information System" (CHRIS) published by the United States Coast Guard (United States Coast Guard), the entire contents of which are incorporated herein by reference. The database may also include information relating to health hazards posed by different compounds, and information on how to treat injuries exposed to the compounds or how to contain and handle the compounds most safely, as well as general and specific handling procedures following a chemical spill incident. The latter information may be found in resources such as "Emergency Response Guide 2000" (ERG 2000) published by the U.S. department of transportation and the U.S. federal fire Administration (USFA), and "Hazardous materials Guide for First-responders" (FEMA) published by the U.S. federal government Emergency management Administration (FEMA), the entire contents of which are incorporated herein by reference. As an example, fig. 6 shows a conceptual diagram of a chemical property database. As shown, the different compounds (denoted as compound A, B, C, etc.) will be related to their respective chemical attributes, such as whether the compound produces a visible vapor (at atmospheric pressure), the density of the vapor relative to air, the color of the vapor (if present), the odor of the compound (if present), the Molecular Weight (MW), and the color, odor and density (if available) of the hydrolyzed state of the compound (i.e., exposure to water). For illustrative purposes, fig. 6 is constructed assuming that compound "a" is ammonia and compound "B" is perchloric acid, and the available attributes of each compound are listed. In the simplest example, the chemical property database may take the form of a computer-read lookup table. However, those skilled in the art will recognize that many more sophisticated database structures also exist. For example, in one preferred embodiment, a computer will build a database table structure in a software base environment (such as Microsoft Visio) and then create a database in an SQL related database, such as Microsoft SQL Server, that allows for the display of unique relationships, references, queries, and information contained in all the guides.

FIG. 5 is a flow chart of a method for indicating the potential presence of a chemical compound or a collection of chemical compounds based on observable conditions at a chemical leak site. When the method is implemented on a computer, it typically takes the form of a "Wizard" or program that looks for information from the user in a series of queries and provides solution information or feedback information based on the information entered. The embodiment of the invention shown in FIG. 5 may be referred to herein as a "chemical property wizard".

The procedure starts at step 20 and will first be confirmed by the user what wizard sub-routine will be executed. In step 21, the program first asks whether water has been used in a chemical leak (e.g., by a fire hose). If the response is positive, hydrolysis conditions may be present and the steps taken are described in steps 30-37 below. Assuming that water is not currently being used in the chemical leak, the user is queried in step 22 whether steam is visible when the user observes the chemical leak. If so, step 23 asks whether there is an observable color in the vapor. In the embodiment of fig. 5, step 24 then provides the user with a selection of the color bar on the touch screen 11. As described above, the database will correlate certain compounds with certain colors. The program typically provides the user with 10-20 color fields with the name of the color printed under the field. It will be appreciated that more than one compound will likely be identified with the same colour. However, identification of color can be used to narrow down the possible number of compounds that may be present in a chemical leak. Once the color is selected, the program moves to step 24 and asks the user if steam is rising. If the steam is rising, then in step 25 it is assumed that the steam density is greater than 1mg/m³(density of ambient air). If the steam is falling, it is assumed that the steam density is less than 1mg/m³. Furthermore, chemical substances having the same density (1.0) as air tend to be uniformly dispersed in the surrounding air containing the substance, and when dischargedWhen exposed to open air, lighter-than-air chemicals will rise and leave the ground. The next step 28 asks whether there is an odor associated with the chemical leak. In a similar way as described with respect to the colours, the user is provided with a description of all chemical odours listed in the database on the touch screen 11 in step 29. For example, the USFA HazMat Guide, which describes the odor of different compounds, describes ammonia as having a "harsh odor", or perchloric acid as having a "bleach-like" odor. After the user indicates whether the scent of the chemical substance leaked matches a certain scent listed on the touch screen 11, the process proceeds to step 38. Step 38 asks the user whether water is now used in the chemical leak. If so, the hydrolysis reaction that occurs may provide a chemical leak with a different color and a different odor. The fact that a compound changes color, odor, or density upon hydrolysis may provide further evidence as to the identity of the compound. It can be seen that by answering "yes", step 32 can reach either the inquiry in step 38 or step 21. An affirmative answer in step 21 anticipates that the system user first observes a chemical leak of used water. The affirmative answer in step 38 anticipates the additional case that a leak is first observed when no water is present and then water is used. Step 31 asks whether a new color of steam has appeared as a result of the hydrolysis. If so, step 32 again displays the color selection to the user on the touch screen 11 and allows the user to select one color before moving to step 33. Subsequently (or when no color change occurs upon hydrolysis), step 33 queries whether steam is rising, and step 34 or 35 makes a decision as to the steam density based on whether steam is rising as discussed above. Steps 36 and 37 will make a new inquiry about the smell of the leaking compound in the hydrolyzed state. Step 39 will then ask whether there is a fire associated with the chemical leak and whether there is accompanying smoke (step 40). If there is accompanying smoke (the user must distinguish between smoke from the fire and moisture from the evaporating water), step 41 will determine that the molecular weight of the combustion material is greater than that of propane (i.e., 44). The conclusion is based onA prerequisite for (1) is that substances having a higher molecular weight than propane produce visible smoke on combustion.

Once all of the queries have been answered, the program returns a list of all chemical compounds for which the database has user-entered attributes. In a preferred embodiment of the invention, a compound is only indicated to be present if all of the input attributes match the compound. It should be understood that a program does not always return a suspect compound, but may often return several compounds that satisfy the observed property.

Another embodiment of the invention relates to a database associating locations with specific chemical compounds. The physical location in question will typically be an industrial plant or other facility that will store chemicals as part of its commercial operations. Information about what chemicals are present in what institution may be obtained from the Local Emergency Planning Committee (LEPC) and the State government Emergency response committee (SERC), or by other methods of entry into the database, such as fire brigade planning data. The U.S. emergency program and community awareness law (EPCRA) has regulated that organizations that store, produce, or use potentially hazardous chemicals must submit catalogs of emergency and hazardous chemicals to LEPCs, SERCs, and local first responders such as fire brigades each year. The organization provides this information in the table of "Tier I" or "Tier II". Many states require Tier II tables and this provides basic agency identification information, employee contact information regarding emergency and non-emergency events, and information regarding chemicals stored or used by the agency. This information is publicly available and may be organized in an electronic database format. Tier I & II information generally identifies the location of an organization by a block address, and the address of an organization may be referred to as a "given map location" of a chemical compound stored in the organization.

