You are on page 1of 20
HOT MIX ASPHALT FOR RAILROAD TRACKBEDS— STRUCTURAL ANALYSIS AND DESIGN YANG H. HUANG,! CHIANG LIN,” and XUEJUN DENG? INTRODUCTION The recent increase in traffic speeds, volumes and loads on railroads has led to new requirements for improved stability and durability of track sup- ports. To reduce both resilient deformations and permanent settlements, and thus minimizing maintenance costs, hot mix asphalt (HMA) has been used for strengthening rail structures. It can be used either as an overlay- ment by placing directly under the ties or as an underlayment by placing under the ballast, as illustrated in Figure 1. With the joint financial support from the National Asphalt Pavement Association and The Asphalt Institute, research has been underway since 1981 at the University of Kentucky on the use of hot mix asphalt for railroad trackbeds. The construction, per- formance and overview of HMA trackbeds were reported in a companion paper by Rose et al. (1), whereas the structural analysis and design of HMA trackbeds are presented in this paper. cra + TF no base course is used, the overlaynent is called full donth fa) OVERLAYMENT SUBCRADE (0) UNDERLAYMENT Fig. 1. Schematic Diagrams of Overlayment and Underlayment. Aprofessor of Civil Engineering, University of Kentucky, Lexington, Kentucky. Graduate Assistant, University of Kentucky, Lexington, Kentucky. 3Lecturer in Civil Engineering, Nanjing Institute of Technology, People’s Republic of China. Due to lack of time on the program, this paper was presented by title only. 475 476 HUANG, LIN AND DENG This paper is concerned with both analysis and design. With the advent of high speed computers and the finite element method, it is not difficult to analyze a complex structure such as a HMA trackbed by the use of a computer model. However, a highly sophisticated model, which accurately simulates the behavior of a structure in minute details, would require con- siderable computer time to run and therefore might not be suitable for de- sign purposes. To provide a model for routine design, some simplifying as- sumptions must be made to keep down the computing cost within the reach of general users. After reviewing several existing models, an efficient and versatile model named KENTRACK was developed (2). Although this model can be used for both analysis and design, it should be noted that the design procedures incorporated in the model are patterned after those used for asphalt highway pavements. As the loading conditions in HMA track- beds are different from those in asphalt highway pavements, the parameter values used for the design of highway pavements may not be applicable to HMA trackbeds. However, in the absence of experience and field data on HMA trackbeds, highway criteria can serve as a guide for the design of HMA trackbeds. These criteria shall be modified as more experience is gained through their use. This paper first reviews several existing computer models, which can be used for the analysis of HMA trackbeds, followed by a description of KEN- TRACK. Results obtained from KENTRACK are then compared with those from the existing models as well as with the measured deflections at the Ravenna Yard test site in Kentucky. The design procedures and their application to Kavenna Yard are presented, and the effects of several im- portant variables on the design of HMA trackbeds are illustrated. Although the proposed design procedures are tentative, they provide the first step to put the design of HMA trackbeds on a rational basis. REVIEW OF EXISTING MODELS The analysis of HMA trackbeds is quite complex and involves the inter- actions among a large number of components. Therefore, the use of numerical methods using computer models is needed. Before KENTRACK was developed, a review had been made on three existing models; viz. FEARAT (3), ILLITRACK (4), and GEOTRACK (5). The limitations of these models for the analysis of HMA trackbeds are briefly described below. FEARAT Model The FEARAT (Finite Element Analysis of Railway Asphalt Track) model was developed at the University of Maryland (3, 6( for the analysis of full depth asphalt trackbeds. The analysis is based on linear elastic theory and is divided into three stages. The tackbed is first considered as a one dimensional beam on a Winkler foundation extending over a length of 70 ties, then as a two dimensional plate on a Winkler foundation over RAILROAD TRACKBEDS 477 ten