HUNGARIAN JOURNAL OF INDUSTRIAL CHEMISTRY VESZPRÉM Vol. 33(1-2). pp. 75-80. (2005) SPLINE FUNCTIONS IN EVALUATION OF EXPLOSION LIMIT CURVES FOR GAS MIXTURES M. MOLNÁRNÉ-JOBBÁGY, K. KOLLÁR-HUNEK1 Federal Institute for Materials Research and Testing, BAM, D-12200 Berlin, GERMANY 1Dept. of Chemical Information Technology, Bp. Univ. of Technology and Economics, H-1521 Budapest, Pf.91. HUNGARY The paper discusses a new method to approximate the sometimes missing apex point of the explosion limit curves of flammable substances with diluents in air. The base of the new method is to vary the frame points of the co-monotonic splines using de Casteljau algorithm. We show several examples for flammable/inert/oxidising gas containing systems – selected by the program TRIANGLE – where the method was applied. Due to the definition of the frame of splines it can be stated that the new method never restricts the explosion range around the apex and shifts the explosion limit curve into the direction of higher inert gas concentrations. This means that the new method can correct the highly “cut down nose” of the co-monotonic splines and gives a safer explosion range of these systems. Keywords: Ternary flammable systems, inerting, explosion areas, co-monotonic spline curves Introduction Inerting of explosive fuel-air mixtures is a frequently applied method in the chemical and related industries to prevent fires and explosions. For this purpose the exact knowledge of the explosion range is required as a function of the flammable, oxidizer and inert gas concentrations. The CHEMSAFE® database [1], which is world wide available through STN International and Internet, contains rated safety characteristics of flammable liquids, gases, dusts and their mixtures, such as explosion limits, flash points, ignition temperatures, etc. The in-house version of CHEMSAFE® allows a graphical representation of the measured explosion range of ternary systems in triangular diagrams as a function of the concentration of flammable (combustible), oxidising or inert gases. The TRIANGLE program [6,7,9,10], created by BAM and extended by the common research group of BME and BAM, is used for processing measured values of ternary systems, it provides 2D triangular diagrams for the data processing phase of the explosion area of the gas mixtures. In our latest research we created a new test method to investigate whether the (last) measured connection point between the upper and lower explosion limit curves – the so called apex point – is the real apex point, or it is not the last point of the limiting curves. Description of the explosion range of ternary systems Beside the explosion limits and the explosion range other characteristics can be also deduced from triangular or Cartesian explosion diagrams, which parameters explicitly define the dangerous area as they are shown on Fig 1. The IAR (minimum Inert gas / Air (oxidising gas) Ratio) and ICR (minimum Inert gas / Combustible Ratio) lines represent limits in the ternary flammable system: the points lying on the right hand side of IAR line or below the ICR line will not cause an explosion regardless of the added amount of flammable gas. MAI (Minimum required Amount of Inert gas) and MXC (MaXimum permissible amount of Combustible) points are intercepts of IAR/ICR lines and the corresponding binary triangle sides. The MOC (Maximum Oxidising gas Content) is given by the 76 tangent line of the explosion limit curve parallel with the flammable-inert side of the triangle. In some cases this line passes through the apex point . Fig. 1: Characteristics of the explosion area The LEL (Lower Explosion Limit) and UEL (Upper Explosion Limit) curves are generated around the explosion area by applying numerical interpolation on the measured data. Having evaluated the characteristic values of the system the TRIANGLE program returns the results in tables and in ternary diagrams both in Cartesian and in triangle co-ordinates. Most of the above mentioned characteristics are deduced from the common point of the LEL and UEL curves, from the so called apex point, and for the ICR line and the MXC point we use a tangent line of the LEL curve. These calculations require the best possible numerical approximation of the explosion limit curves and of their apex point. Application of co-monotonic vector splines Co-monotonic parametric vector splines possess the best numerical properties for the approximation of ternary explosion limit curves [2,3,4,5,8]. The earlier used Akima splines failed to describe several systems, where the LEL or the UEL curve was not monotonic or the concatenation of the two explosion curves couldn’t be considered as an only function of the flammable gas concentration. As a result of our previous research [6,9], we created subroutines for the co-monotonic parametric vector splines and built them in the TRIANGLE program. Testing this extension of the software we have found several ternary data sets, which