FIG. 7 illustrates a flow chart of a method of associating user-entered location information with a database having Tier I or Tier II data. This method is typically implemented in the form of computer software and is sometimes referred to herein as a "chassis wizard" program. Step 50a represents the initialization of the organization wizard. In step 50b, the program will receive input indicative of the physical location of interest. When the mechanism wizard is run in a hardware configuration such as that shown in figure 4, the location input will be automatically provided by the Global Positioning System (GPS) unit 7, so the "system location" is the location of the GPS device. The GPS position data may be standard American Engineers and Mechanics Association (NEMA) GPS format data including degrees and minutes of latitude and longitude with the measurement of the minutes being accurate to 6 digits after the decimal point. However, there are other well-known GPS message formats that accomplish the same purpose. Since Tier I and Tier II data typically provide the organization location via a block address, the organization wizard uses a conventional mapping program, such as Microsoft Map-point with address geocoding features, to correlate latitude/longitude data with the closest corresponding block address data.

The institution wizard next gives the name of the institution that relates Tier II data to the selected location and asks the user in step 51 whether the institution is the location where the user is confident that a chemical leak has occurred or is the location from which the leaking material has been dispensed. If the user's response is negative, step 56 will automatically return a list of all chemicals reported by the institution within a given radius of the input location. In one embodiment, the radius has a default value of 0.5 miles, but the radius may be increased or decreased by user input. Alternatively, if the query in step 51 is answered in the affirmative, then only the chemicals reported by the confirmed agencies are listed, as depicted in step 52. Next, step 53 will ask the user if the leak is from a known storage container. If the user can confirm the container emitting the leaking material, step 54 of the facility wizard will provide the user with a selection of container profiles on the touch screen 11. FIG. 8 shows examples 61a-64d of four container profiles that may be provided to a user. The fact that a leaking substance emanating from a particular container type does not identify the compound being leaked, but can define the type or class of compound that may be stored in that container. All information about the storage container and the type of product associated with the container can be publicly available from certain sources, such as the Hazmat guidelines of the united states federal fire administration (USFA).

Whether the institutional wizard branch is taken to step 56 or step 54, the next step is to ask the user in step 55 for the presence of an information statement, such as the National Fire Protection Association (NFPA) statement shown in the screen shot of fig. 9. As is well known in the art, the NFPA ticket 63 is divided into four diamond-shaped areas, shown as 64 a-64 d. Area 64a is blue (health risk), area 64b is red (flammable), area 64c is yellow (reactive), and area 64d is white (of particular interest). The regions 64 a-64d also include numbers ranging from 0 to 4 in the region, where the numbers indicate the degree of growth of each category, health hazard, flammability, or susceptibility to reaction. Region 64d will have a special letter code included therein to indicate particular concerns, such as hydrolysis. Step 57 prompts the user to select a particular NFPA ticket, which in the embodiment shown in fig. 9 would require the user to regenerate the ticket he or she is viewing on the touch screen 11. This is accomplished by touching the adjacent numbers or letters of the areas 64 a-64d to recreate the statement value observed at the site of the chemical leak. Because of the use of a container profile, the NFPA statement does not necessarily provide the exact compound, but there is typically a set of compounds that correspond to a statement with a specific set of numbers.

In step 58, the institutional wizard will initialize the chemical properties wizard discussed with reference to FIG. 5, and the chemical properties wizard will present the user with a set of queries and return the possible chemical compounds indicated based on their validation process. In step 59, the results of steps 50-57 of the organization wizard and the results of the chemical property wizard are compared and a chemical compound common to both is provided as the final result of the organization wizard.

Another embodiment of the invention relates to a method (typically computer-based) of identifying possible chemical compounds based on observed information about highway tank trailers or railroad container vehicles. FIG. 10 illustrates a flow chart known as the "Railcar Wizard" (Railcar Wizard) used when a chemical leak is emanating from a Railcar. It will be appreciated that although described in terms of a railway container truck, the process is equally applicable to highway tank trailers and the invention has clearly considered this. Step 70 initiates the present routine on the touch screen 11 (see fig. 4). In step 71, the user will be prompted to observe whether there is a U.S. department of transportation (DOT) statement on the railcar that is associated with a chemical spill. FIG. 11 illustrates several exemplary DOT statements that may be displayed on the touch screen 11, as set forth in step 72. In some cases, the statement will clearly indicate the chemical in the container truck (see, e.g., statements about oxygen and gasoline). However, in most cases, the DOT ticket will only provide certain general attributes (e.g., flammable, toxic, etc.) that can be used for several chemical compounds in the case of the NFPA ticket described above. The program will note all compounds associated with the selected ticket as possible results of the railroad car guide program.

Regardless of whether the available DOT ticket is validated, the next step the railcar wizard program will ask the user if there is a United Nations (UN) number. The UN code typically appears as four digits in a ring and may appear on a DOT ticket or on a separate ticket. The UN code is the unique code assigned to each chemical compound and is typically found in any published hazardous materials guide. In one embodiment of the invention, the user will use the touch screen to select the number sequence that makes up the UN code that is observed at the site of the leak.

Following the query for the UN code, the next railcar wizard program will query the user at step 75 whether there is a railcar profile in the database that matches the leaking railcar. The USFA "hazardous materials guidelines for first responders" includes the railway tanker profile typically used in the railway industry. The USFA "hazardous materials guidelines" indicate that certain types of tank trucks are designated for certain chemical compounds. For example, DOT 105J100W tank trucks are designated for ethylene oxide, liquefied petroleum gas and liquefied hydrocarbon gas, and DOT 105J200W tank trucks are designated for sulfur dioxide, vinyl chloride and liquefied petroleum gas. Fig. 12 shows a screen shot illustrating how the railcar wizard guides the user to select the correct railcar to observe at the accident site. Using up/down selection arrow 81, the user will look over the different profiles in the database and compare profile 80 to the observed railcar. The user will simply touch the appropriate railcar to indicate that this is what is observed on site.

After selecting the profile in step 76, the railroad car wizard will call out the chemical properties wizard discussed above in step 77, and the chemical properties wizard will present the user with a set of queries and return any possible chemical compounds indicated based on their validation process. In step 78, the results of steps 70-76 and the results of the chemical property wizard will be compared and the chemical compound common to both will be provided as the final result of the railroad car wizard. Although the UN code is the most accurate identification of a compound when it is correctly updated on a container, there is no guarantee that the UN code has in fact been updated. Thus, a compound identified by a UN code may only be identified by a program when it also corresponds to a chemical or a chemical identified by other identification (profile, DOT/NFPA statement, chemical property action wizard program).

2. Spectral identification of compounds

Another main aspect of the present invention is the use of spectroscopic analysis to remotely identify unknown compounds in chemical leaks. As discussed above, one key parameter in performing reliable spectral analysis is the accurate background spectrum. The present invention provides a new and advantageous method of generating an accurate background spectrum. However, as a preliminary process for determining the background spectrum, fig. 13 illustratesThe invention is how to perform several initialization steps when obtaining the initial spectral reading. An initial step 122 involves an "alignment mode" process, which is performed by existing software in the spectrometer. These steps are discussed briefly with reference to fig. 14. In this mode, the spectrometer will take one interferogram scan and convert the scan (by FFT with triangle apodization function) to a value of 650cm at v^-1～4200cm^-1Sample single beam Spectrum (SB) over a range of frequencies_i ^S). (step 141).