ties, and finally as a two dimensional plane strain continuum on an elastic solid foundation over four ties. The model can only handle a full depth hot mix asphalt on a soil subgrade but can be easily modified to suit the conditions at the Ravenna Yard, where a layer of granular material exists between the full depth asphalt and the subgrade. This model involves many assumptions and approximations. [ the plane strain analysis, it is assumed that the most critical section is located along the edge of the plate; whereas in reality it is located near the rail, which is quite far away from the edge. Although the model predicts the maximum stresses and strains in the trackbed with reasonable results, it is inadequate to determine the vertical deflections because the deflections are determined in the second stage, where the trackbed is considered as a plate on a Wink- ler foundation. Once the plate deflections are determined, they are super- imposed on the surface and on the boundary of the asphalt layer so that the same deflections are obtained in the third stage. Therefore, the use of the plane strain analysis is merely for the purpose of computing stresscs and strains in the trackbed and contributes nothing to the vertical deflec- tions. This may explain why a large discrepancy in vertical deflections exists between FEARAT and the other models. ILLITRACK Model ‘This model was developed at the University of Illinois (4) and designed principally for conventional railroad tracks composed of ballast, subballast and subgrade. As the properties of granular materials depend strongly on the state of stresses, the nonlinear properties of each material are consid- ered. The analysis is divided into two stages. The trackbed is first con- sidered longitudinally as a two dimensional plane strain continuum with 26 ties on each side of the plane of symmetry and then transversely as a two dimensional plane strain continuum. The model can be applied to HMA. trackbeds by replacing the properties of ballast by those of hot mix asphalt. However, the tensile strain at the bottom of hot mix asphalt thus deter- mined may not he accurate due to the relatively thick layers of finite ele- ments employed. A major difficulty in using ILLITRACK is the assignment of a tie bear- ing length under each rail for longitudinal analysis and an angle of distribu- tion for pseudo plane strain analysis. It was suggested that a tie bearing length of 18 in. (457 mm) and an angle of distribution of 10 deg be used. However, when the ballast is replaced by hot mix asphalt, both values should increase. Unfortunately, the assignment of different values has sig- nificant effects on the results obtained. GEOTRACK Model This model was developed at the University of Massachusetts (5) for conventional railroad tracks but can be applied to HMA trackbeds by re- placing the properties of ballast by thosc of hot mix asphalt. The model treats the ballast, subballast and subgrade as a layered system, thus the 478 HUANG, LIN AND DENG Burmister’s layered theory, so well known in the design of highway and airport pavements, can be applied. Although the layered system is assumed to be infinite in a real extent, while the hot mix asphalt is of limited width, this assumption should not affect the accuracy of results because the rail load is applied 2 to 3 ft (0.6 to 0.9 m) from the edge. The model considers nonlinear properties and involves fewer approximations as compared to the above two models. It obtains the required solutions in a single stage rather than two or three stages. One important assumption made in GEOTRACK is that each wheel load, when applied over a central tie, is distributed over 11 ties, five on each side of the central tie, Any tie at a distance of six or more ties away from the central tie does not carry any load. This assumption is reasonable for de- sign purposes since the most critical stress and strain lie in the vicinity of the central tie. It allows the application of the simple superposition prin- ciple to determine the stresses and strains under multiple wheel loads and thus saves a great deal of computer time. A major drawback of GEOTRACK is that the rail load is not applied to the tie directly above one of the reactive points between the tie and the layered system. As a result, the stresses and strains computed in the layered system are those under the reactive points, instead of those under the rail load, which are