didn’t contain enough measurements around the most critical apex of their explosion curves. To select these data sets, we made the program to give an alert in these cases, and to offer the user the possibilities of sketching an “open apex” curve, typing in an apex from another source, or trying to make up the apex based on the last two points of the LEL and of the UEL curves. The make up of the apex, based on the last two points of the LEL and of the UEL curves gives also a possibility to investigate other - “full” - data sets whether the “common” point of the LEL and UEL can be considered as a real apex point, or it rather belongs only to the LEL or only to the UEL curve. The theoretical background of the apex-make-up is the de Casteljau algorithm. This algorithm offers a numerically very simple way to evaluate an n- dimensional point of a parametric (cubic) co-monotonic spline curve. The algorithm and the resulted point are shown on Fig.2. t=1/3 p2 p3 p11 p21 p12 p03 p0 p1p01 p02 32 1 2 21 1 1 10 1 0 )1()( )1()( )1()( ptpttp ptpttp ptpttp ⋅+⋅−= ⋅+⋅−= ⋅+⋅−= 1 2 1 1 2 1 1 1 1 0 2 0 )1()( )1()( ptpttp ptpttp ⋅+⋅−= ⋅+⋅−= ⎣ ⎦1;0 )1()( 2 1 2 0 3 0 ∈ ⋅+⋅−= t ptpttp Fig. 2.: The four frame points of a co-monotonic spline and evaluating p0 3 at t=1/3 To determine the p0, p1, p2 and p3 frame points of the co-monotonic spline we have used the last two points of the UEL curve, and the last two points of the LEL curve. If the measured data set contained an (assumed) apex point, in the first step of our algorithm we truncated the data set by this assumed apex. We evaluated a co-monotonic spline based on the remained last four points. In Tables 1 and 2 we show the steps of the frame point calculations for the systems NH3+N2+Air and CH4+CO2+Air, in this second case we give a description of the steps in every detail. One can see the frame points (black squares), and the result of the apex test for the NH3+N2+Air system on Fig 3, and for the CH4+CO2+Air system on Fig. 4. NH3+N2+Air T=24C P=1,03 bar For the secant-intercept: N2 NH3 14 15,8 LEL 16 16,15 16 16,9 UEL 14 18,05 de Casteljau frame t*= 0,66 xo yo slope x y secant1 16 16,15 0,175 p0 16 16,2 secant2 16 16,9 -0,58 p1 16,7 16,3 intercept 17 16,325 p2 16,7 16,5 delta(x)1,00 p3 16 16,9 1,00 ratio(2/1)1,00 slope ratio3,3 Table 1: Frame point calculation (symmetric case) 77 NH3 + N2 + Air 15,5 16,5 17,5 18,5 14 15 16 17 NH3 Mpoint Fig. 3: Apex test (result: the real apex was measured) In the case of NH3+N2+Air system the skipped apex lies on the approximating spline as a real apex – this means that the measured apex is a real one. The apex test suggests for the CH4+CO2+Air system that the assumed apex is not a real one. The measured point lies on the UEL curve, and a new apex point is offered showing 1% relative difference in the inert gas concentration. In this case the steps of the frame point calculations show asymmetry because the inert gas differences of the last two points on the LEL and UEL curves are not equal. The slope ratio is also in this system not too big, what means that the asymmetry is only in the different distances of the measured points. CH4+CO2+Air T=24C P=1 bar For the secant-intercept: CO2 CH4 25 5,7 LEL 28,5 6 27,8 7 UEL 25 7,9 de Casteljau frame t*= 0,75 xo yo slope x y secant1 28,5 6 0,086 p0 28,5 6 secant2 27,8 7 -0,32 p1 29,5 6,1 intercept 30,40 6,16 p2 29,8 6,4 delta(x)1,90 p3 27,8 7 2,60 ratio(2/1)1,37 slope ratio3,75 Table 2: Frame point calculation (asymmetric case) To see clearly the meaning of the results now we show step by step the frame point algorithm and the apex calculation of the asymmetric case. The bold points of Table 2 on LEL / UEL (28,5 ; 6) and (27,8 ; 7) are chosen for the frame points p0 and p3. Using the other two points – (25 ; 5,7) on LEL and (25 ; 7,9) on UEL – we determined the equation of the lines passing through the given points of LEL/UEL. The intercept (pm) and the slopes (s1 and s2) of the two lines and the ratio of the two slopes contain the shaded cells of Table 2. The next step is to calculate the t* parameter that determines the p2 frame point on the line segment of p3 and pm: If abs(s2/s1)<5 then t*=abs(s2/s1)/5 else t*=0,99 with this t* : p2 = p3 + t*⋅(pm-p3) (1/a) Now we calculate the ratio of the horizontal distances between the secant line intercept and p0 / p3 LEL / UEL points: ratio(2/1)=2,6/1,9=1,37=a21 with this a21 : p1 = p0 + t*⋅(pm-p0)/a21 (1/b) Having determined the missing two frame points, any arbitrary point of the interpolating co-monotonic spline can be calculated by the de Casteljau algorithm. CH4 + CO2 + Air 5 6 7 8 25 26 27 28 29 30 CH4 Mpoint Fig. 4: Apex test (result: not the real apex was measured) The apex calculation is very simple: we have to determine the maximal abscissa of the co-monotonic spline. The key of this apex calculation lies in the parameter t* that determines by the missing two frame points the apex, too. The original formula for t* is the following: If abs(s2/s1)