Next, the system estimates the total incident infrared power in the signal by measuring the difference between the minimum and maximum values of the interferogram (peak-to-peak value) (step 142). The magnitude of the IR signal power may be represented by any conventional digital or graphical user interface. If the signal power is greater than the spectrometer's recommended power range, the user will place an IR filter opposite the spectrometer's lens and take another reading of the signal power. If the signal power is less than the recommended range, the user will take additional readings at a different location, which will likely result in more IR energy being emitted into the spectrometer. If no viewing angle can provide sufficient signal energy, there may be no spectral reading under the existing conditions.

Once the system has confirmed that there is sufficient signal energy, the system performs several further steps to provide additional data. As shown in fig. 13, for each spectral reading, the system will assign a specific identification number to the reading (step 125), obtain the compass direction (step 126), obtain distance measurements for background objects (step 127), obtain temperature and humidity readings (step 128), obtain GPS readings (step 132), obtain camera focus position (step 131), store metadata from the video capture (step 130), start video capture (which may be stored in a conventional JPEG format) (step 129), start video time stamping (step 133), perform a predetermined number of scans in preselected wave intervals to construct an interferogram (step 134), end time stamping (step 135), stop video capture (step 136), and then process the interferogram (step 137). In a preferred embodiment, 32 scans are performed with respect to the interferogram at wavenumber intervals of 0.5 in order to create a composite interferogram. The manipulation of interferograms and the processing of interferograms is well known in the art and need not be described further herein.

One step shown in fig. 14 (step 143) is to perform an initial estimate of the background temperature shown in fig. 15. Before studying the steps in fig. 15, it is necessary to understand that in order to perform this estimation, the system will use a database of known sample single beam spectra for different gas mixtures with known gas and background temperatures.

In a preferred embodiment, 24 single-beam background spectra are used in estimating the sample background spectrum. These 24 spectra represent the background spectrum between 0.0 and 142 ℃ and are illustrated in fig. 19; these spectra were recorded using a Model340 blackbody radiation source, manufactured by Mikron infra, inc.

Another experimental run of trials for generating absorption spectra included a heated unidirectional gas sample cell (with an absorption path length of 10 cm) and a solid, grey "hot plate" of 25cm diameter. A type K thermocouple is attached to the surface of the hot plate, and an absorption tank supplies a sample gas (T)_S) Temperature of (2) and background temperature (T)_B) The measurement result of (1). By filling the field of view of the spectrometer with a hot plate and placing a gas cell between the spectrometer and the background, the spectra that formed the collection were recorded. These absorption spectra are used to generate the temperature contrast Δ T ═ T discussed in equations 18-20 below_A-T⁰ _B。

After the initial absorption spectrum is generated, the spectrum passes through a plurality (-200) of small frequency ranges (corresponding primarily to H)₂O and CO₂) The gas phase absorption elements are removed and replaced with a cubic spline interpolation approximation, modifying the spectra to a relatively smooth underlying background emission spectrum. The spectra may also be scaled to set their maxima uniform. FIGS. 19 and 20 illustrate the wavenumber region of 650 to 1400cm^-1Last background referenceAnd (4) collecting. Although fig. 19 illustrates a complete set of background spectra, for the sake of clarity, fig. 20 only shows spectra representing minimum and maximum temperatures (0 ℃ and 142 ℃) of a set of reference background spectra.

As shown in FIG. 20, the parabolic curves shown fit each of these spectra (0℃. and 142℃.), and the maximum height or center of the parabolas is defined by Vc₁And Vc₂And (4) showing. When several single-beam spectra are generated over a given temperature range (e.g., 0 deg.C-142 deg.C) (see FIG. 19), the center of the parabola for these single-beam spectra can be estimated and a relationship representing the background temperature versus the center of the parabola can be plotted, as shown in FIG. 21. Using this relationship, it can be understood how the estimated parabolic center of the sample spectrum is given, and by reading the point corresponding to the parabolic center on the temperature axis, the estimated temperature can be obtained.

Returning now to FIG. 15, to implement the process as shown in this figure, the system first acquires an analysis range from the spectrometer (step 146). The database comprises v_AAnd v_BWhich in a common embodiment may be defined as v_A＝650cm^-1And v_B＝1100cm^-1. All spectral information for the described embodiment is present in a standard disk file, which has a format developed by Thermo practical, inc. The disk file includes several blocks with header information followed by an ordered list of spectral y-values representing absorbance or infrared intensity depending on the type of spectral data. In addition to the data, the header information also contains the frequency of the initial y data points (FFP), the frequency of the last y data points (FLP), and the number of points in the file (NPTS). The value of the x-axis associated with the y-data is not stored in the file, but can be calculated from the data location in the list, FFP, FLP and NPTS, if desired. For example, the frequency associated with the 13 th y-value in the file is equal toThe embodiment uses all spectral files with the same values of FFP, FLP and NPTS, so the SPC file format lends itself to easily constructing columns in the design matrix D. The position of the corresponding y value in the disk file list can be calculated by the upper frequency of the desired analysis range and the header information. This value is read from the disk file and placed in the first row of the design matrix column. Subsequent values of y (up to data corresponding to the lower frequency of the analysis range) can be easily read sequentially from the disk file and placed sequentially at the lower row position of the same column of D.

The system will then define the elements needed to implement the matrix represented by equation 2. For example, by SB in the scope of the analysis defined above^S _iVector a is defined (step 147 of fig. 15). The vector element is the analysis range A_i＝(SB)^S _iWherein v is_i＝v_A～v_BSample single beam value of (1). The system will then construct the design matrix D as described in step 148 of fig. 15. Column of D includes a constant, value v_iSum value (v)_i)²To provide a matrix:

equation (12)

The design matrix representation with equation 12 is over a frequency range v₁～v_NSample single beam spectrum on; in the present application, v₁＝650(cm^-1) And v is_N＝1100(cm^-1). Equation 9 yields the parameter X when the elements of the row vector a are set equal to the elements of the sample single-beam spectrum over the frequency range and the initial (k-0) P matrix is set to the identity matrix I_L ^k(L ═ 0, 1, 2,. N), wherein the estimated parabola is given by:

equation (13)

And k represents the number of iterations of the "weight iteration" CLS ("WICLS").

WICLS (step 149 of fig. 15) is a new CLS analysis method performed by the present invention in order to obtain a more accurate estimation of the parabola (step 150) than the CLS method of the prior art. The WICLS technique uses iterative adjustments in the weight matrix P (equation 5) based on the residual V (equation 7). In many cases, the weighted sum of residual squares, V, is performed in successive iterations²The change in (equation 8) will eventually be below a predetermined positive value, informing the end of the iterative process. This technique results in effective suppression of certain CLS analyses in which a subset of the measured data A (equation 2) cannot be represented by the design momentsThe linear model expressed in array D (Eq. 3) is well represented. WICLS estimate X, as used by the present invention_jAnd its associated MSD value Δ_j(equations 9 and 11) outperformed the corresponding values provided by standard CLS analysis.