usually the most critical. In other words, GEOTRACK may miss this maximum values and result in a critical stress or strain which is too small. This situation is avoided in KENTRACK by placing the rail at one of the reactive points. DESCRIPTION OF KENTRACK The above review indicates that GEOTRACK is a more realistic model for track simulation and can be used as a framework upon which KEN- TRACK is to be built. Figure 2 shows the various components in a tack system as considered in both GEOTRACK and KENTRACK. Starting from top to bottom, the track is divided into rails, spring fasteners, ties and layered system. The vertical stress between tie and layered system is dis- tributed over circular areas of equal diameter, so the Burmister’s layered theory applies. Differences between KENTRACK and GEOTRACK In addition to the use of more efficient methods and realistic approxi- mations to reduce the computer time required, the major difference be- tween KENTRACK and GEOTRACK is that KENTRACK is a finite ele- ment model and GEOTRACK is not. In GEOTRACK, rails and ties are considered as integral beams and are not divided into finite elements. Con- sequently, they must be homogencous and uniform in cross section and cannot have joints or discoiscinuities. KENTRACK, on the other hand, divides rails and ties into finite elements. Although jointed rails are not considered in KENTRACK, they can be easily incorporated by including RAILROAD TRACKBEDS 479 LAYER 1 LAYER N-1 LAYER N, SUBGRADE Fig. 2, Components in a Track System. the rotational stiffness of joint as an input parameter, KENTRACK can be applied to concrete ties with variable cross sections. By assigning a zero moment of inertia to a certain finite element ina tie, the case of discon- tinuous ties can be modeled. In KENTRACK, each section of rail hetween two ties is considered as a beam element with two nodes. Any wheel load applied to the rail between two ties is distributed to the nodes by statics. The rails and ties are con- nected by tie plates modeled as a spring element. The tie is also divided into a number of beam clements, The loads from tic plates and the reactive forces from the layered system are applied to the tie at the nodes, The stiffness matrix of the layered system, which relates the reactive forces to the vertical displacements of the ties, are determined from the Burmister’s layered theory and added to the stiffness matrices of rails, springs and ties to form the overall stiffness matrix. By equating the nodal forces due to nodal displacements to those due to wheel loads, a set of simultaneous equations is obtained for solving the nodal displacements. Once the nodal displacements are determined, the stresses and strains in rails, ties, and the layered system can be computed. A significant addition to KENTRACK is the incorporation of two failure criteria for the design of HMA uackbeds: one is the maximum horizontal tensile strain at the bottom of asphalt layer for fatigue cracking, and the other is the maximum vertical compressive stress on the surface of subgrade for permanent deformation. The tensile strain at the bottom of asphalt 480 HUANG, LIN AND DENG layer and the compressive strain on the surface of subgrade have frequently been used for the design of highway pavements. However, the HMA track- beds due to the difference in loading conditions, it was found that the ver- tical compressive stress, instead of strain, on the surface of subgrade was a better indicator (7). These stresses and strains are used for damage compu- tations based on the cumulative damage concept. Depending on traffic and weather conditions, each year is divided into a number of periods up to 12 months, Damage computed for each period, based on the given traffic and material properties, is accumulated up to a damage value of 1.0, using the hypothesis of the linear summation of cycle ratios for both distress modes. The design life and number of load repetitions to failure are summarized for both cracking and detormation, and the one with a lesser lite controls the design. Model Capabilities KENTRACK can be used not only for HMA trackbeds but also for con- ventional ballast trackbeds as well as cement concrete slab tracks. Its capa- bilities, as applied to HMA trackbeds, are summarized as follows: (a) By superposition, any number of axle loads can be analyzed. For conventional trains, only the two axle loads at the end of one car need be considered