Figure 16 illustrates the steps performed to implement the WICLS technique. The matrices a and D are sampled in step 156. Prior to the first iteration of the WICLS analysis, the weight matrix P is set equal to the identity matrix I (step 157), i.e. all data involved in the analysis are given the same weight. Next in step 158, an initial (k ═ 0) parameter estimate X is calculated as described above_jResidual V, variance and V²And marginal standard deviation Δ_j(see equations 8-11). In step 159, the weight P is calculated and the iteration count k is incremented. Redefining the weights according to the following formula based on the residual error of the previous iteration:

equation (14)

And the average thereof is:

equation (15)

Where the superscript k > 0 indicates the number of iterations. Weight P for the next (k +1) th iteration_ij ^k+1Determined by a non-linear relationship

Equation (16)

And isFor i ≠ j

Where α ≧ 0 is an experimentally determined adjustable parameter for each WICLS application. In a preferred embodiment of the invention, the value α is 1. Using the final weight matrix P, in step 160, the parameter estimates X are recalculated using equations 8-11 again_jResidual V, variance and V²And marginal standard deviation Δ_j. In many cases, the values (X) are iteratively repeated using equations 12-16_j)^k、( Δ_j)^kAnd (V)²)^kConvergence, where the superscript k > 0 still indicates the number of iterations. In this case, (V) for the next k value²)^kThe value of (b) satisfies:

equation (17)

When the relationship expressed in equation 17 is satisfied (step 161), the value (X)_j)^k+1And (Δ)_j)^k+1(and all other related estimates) are employed as values for the WICLS analysis. In a preferred embodiment, the value L is 0.01.

Application of the WICLS process described by equations 12-17 will provide information about X through the parabola described by equation 13₀ ^k、X₁ ^k、X₂ ^kThe value of (c). The finally estimated center frequency v of the parabola of interest_C(step 151) may be calculated as follows:

equation (18)

As discussed above, the WICLS estimate v shown in FIG. 21_CA curve against a known background temperature will provide a basis for estimating the background temperature. Polynomial regression on the experimental data represented in fig. 21 resulted in an estimate of the function having the form:

equation (19)

When the process described in equations 16-19 is applied to a sample single beam spectrum of unknown background temperature, they produce an estimate of the center frequency of the spectrum, v_C ^S. Initial estimated background temperature T of the spectrum_B ⁰Given by:

equation (20)

The system displays a digital or graphical representation of the background temperature. This allows the system to quickly provide an updated IR energy pattern output and temperature contrast Δ T ═ T_A-T⁰ _B(see equations 18-20 and FIG. 28).

Once the system determines an initial estimate of the background temperature (step 176 in fig. 18), it will continue to construct a background spectrum. Background temperature T_B ⁰Is used to select a "reference background Single Beam (SB) pair", which are two background spectra, one of which is indicated at T_B ⁰The top closest spectrum, and the other at T_B ⁰The next closest spectrum (step 178). For illustrative purposes, the bottom most spectrum (0 ℃) in FIG. 19 is used, and the adjacent spectrum (6.8 ℃) is used. If T is_B ⁰Between 0 ℃ and 6.8 ℃, then the two spectra will be a reference background SB pair.

By fitting the sample SB spectrum with the reference background SB pair, an estimated background SB spectrum will be generated by means of the WICLS method described above (step 179). Will be at a frequency v_iThe intensity value of the reference background SB pair at (A) is denoted as S_i ¹And S_i ²Design matrix at v₁～v_NIs formulated as:

equation (21)

At the same time at v₁～v_NThe corresponding value of the sample single beam spectrum over the frequency range of (a) is used as the input row vector a (see equation 2). As noted above, the value used by the current application is v₁＝650cm^-1And v_N＝1100cm^-1. Application of the WICLS process described by equations 12-17 results in two estimates X₁And X₂(ii) a Estimated background SB Spectrum SB_i ^BGiven as follows:

equation (22)

FIG. 17, step 171, proposes that the background temperature (T) is ultimately performed by the system_B) To the final estimate of (c). Determination of initial background temperature T Using the above description_B ⁰Except that the row vector A is set equal to the estimated background SB spectrum SB (equation 12-20)_i ^B(equation 22) instead of the sample SB spectrum as shown in fig. 19. The background temperature T is given by equation 23_BThe final estimate of (step 180, fig. 18), which is similar to equation 20:

equation (23)

Once the cover is closedThe estimated background SB spectrum SB is determined_i ^BThen the sample SB absorption spectrum SB can be calculated using equation 1_i ^S(step 181, fig. 18), as is known in the art.

Once the sample absorption spectrum is obtained, the present invention will begin a discrimination process that identifies which spectra of individual compounds appear in the absorption spectrum. The system first uses a variety of methods to initially identify a set of compounds represented in a sample spectrum. The system then uses a second method with increased certainty to determine whether the compound identified in the first search is indeed represented in the reference spectrum.

Initial identification of compounds:

in performing the initial identification of a compound, which is represented in a sample spectrum, one method used by the present invention is to compare the sample spectrum to the known absorption spectra of different compounds. The absorption spectra of many compounds are known and can be obtained from sources such as Thermo Galactic of Salem, NH, mentioned above. As described above, the spectral information provided by Thermo galenic is compiled in a standard disk file in a format developed by Thermo galenic, which is referred to as "SPC" format. The construction of the design matrix D is the same as described above.

By means of a database correlating compounds with a known spectrum or spectra (hereinafter referred to as reference spectra), the system of the invention can CLS compare the reference spectrum with the sample absorption spectrum and determine the likelihood that the reference compound is represented in the sample absorption spectrum. FIG. 22 illustrates one method of the present invention to make this comparison. In step 196, the system will define a search compound, which may be the first compound in the database. Once the search compound is confirmed, the scope of the analysis associated with that compound will then be confirmed. This analytical range is defined based on the IR frequency at which the compound is demonstrated to absorb. For example, FIG. 23a shows the analysis of the extent of analysis of hypothetical Compound B, its occupancyThe range includes two non-adjacent sets of infrared frequencies. The analysis range of the compound B comprises a frequency part B1 (810-850 cm)^-1) And B2 (880-900 cm)^-1). Since there is only one compound considered, the range of analysis included in the design matrices (R1 and R2) will include these two non-adjacent frequency groups. However, if H₂O and CO₂Overlap with either R1 or R2, the design matrix will also include information about H₂O and CO₂The column (c). For example, if H₂The analysis range of O is 1300-850 cm^-1Then one column in D will include information about H₂The frequency range of O R1 and the reference spectral value of R2. Likewise, if CO₂The analysis range of (a) is 650-750 cm^-1Then the design matrix for compound B described above will not contain a design matrix corresponding to CO₂The column (c). Note that H in the assay₂O or CO₂Does not affect the scope of the analysis used in the analysis (i.e., the rows of the design matrix), but only the number of columns of the design matrix.

Certain analytical ranges are associated with each compound comprised in the database that occupies at least one reference spectrum, and these analytical ranges are also stored in the database. These analysis ranges will involve a large fraction limited by the subset of adjacent IR frequencies at which absorption of the compound is demonstrated, while also demonstrating the favorable infrared intensity of the background spectrum at that IR frequency. Furthermore, the selection of the appropriate subset of frequencies relates to the set of frequencies that other compounds included in the database are proven to absorb; among these other compounds, H in the ambient air₂O and CO₂Is ubiquitous and therefore of paramount importance. However, certain other types of compounds may also be of particular interest, these being compounds which may be present in the sample of interest, or compounds having a particularly strong absorption characteristic, or compounds having a particularly broad absorption range. The selection of an appropriate analytical range will also relate to the intensity and uniqueness of the spectral absorption range of each compound. Furthermore, for those compounds that are associated with two or more assay domains, the assay domains will be assigned"grades" which distinguish these analytical ranges from other ranges associated with the compound, and these grades will also be stored in the database. The spectral search and CLS analysis are adjusted to include all analysis ranges with levels below the adjustable maximum level. This technique allows additional spectral data about the compound to be used in spectral search and CLS analysis, if desired; in laboratory testing and field applications, the adjustment is performed in response to the quality of the spectral search results and/or CLS analysis results.

Figure 24 illustrates possible selection and fractionation processes for the compound ethanol. As shown by the following traces, which represent the absorption range of ethanol, there are three distinct absorption bands, labeled as region 1 to region 3, at 800 to 1300cm^-1In the meantime. Between these limits, the background intensity, although varying, is relatively constant. In contrast, a strong H can be seen₂The O absorption features become more intense in the higher wavenumber portion of the spectrum. The combination of these factors illustrates that in the presence of water, zone 2 will produce the best analytical results with respect to ethanol, while zone 1 will produce the worst results. This results in region 2 being designated as "level 1", region 3 being designated as "level 2", and region 1 being designated as "level 3". In a preferred embodiment, the subroutine is arranged to use only regions having level 1 in the design matrix. However, the analysis subroutine may also be adjusted to include regions having levels 1 and 2; alternatively, the analysis subroutine may even be adjusted to include all levels (1, 2 and 3) for the compound. Moreover, one skilled in the art will recognize that each compound will likely have any number of ranges, and that these ranges may be assigned any rank. Thus, the present invention is not limited to the three-level example described above and to an example in which there is one area per level.

Since the system again uses CLS comparison, the design matrix D for the CLS process must be the design matrix as defined in step 197 of FIG. 22. At least one column in the design matrix represents a reference spectrum of the selected compound (and if multiple reference spectra are selected, byMultiple column representations). The rows of the design matrix include only infrared frequencies that are in the analytical range of the search compound. Since the CLS analysis is not the WICLS analysis, the weight matrix is set equal to the identity matrix (P ═ I) in step 198. Calculating the parameter estimates X, residual V, variance and V for the search compound by equations 8-11 as described above²And a marginal standard deviation Δ (step 199). These values are sought for the purpose that they can provide a "quality index" Q_jIt means CLS determination parameter X_jAn indicator of quality of (a). Using the sample absorption vectors a, Q from step 200_jCan be defined as:

equation (24)

U and Θ are defined below and need only be understood at present for use in the present invention, with the associated quality index Q falling below a specific value_jCLS value of (1)_jIs referred to as "pending". At Q_jThe following compounds are considered to be the undetermined ones of this Q_jThe values vary from CLS application to CLS application. Preferred embodiments of the present invention recognize Q less than 90_jIs undefined X_jThe value is obtained. The determination is made in FIG. 22Expressed in step 201, where Q is greater than 90_jIs considered a positive ("determined") result (step 202), with a lower Q_jIs considered a negative ("undetermined") result (step 203). Finally, the results of this search are reported in step 204. Thus, the method described above, referred to herein as "CLS search," provides an example of one process for searching for and identifying compounds that may be represented by the sample absorption spectrum.

Another method used in the present invention in performing the initial identification of compounds was developed by the Thermo Galactic of Salem, NH. Although this method is used in the present invention ("Galactic search"), Galactic search itself is not subject to the present invention. The Galactic search method (see FIG. 25) first determines (step 226) the analytical range of the chemical to be compared to the sample spectrum (step 227). As in the method of fig. 22, the Galactic search uses a database that relates the absorption characteristics of different chemical compounds over a given range. The system will correlate the absorption characteristics to the frequency range in which the compound absorbs IR energy. This "analytical range" will correspond to the compound being searched. Once the analysis range is determined, the Galactic search will use a number of different algorithms in making the search validation. An important difference between this search and the search described above by way of FIG. 22 is that the Galactic search does not include any representation of CO₂Or H₂And O, the data of the O. As described in step 29 of fig. 25, the search type of the Galactic search includes a first derivative based on the correlation, the first derivative, the euclidean distance, and the euclidean distance.

The Galactic search is initially set to begin searching by one of the specific search types listed above (step 228). The search will generate a list of all reference spectra in the database, arranged in order of the degree of correlation of the same spectrum (step 230). If the reference spectrum of the selected compound is the first in the rank queue, then the percent separation between the first and second ranks is N (step 231), then the selected compound is deemed to have reliable results and the reliability index associated with that compound is incremented by 1 (step 232). The degree of separation N may vary for different embodiments. By way of example, the degree of separation N may be 0 percentage points in one embodiment, 25 percentage points in another, or any other percentage found to give an accurate result. The method is then repeated for each of the five search types described above, and the reliability index is increased (to a maximum possible value of 4) when the condition of step 5 is met. The method is then repeated for each compound in the Galactic database (step 233). The final output of the Galactic search will be a list of compounds and compound reliability index values (step 234), the compounds in the list having a reliability index of at least 1.

By using CLS search, Galactic search and wizard program search, the present invention will create an initial list or tentative list of chemical compounds represented by the sample absorption spectrum. "wizard search" refers to a chemical property wizard, institutional wizard, or railroad car wizard as described above that indicates a possible compound to be present. The present invention will then process this initial list of compounds in a new way in order to determine with greater certainty which compounds in the initial list are most likely represented by the sample absorption spectra. FIG. 26 illustrates how the first step (241) in this process is to create the primary and secondary lists based on the consistency of the chemicals that intersect between the three search lists (242 a-242C), hereinafter the CLS search list is represented by the notation { C }, the Galactic search list by the notation { G } and the wizard search list by the notation { W }.