in design. The two axle loads in the adjoining car result in neg- ative moments and actually reduce the critical stress and strain, so the use of two axles, instead of four, is on the safe side. (b) Any layer of granular materials can be nonlinear elastic with a resil- ient modulus determined by M, = k, 6 a in which M, = resilient modulus; @ = bulk stress, or the sum of three prin- cipal stresses including geostatic stresses; and k,, k, = regression constants reflecting material properties. The constant k, can vary thronghont the year but k, is assumed to be constant. When k, = 0, a linear layer is im- plied. When one or more layers are nonlinear, an iteration procedure is invoked. First, a set of moduli is assumed and the bulk stress at the mid- depth of each nonlinear layer under the three ties closest to the maximum load are determined and averaged. Based on Equation [1], a new set of moduli is computed. The process is repeated until the moduli converge to a specified tolerance. (c) The dynamic modulus of hot mix asphalt is linear, independent of the state of stresses but dependent on temperature. Equation [2] devel- oped by Hwang and Witczak (8) for The Asphalt Institute can be used. E, = 108 x 10% (2a] By = By + 0.000005 f, - 0.00189 pf ** [2b] By = Bao THs [2c] RAILROAD TRACKBEDS 481 Bs = 0.553833 + 0.028829 Prog £7 - 0.03476Vy [2d] + 0.070377n + 0.931757£0027 Bi: = 0.483V_ [2e} Bs = 1.3 + 0.49825 log f 124] in which 8,, 62, 85, Bs, 8s = temporary constants; E, = dynamic modulus of hot mix asphalt in psi; T = temperature in °F; f = load frequency in Hz; Pao = aggregate passing no. 200 sieve in percent; Vp = volume of bitumen in percent; Vy = volume of air voids in percent; n = asphalt viscosity at 70 F (21 C) in 10° poses, and log = base 10 logarithm. (d) The subgrade soil is linear elastic with an elastic modulus varying throughout the year. (e) Because Poisson’s ratio has a relatively small effect on stresses and strains, a constant Poisson’s ratio is assumed throughout the year. COMPARISON OF RESULTS To compare the results obtained from KENTRACK with those from other models as well as with the measured deflections from the field, an analysis was made on the overlayment installation at Ravenna Yard, KY. Figure 3 shows the 8 in. (200 mm) HMA trackbed at station -3+50, where the deflection profiles of rail and asphalt layer were measured. Two hot mix asphalt samples were cored at this vicinity and shipped to the Asphalt Institute for laboratory tests. It was found that the samples had an average dynamic modulus of 1.9 X 105 psi (1.3 GN/m?) at 77 F (25.0 C) under a frequency of 1 Hz. Because the temperature of the as- phalt layer during the deflection measurement was about 77 F (2.5.0 C), as indicated by thermisters embedded in the hot mix asphalt, 2 modulus of 1.9 X 10° psi (1.3 GN/m?) was used for the analysis. Laboratory tests also show that the cored samples have the following properties: P99 = 6.6 per- cent, Vp = 11 percent, Vy = 10 percent and n = 2.5 X 10° poises. If these values are substituted into Equation [2], about the same modulus value is obtained, indicating that Equation [2] gives a good prediction on the dy- namic modulus of hot mix asphalt. Based on the results of Road Rater tests supplemented by laboratory CBR and field Dutch cone tests at this location, a modulus of elasticity of 15,000 psi (104 MN/m?) and 7,500 psi (52 MN/m?) was selected for the old subballast and the subgrade, respec- tively. The depth from the surface of hot mix asphalt to bedrock is 16 ft (4.9m). Details about the Ravenna Yard test site can be found elsewhere (9). Comparison of Computer Solutions The following parameters, based on the actual conditions at the test site, were uscd in the computer analysis: (a) rail (RE132)—scction modulus 482 HUANG, LIN AND DENG ATS OF SyaMeTRY 35 in. | ween LoD (55,000 15) ASPHALT _E = 190,000 psi ‘SUBBALLAST E = 15,000 psi -v = 0.85 30 in. SUBERADE, E= 7,500 psi -v = 0.00 156 in, BEDROCK °™ SS eF—>7 ° 2 2 * poe at auto gi ia Ooo = 1 wasp 22 sl gon Eoasf 2 LeceND : [FIELD MEASUREMENT © FEARAT = revraace © TALITRACK, CASE T 2 GeoTRAcK fo TILITRACK, ASE 11 Fig. 3. Comparison between Computed Deflections and Experimental Measurements. (1 in. = 25.4 mm; 1 Ib = 4.45 N; and 1 psi = 6.9 kN/m?), 22.5 in.? (3.69 X 105 mm), moment of inertia 88.2 in.4 (3.67 X 107 mm‘), modulus of elasticity 3 X 10” psi (210 GN/m’), gauge (center to center of rail head) 59.2 