The primary list { P } includes all compounds identified in list { C } and those found in both lists { W } and { G }. Expressed in aggregate notation, { P } is equal to { C } { { W } { { G } }. The set { S } is the complement of { P } with respect to the entire union { C }. U }. U.E }. G }. In other words, { S } denotes any compound found in the lists { C }, { W }, or { G } and not listed in { P }. The list { P } represents those compounds most likely represented by the sample absorption spectrum, while the list of compounds in { S } is considered unlikely to beNow in the sample, but some spectral analysis attention is also needed. Next, in step 243, the system will define the analysis range R for CLS from { P }, the design matrix D, and the sample input vector A. The analysis range R is the union of the analysis ranges for all compounds in P and is determined in a similar manner as described above with reference to fig. 23a, as shown in fig. 23 b. As an example, if the list { P } had only two compounds a and B (including their respective frequency ranges a1, a2 and B1, B2), the analysis range R would be those IR frequencies for which any compound is declared to absorb. Analogously to the above, if Compound H₂O and CO₂To the extent that they overlap with the IR frequencies defined by P, they are also included in the analysis, but their frequency ranges cannot otherwise be used to define R. The elements of vector a are the sample absorbance values over the analysis range R. Column D includes reference spectrum A_ij ^RAnd wave number v over the same frequency range_iAnd v_i ². The design matrix D can thus be represented as

Equation (25)

It will be appreciated that each column in the matrix D (except the two rightmost columns) is represented in the analysis range v₁～v_NReference spectrum A of the compounds in the above list { P }^R。

The next step (step 244) in fig. 26 is to perform an "adjust CLS" analysis on the list of compounds { P }. FIG. 27 illustrates the steps employed in performing an adjusted CLS analysis. The sample vector a and the design matrix D are as defined directly above (step 276). In step 277, the weight matrix P (see equations 5 and 6) is set to the identity matrix I (so that the form of this CLS analysis is a "weightless" form). Step 278 requires next calculating a parameter estimate X_jResidual V, variance and V²And marginal standard deviation Δ_j(see equations 7-11).

The system will also calculate the optical depth Θ using the following relationship_j ^SAnd uncertainty U_j ^S：

Equation (26)

Wherein:

Θ_j ^Rlight depth (ppm-m value) of a single reference spectrum of (pure) jth compound;

T_j ^Rrecord A_ij ^RAbsolute temperature of time;

P_j ^Rrecord A_ij ^RAbsolute pressure in time;

T^Srecord A_i ^SAbsolute temperature of time;

P^Srecord A_i ^SAbsolute pressure of time. Similarly, Θ_j ^SMSD (1. sigma. uncertainty) in (1. sigma.) with U_j ^SWhich is given by the formula:

equation (27)

Step 279 then calculates the light depth Θ of the sample spectrum_j ^SGiven light depth theta corresponding to the reference spectrum_j ^RA comparison is made. The depth of light is an important parameter because its value, in combination with the optical path length through the sample gas, will provide an estimate of the concentration of the compound. For many compounds, the infrared spectral database includes multiple reference spectra recorded at different depths of light (i.e., different concentrations). For this compound, the most accurate available CLS results are based on the specific reference spectrum of the compound having a depth of light closest to that found in the sample gasAnd the particular reference spectrum need not be the same reference spectrum selected for the initial design matrix D calculation.

To illustrate the source of this potential error, step 280 performs the following analysis. It will be appreciated that for each compound, the current CLS result Θ_j ^SNecessarily fall into one of three intervals: a) theta below the depth of light of the reference spectrum of the compound having the lowest depth of light_j ^SOr b) Theta between the depths of its two reference spectra_j ^SOr c) a depth of light Θ higher than the reference spectrum of the compound having the highest depth of light_j ^S. These three cases are described below. If the result is case (a), the current CLS result for the compound in question is consistent and no further action is required. If the result is case (b), then for a particular compound (j), it may be necessary to repeat the CLS analysis after replacing the corresponding column of the design matrix with a value from a different reference spectrum for a different light depth for that compound. WhereinThe following parameters may be defined:

andequation (28)

In general, the results of CLS may be defined to be consistent according to any suitable mathematical comparison of the β and γ values; this definition may be relevant for a particular compound and may require additional iteration constraints. For example, some allowable ranges for β and γ values may be predetermined (e.g., 0.5 ≦ β ≦ 5 and 0.5 ≦ γ ≦ 5), and this may be used to determine whether the CLS result is consistent.

In an alternative embodiment, if the current reference spectrum is not maintained at the optical depth Θ_jHIGH ^RThe result is defined as not conforming and only one change of the reference spectrum is allowed for any compound, regardless of the number of iterations required in the overall process. The process of fig. 27 uses this embodiment, which can be better understood by way of example. It is assumed that three reference spectra S1, S2 and S3 (respectively) with a light depth D1-100 ppm-m, D2-250 ppm-m and D3-500 ppm-m can be used for the analysis of a specific compound. Then assume that the current CLS analysis uses spectrum S3 and returns the resultThe result falls between D1 and D2, and thusWhileUsing the criteria given above, the results fall between D1 and D2And thus this is a non-conforming result, one of the reference spectra associated with these optical depths (S1 or S2) will likely provide more reliable results in current CLS analysis using reference spectra (S3 with D3-500 ppm-m). Step 4 in fig. 27 performs this function and also checks whether the previous result has been adjusted. If not, the design matrix is modified to include the values from S2 (instead of S3) and the analysis is repeated (step 281). However, if the design matrix has been adjusted in the previous CLS analysis as a result of the disagreement of the compound, further modifications are made, even though the disagreement results are produced by the current CLS analysis. This technique is needed to avoid allowing a computer program to enter an infinite logic loop.

If the result is case (c), then when the strong CLS result is a match, but the result is flagged as a potentially underestimated value of the actual sample light depth. Finally, step 282 provides for determining an accurate quality indicator Q_j(equation 26) to perform further calculations.

Returning to FIG. 26, upon completion of the adjusted CLS analysis in step 3, the most interesting result is the ppm-m estimate Θ_j ^SAnd its associated (new) quality index Q_j。Q_jAll compounds below certain selected values (typically close to 90 in preferred embodiments) are considered "indeterminate". If there are undetermined results, step 248 will remove the chemicals with undetermined results from the list { P } and add these compounds at the top of the list { S }. Using the new shortened list { P }, step 245 redefines the CLS analysis range R, the design matrix D, and the sample input vector A in a manner similar to step 243. The iterative process of steps 244 and 247 then continues until the results for all compounds in P are determined.

In the absence of further undetermined results, step 246 will put the initial compound { S } at the top of the list { S }₀Add to the last list { P } and remove { S } from { S }₀. Using the new list { P }, the CLS analysis range R, the design matrix D, and the samples are redefined in step 249The vector a is now input and another adjusted CLS analysis is performed on the new list P in step 250. If the most recently added { S }₀If the compound is specified in step 252, then step 251 will take the next { S }₀Add to { P } (and remove { S } from { S })₀) And step 249 is repeated. If { S } in step 252₀If the compound is undetermined, step 254 will determine if S is empty. If not, step 253 will remove { S } from the lists { P } and { S }₀And the next top compound of the list { S }₀Added to the list P. This process continues until S is found to be empty in step 254.

Before performing the final confirmation, i.e. confirming which compounds identified by the system are represented by the sample spectrum, it will be confirmed whether there are compounds that are IR radiation sources. Step 269 asks whether there is an estimate X_jIs less than 0. If all CLS results (and the associated ppm-m estimates Θ)_j ^S) Is zero or positive, the final step of the analysis will be to calculate an estimate of concentration C as described below_jAnd Γ_j. If there is a negative evaluation value X_jThen steps 255-265 are performed to determine if these suspected radiation sources are false results or if there is substantial evidence in the sample that the compound emits IR radiation. Step 255 will define a subset { R } of { P }, with X > 0, and will define a subset { N } of { P }, with X < 0. The set R includes compounds in the sample that are indicated as infrared light absorbers by final CLS analysis of P and N includes those compounds that are indicated as infrared light emitters. Step 256 will define a new design matrix D from R without redefining the extents (i.e., using the extents found in the final iteration of step 249). The columns of matrix D thus correspond only to those compounds that are indicated in the sample as being infrared light absorbers, but the spectral range corresponds to all compounds found in steps 241-254 with good established CLS results. Step 257 will perform an adjustment CLS for { R } in the same manner as above. If there are undetermined results, step 258 removes the undetermined compounds and repeats steps 256-259 until there are undetermined resultsUntil no undetermined results exist. Step 260 will then search the set N for the residual V on the chemical. If the compound that indicates that it is the infrared radiator (see step 269) is indeed present in the sample as well, the residual matrix V should contain evidence about them, thus subjecting the residual to the spectral search operation, i.e. the previous step 1 (i.e. the Galactic and CLS searches).

Step 261 determines whether a compound in the set N is found. If so, the residual spectrum includes a recognizable template with at least one compound that is included in the database and excluded by the CLS analysis used to generate the residual. The appearance of this recognizable template indicates that at least one absorbing or radiating compound not included in the analysis may be present in the sample. In this case, the results of steps 241-254 using the set { P } are the most available, and step 266 returns to the results of the set { P } confirmed near step 254. If the search in step 261 does not indicate the presence of any compound from the set { N }, then all negative results in steps 241-254 are considered spurious. The current set of compounds { R } is considered accurate except for the need to reduce the current range of analysis, so the method proceeds to steps 262-265. In step 262, the design matrix D is redefined using the compounds in { R } and the ranges for the compounds in { R }. Then, the CLS is adjusted using the redefined range. If there is an undetermined result in step 265, the compound with an undetermined result is removed in step 263 and the process returns to step 262. If there are no pending results, the current result is considered to be the final result and passed to step 267.

As a final step in the process of identifying chemical compounds, the present invention will determine the concentration C of each compound_jAnd an uncertainty Γ associated with the concentration measurement_j. These values can be calculated by the following formula:

equation (29)

And

equation (30)

Quantity F in equations 29 and 30_TIs a correction factor describing the observed contrast with temperature Δ T ═ T_A-T⁰ _BChange in the ethylene absorbance of (c). FIG. 28 illustrates the definition of F in this work_TAnd cubic polynomial regression. It can be seen that the sample absorption path length L_SMust be known. L is generally estimated by the user during the initial measurement process described in step 127 of FIG. 13_SAnd inputs it into the system (step 268).

FIG. 29 illustrates some additional functions performed by the software of the present invention. The subroutine section involves taking spectrometer readings of the gas while the gas is held in a glass sample vial or "cell". Taking a gas sample reading in a cell is well known in the art. However, several steps, including steps 300-305 and steps 314-316, may be performed independently, regardless of whether a gas cell is being utilized. Step 300 will initialize the subroutine and validate the subroutine for the user. Step 301 will ask the user if he or she knows the chemical compound that the spectrometer will identify. This allows for the situation that a user arriving at a chemical spill site has knowledge of potentially leaking compounds, particularly toxic or otherwise dangerous compounds. The user may wish to take readings of the spectrometer in several different directions at once to determineWhether the present compound can be identified. To accomplish this, step 302 will query the user to select a known or suspected compound from an alphabetical list (step 303) on the touch screen 11 (FIG. 4) listing all chemical compounds in the database. Once a known compound is selected, step 302 will return a reference spectrum for that compound to step 305. Confirmation of the presence of the particular compound can be performed very quickly if the discrimination software only needs to compare the spectral reading with a particular reference spectrum. The user will take a spectral reading in step 314 (described in detail in fig. 13), generate a background SB and absorption spectrum in step 315 (see fig. 18) and then perform the CLS search of fig. 22 in step 316. After the CLS search is performed, the result is sent to the subroutine shown in fig. 26. Only one reference spectrum appears in the lists 242 a-242 c, rather than the lists 242 a-242 d providing multiple reference spectra. Referring to fig. 26, the reference spectrum (along with those ubiquitous H) is processed as described above₂O and CO₂Spectrum of (d). The output of this determination will inform the user whether the spectrometer detected a known compound.

Returning to step 305, if the user determines that it is desired to use the gas cell (i.e., capture a gas sample into the cell and then perform a spectroscopic analysis of the gas in the cell), the program will proceed to step 306 and query whether the gas cell has been purged (i.e., flushed with an inert gas such as nitrogen to remove the residue from the previous use). If not, the user is notified to purge the gas cell, or if the spectrometer is properly equipped with additional hardware for automatically performing this function, the spectrometer does so. If the gas cell has been cleaned, the software will ask if the current background spectrum is available. It will be appreciated that a known predetermined background radiation source will be used in taking spectral readings of the sample in the gas cell. This is possible because the gas sample is contained in a gas cell that can be easily positioned at any distance between the spectrometer and the user-selected radiation source. It is clear that it is not the same subroutine as used to estimate the context based remote source as discussed above with respect to fig. 18. If the current context is available (i.e., the conditions under which the context is obtained do not change), the process moves to step 310. If the current background is not available, a new background SB spectrum will be used. This typically requires filling the cell with an inert gas such as nitrogen and recording the spectrum obtained when the cell is placed in front of the radiation source. Step 310 will ask if the gas sample is in the cell. If not, the program will notify the user to introduce a gas sample, or if the spectrometer is equipped with appropriate external hardware, the program will automatically open a valve on the gas cell and draw a sample of ambient air into the cell. Steps 312 and 313 will then take spectral readings of the gas in the cell and generate a single beam spectrum (see fig. 13) and an absorption spectrum (see equation 1) of the gas in the cell. The program then proceeds to the sub-routine of fig. 22 and 26 described above in step 316. It should be understood that steps 306-313 are generally well known in the art and do not form part of the present invention in and of themselves.

While the foregoing disclosure describes the invention in terms of specific embodiments, those skilled in the art will recognize that there are numerous modifications that are within the scope of the invention. For example, although the "wizard" subroutine described is used in conjunction with the spectral identification of chemical species, it is contemplated that the wizard itself (i.e., without involving the spectral analysis component) may be used to provide useful preliminary identification of unknown chemical compounds. All such modifications and variations are intended to be within the scope of the appended claims.