in. (1.33 m), and fastener stiffness 1 X 10° Ib/in. (180 MN/mm); (b) tie—thickness 6.75 in. (170 mm), width 9 in. (230 mm), length 102 in. (2.59 m), center to center spacing 21 in. (530 mm), and modulus of elasticity 1.25 X 10 psi (8.63 GN/m?); (c) hot mix asphalt— thickness 8 in. (200 mm), width for use in FEARAT 138 in. (3.51 m), modulus of elasticity 1.9 X 10 psi (1.3 GN/m?), and Poisson’s ratio 0.45; (d) subballast—thickness 30 in. (760 mm), modulus of elasticity 1.5 x 10* psi (104 MN/m?), Poisson’s ratio 0.35, and spring constant for use in FEARAT 200 pci (55 MN/m?); subgrade—modulus of elasticity 7.5 X 10° psi (52 MN/m?), and Poisson’s ratio 0.4; (e) load—number of axles 2, wheel load 33,000 lb (147 KN), and axle spacing 70 in. (1.8 m). RAILROAD TRACKBEDS. 483 Table 1 shows a comparison of solutions obtained from the four com- puter models. Two cases are considered for ILLITRACK: Case | with a tie bearing length of 18 in. (460 mm) and a distribution angle of 10°, and Case Il with a tie bearing length of 36 in, (920 mm) and a distribution angle of 20°. It can be seen that the two different sets of assumptions result in a large difference in solutions. As the suonger hoc mix asphalt will disuribute the load over a larger area, the use of Case II is more reasonable than Case I. However, the tensile strain at the bottom of asphalt layer is much smaller in Case II and does not check with the other models. Because GEOTRACK and KENTRACK are based on a subgrade of infinite depth, whereas FEARAT and ILLITRACK consider the bedrock at a depth of 16 ft (4.9 m) below the surface of asphalt concrete, the computed deflections of the former are reduced by an amount equal to the deflections at the 16 ft (4.9 m) point. Table 1, Comparison of Solutions by Different Models Maxima Values FEARAT TLLITRACK GEOTRACK KENTRACK case T Case IT Deflection at Rail (ia-) 0.0398 - - 0.0802 0.0790 Deflection at Tie (tn) 0.0390 0.1285 0.0642 0.0702 0.0693 Deflection at top of asphalt layer (in.) 0.0447 0.1283 0.0640 0.0710 0.0693 Deflections at top of subballast layer (in) 0.0439 0.1280 0.0638 0.0708, - Deflection at top of subgrade (n+) 0.0321 0.1000 0.0509 0.0576 0.0542 Tensile stress at bottom of asphalt layer (pat) 40.72 28.58 20.76 51-01 62.48 Coupressive stress at top of subgrade sotl (pot) 3.34 13.92 9.99 bez gus Tensile strajn at bottom of asphalt layer (x 10™%in./in.) 2.00 1.88 1.25 1.96 2.39 Compressive strain at top of subgrade sot (x 10~tn./in.) 5.47 11.83 5.92 6-81 6.70 Note: 1 in. = 25.4 mm and 1 pei = 6.9 kN/a? Tt can be seen from Table 1 that, even though there is a large difference in deflections between FEARAT and the other three models, the computed tensile strains at bottom of asphalt layer and compressive stress on the sur- face of subgrade, which govern the design of HMA trackbeds, are not too much different. Note that the deflections and the compressive stress at top of subgrade obtained from KENTRACK are nearly the same as those ob- tained from GEOTRACK; however, the large differences in the tensile stress and strain at the bottom of asphalt layer are due to the fact that GEOTRACK misses the most critical point. 484 HUANG, LIN AND DENG Figure 3 shows a comparison between the computed deflections ob- tained from the four models and the experimental measurements. The de- flection measurements were made on the rail and on the top of asphalt con- crete adjacent to the rail at each tie location. Metallic rulers, graduated to 1/64 in. (0.4 mm), were attached to the rail as well as to the steel rod rest- ing on the top of hot mix asphalt through a hole in the ties. The elevations were taken by using a transit, first before and then after the whccl loads were applied. The difference between the two elevation readings is the de- flection due to the applied wheel loads. In Figure 3, the experimental measurements are shown by the symbol I, the height of which indicates the possible range of measured values because the measurements could be estimated only to the nearest 1/64 in. (0.4 mm). It can be seen that the results from KENTRACK, GEOTRACK and ILLITRACK (Case II) check tavorably with the measured deflections. DEVELOPMENT OF DESIGN PROCEDURES One of the uncertainties in the design of HMA trackbeds is the selec- tion of proper failure criteria. In the absence of past experience on the use of HMA trackbeds, the best resort is to study the failure criteria used in asphalt highway pavements and make