Claims (35)

1. A method of determining the temperature of a background object for which a sample spectrum is recorded, the method comprising the steps of:

a. providing a predetermined relationship between a parabolic center frequency and background temperature, wherein the parabolic center frequency is the frequency of a single beam spectrum of a reference background at a known temperature;

b. providing a sample spectrum recorded against a background of unknown temperature;

c. determining a parabolic curve most fitting to the sample spectrum;

d. determining a sample parabolic center frequency of the best fit parabolic curve;

e. comparing the sample parabolic center frequency to a predetermined relationship of the center frequency to a background temperature;

f. estimating a temperature of the background based on the comparison result.

2. The method of claim 1, wherein said step of determining a best fit parabolic curve further comprises the step of using a conventional least squares analysis.

3. The method of claim 2, wherein said step of using a conventional least squares analysis further comprises the step of using a weighted iterative conventional least squares analysis.

4. The method of claim 3, wherein the weighted iteration continues until the minor change in the sum of variances is less than a predetermined minor amount.

Background spectrum

5. A method of generating a background spectrum for spectroscopic analysis, the method comprising the steps of:

a. providing a sample spectrum and an estimated temperature of a background object;

b. selecting at least two known temperature spectra from a set of known temperature spectra for a known background temperature, which represent background temperatures above and below the estimated temperature;

c. comparing the sample spectrum to the known temperature spectrum to facilitate determination of a sample background spectrum.

6. The method of claim 5, wherein said step of comparing said sample spectrum to said known temperature spectrum further comprises the step of using a conventional least squares analysis.

Chemical identification

7. A method of generating a temperature compensated absorption spectrum, the method comprising the steps of:

a. providing a sample spectrum and an estimated temperature of a background object;

b. selecting at least two known temperature spectra, representing background temperatures above and below the estimated temperature, from a set of known temperature spectra for known background temperatures;

c. comparing the sample spectrum to the known temperature spectrum to facilitate determination of a sample background spectrum;

d. an absorption spectrum is calculated from the sample spectrum and the background spectrum.

8. The method of claim 7, further comprising the step of comparing said absorption spectrum to a reference spectrum of at least one chemical species to facilitate identification of the chemical species represented in said absorption spectrum.

9. The method of claim 8, wherein the step of comparing further comprises using CO₂And H₂O, to facilitate identification of the chemical species.

10. A method of identifying a chemical species represented in an absorption spectrum, the method comprising the steps of:

a. providing a sample absorption spectrum;

b. the absorption spectrum is combined with CO₂And H₂The spectrum of O is compared and compared to at least one chemical reference spectrum to identify the chemical represented in the absorption spectrum.

11. The method of claim 10, wherein the CO is₂And H₂The analytical spectral range of O and the chemical reference spectrum are compared to the absorption spectrum.

12. The method of claim 11, wherein a plurality of chemical reference spectra are compared to the absorption spectrum.

13. The method of claim 12, wherein a conventional least squares analysis is performed to separate the CO₂、H₂The reference spectra of O and the chemical are compared to the absorption spectrum.

14. The method of claim 12, wherein the plurality of chemical reference spectra comprise a primary set of chemical reference spectra, and the primary set of chemical reference spectra are formed by a plurality of chemical spectral discrimination methods.

15. The method of claim 12, wherein the reference spectra of the plurality of chemicals comprise a main set of chemical reference spectra, and comparing the main set of chemical reference spectra with the absorption spectra comprises the steps of:

a. comparing the main set of chemical reference spectra to the absorption spectra;

b. determining whether there is an undetermined result in the comparison result with respect to any of the reference spectra;

c. removing the reference spectrum with undetermined results in the primary set if undetermined results do exist in the comparison results;

d. making another comparison of the main set of chemical reference spectra with the absorption spectra;

e. repeating steps b-d until no indeterminate results exist.

16. The method of claim 15, wherein after step (e), a new reference spectrum is further acquired from the secondary set of chemical reference spectra and the new reference spectrum is added to the primary set.

17. The method of claim 16, wherein steps b-d are repeated until there are no undetermined results in said primary set.

18. The method of claim 17, wherein the step of moving a new reference spectrum from the secondary set into the primary set is repeated until the secondary set is empty.

19. The method of claim 17, wherein said comparing comprises a conventional least squares analysis.

20. The method of claim 19, wherein all negative scale factor values associated with the reference spectrum form a set of residues.

Wizard program

21. A computer system for identifying chemical compounds using observable features, the system comprising;

a. a chemical substance database that correlates predefined characteristics and attributes with a plurality of chemical compounds;

b. a user interface that receives input of an observable feature;

c. a processor running software that directs the processor to identify a chemical compound in the database that corresponds to the observable feature.

22. The computer system of claim 21, wherein the observable features include color and odor.

23. The computer system of claim 22, wherein the observable features include vapor density and molecular weight.

24. A computer system for identifying a chemical compound based on a location at which the compound is observed, the system comprising:

a. a chemical/location database associating at least one chemical compound with a map location of the distribution of the compound;

b. a location input for inputting a system location into the system;

c. a processor running software that directs the processor to identify a chemical compound in the database that corresponds to the system location.

25. The computer system of claim 24, wherein the location input is provided by a global positioning system.

26. The computer system of claim 24, wherein said assigned map location in said database is in the format of a neighborhood address.

27. The computer system of claim 26, wherein the system location is in latitude/longitude format.

28. The computer system of claim 24, wherein said system identifies all compounds in said database that are within a predetermined radius of said system location.

29. The computer system of claim 24, wherein the system identifies one or more compounds associated with a distribution map location that is closest to the system location.

30. The computer system of claim 24, wherein the chemical/location database includes information from Tier1 and 2 sources.

31. The computer system of claim 24, wherein the system further comprises a container database correlating predetermined container shapes to a set of compounds and a user interface, wherein the user interface displays the container shape selected by the user.

32. The computer system of claim 24, wherein the system further comprises a listing database and a user interface relating predetermined listings to a set of compounds, wherein the user interface displays listing information selected by the user.

33. A system for remotely identifying a chemical compound, the system comprising:

a. a passive infrared spectrometer;

b. a positioner;

c. a range finder;

d. a user interface;

e. a database comprising data representing reference spectra of chemical substances, data associating observable properties with certain chemical compounds, and data associating locations with certain chemical compounds;

f. a computer processor in communication with the spectrometer, the locator, the range finder, the user interface, and the database; and

g. software for comparing data from the spectrometer, the locator and the rangefinder to the database in order to identify the possible presence and concentration of one or more chemical compounds.

34. The system of claim 33, further comprising a weather station and a compass in communication with said computer processor.

35. The system of claim 34, wherein the user interface comprises a touch screen.