certain modifications if needed. It is believed that the use of highway criteria for railroad design is quite conservative because the materials in railroad trackbeds are in a more confined state. Therefore, highway criteria can be used conservatively as a starting point for railroad design. As more experience is gained, the criteria can be modified to suit railroad conditions. Fatigue Criterion The horizontal tensile strain at the bottom of asphalt layer is used to predict the fatigue life of hot mix asphalt. Figure 4 shows the tensile strain criterion used in The Asphalt Institute’s thickness design manual (MS-1), as indicated by the solid lines, and in the Shell Petroleum Company’s method, as indicated by the dashed lines (10, 11). It can be seen that under large strains The Asphalt Institute criterion permits a greater number of repeti- tions, whereas under small strains the reverse is true. Because the deforma- tion criterion seer by Shell was used in the analysis that follows, the Shell criterion was also used in the fatigue analysis. The criterion, as indi- cated by the dashed lines in Figure 4, can be represented by Ne = 0.0685 ef $47 By? 3] in which N¢ = number of repetitions to failure and ¢; = tensile strain at hottom of asphalt layer RAILROAD TRACKBEDS 485 4 = Sxt0* psi (3.5x108 Pa) ASPHALT LAYER TENSILE STRAIN (x10°4) NUMBER OF REPETITIONS Fig. 4. Tensile Strain Criterion for Fatigue Cracking. Deformation Criterion The vertical compressive strain on the surface of subgrade is used to pre- dict the permanent deformation of asphalt highway pavements. Figure 5 shows the relationship between strain and the number of repetitions to failure developed by the Shell International Petroleum Company (12) and the Chevron Research Company (13). The Chevron criterion was also in- corporated in The Asphalt Institute’s method of pavement design (10, 11). The small circles and the least square line are obtained from the AASHTO (American Association of State Highway and Transportation Officials) equation, as will be cxplained later. It can be seen that for a given com- pressive strain, the allowable number of repetitions based on Shell criterion 20 SUBGRADE COMPRESSIVE STRAIN (x1074) 105 108 107 10® 10 NUMBER OF REPETITIONS Fig. 5. Subgrade Strain Criterion for Permanent Deformation. 486 HUANG, LIN AND DENG is about ten times greater than that based on Chevron criterion. In view of the fact that highway criteria are conservative when applied to railroad loadings, Shell criterion based on vertical compressive strain was used for conversion to the stress criterion. From the straight line in Figure 4, Shell criterion can be represented by Np = 6.15 X 107 ec#? [4] in which e. = compressive strain on the surface of subgrade, Fquation [4] can be converted to a stress criterion based on the AASHTO equation. The reason that Equation [4] cannot be applied to railroad trackbeds is because, unlike highway pavements, the vertical compressive strain is un- predictable and varies in an erratic manner (7). Theoretically, the vertical compressive stress is the best parameter for predicting permanent deforma- tion. However, in using the compressive stress, the stiffness or modulus of subgrade should also be specified because a stiffer subgrade permits a greater stress. In highway pavements, the vertical strain depends to a large extent on the vertical stress. Because the modulus of subgrade has already been considered in computing the strain, the number of repetitions to fail- ure depends on the compressive strain only, independent of the modulus of subgrade. In railroad trackbeds, due to the large area of load distribution, the vertical strain depends strongly on the horizontal stresses and therefore is not a good indicator of permanent deformation. Equation [4] can be converted to Ne = 4.837 X 10°8 og 3-784 B,3-583 U5] in which o, = vertical compressive stress on the surface of subgrade and E, = elastic modulus of subgrade. The derivation of Equation [5] was re- ported elsewhere (2) and is briefly described below. The conversion from strain to stress is based on the AASHTO equation relating the number of load repetitions to the structural number and the soil support value (14). Por a typical full depth asphalt pavement under an 18,000 lb (80 KN) single axle load, the structural number, which depends on the pavement thickness, and the soil support value, which depends on the modulus of subgrade, were varied and the maximum vertical compres- sive strain and the corresponding vertical compressive stress were computed by the Burmister’s two-layer theory. The relationship between the number of load repetitions and the compressive strain were plotted in Figure 5 as indicated by the small circles. It can be seen that these circles lie mostly between the Shell and Chevron criteria, indicating that the use of full depth asphalt pavement in the analysis yields reasonable results. By substituting each strain into Equation [4], the number of repetitions for each corre- sponding stress was determined. A plot of compressive stress versus num- ber of repetitions on a log-log scale results in a series of nearly parallel straight lines, one for each subgrade modulus. The average slope of these straight lines is 3.734, which is the exponent of u¢, as indicated in Equation RAILROAD TRACKBEDS 487 [5]. By setting o, = 1 psi (6.9 KN/m?), the coefficients 4.837 X 10° and 3.584 can be determined by plotting log E, versus log Ng. Application to Ravenna Yard Equation [2], [3], and [5] were used with KENTRACK for determining the design life of the track at Ravenna Yard. Because of the very low speed permitted, a frequency of 1 Hz was assumed. In the analysis, a year was divided into four scasons and the dynamic modulus of hot mix asphalt dur- ing each season was obtained from Equation [2] using the temperature of hot mix asphalt actually measured in the field. In the damage analysis, the old subballast with an elastic modulus of 15,000 psi (104 MN/m?) and the subgrade with an elastic modulus of 7,500 psi (52 MN/m?) were combined to form a composite subgrade with an elastic modulus of 12,000 psi (83 MN/m?). This combination is necessary because the subballast is very thick and its modulus is not too much different from that of subgrade. Although KENTRACK can consider the subballast and the subgrade as two separate layers and determine the expected life of each based on the compressive stresses of the subgrade and the subballast, respectively, the life based on either stress will be too long. The life based on the subgrade stress is too long because of the very small compressive stress due to the great thickness of subballast used; while that based on the subballast stress is also too long because the allowable number of stress repetitions is based on the modulus of subballast alone, regardless of the weaker subgrade. It is therefore recommended that, whenever a thick layer of material with a modulus not too much different from the subgrade is employed, a composite modulus should be used. Although the modulus of composite subgrade could have a different value for each season, a constant modulus of 12,000 psi (83 MN/m?) was assumed. It is estimated that the HMA trackbed will be sub- jected to 3 trains per day each carrying 100 cars with a wheel load of 33,000 lb (147 KN). Assuming that traffic is uniformly distributed throughout the year, the total number of load repetitions during each sea- son of three months is 91.5 X 3 X 100 = 27,450. The results of the analysis are shown in Table 2. The damage ratio is a ratio between the actual number of repetitions and the allowable number of repetitions, as determined from Equation [3] or (5]. It can be seen from Table 2 that during the summer more damage is caused by permancut deformation but less damage by fatigue cracking; while during the winter more damage is caused by fatigue cracking but less damage by permanent deformation. The reciprocal of the total damage ratio summed over the four seasons is the design life, which is 12.3 years based on fatigue cracking and 13.9 years based on permanent deformation. It should be noted that the above design is based purely on highway cri- teria. A design life of only 12.3 years does not appear satisfactory. The design life can be incrcascd by changing thc failurc critcria. Thc most ap- propriate criteria cannot be determined until more HMA trackbeds are constructed and analyzed. 488 HUANG, LIN AND DENG Table 2. Damage Analysis of HMA ‘Trackbeds at Ravenna Yard Season, Spring Suamer Fall Winter Total Design 48.2°F = -75.2°F 59.0°F —-36.5°F Damage Life ‘Tenperature (9.0%) (24.0%) (15.0%) (2.5%) | Ratio (Year) Elastic Modulus 682,980 204,192 429,598 1,090,399) (pst) haw Tenetle Strain cao) 1.98 2.96 2.38 1.59 Damage Ratio 0.0243 0.0132 0.0229 9,021 | 0.0815. 12.3 Compressive Stress 11.5 14.8 12.9 10.2 lsubgcadd con Damage Ratio 0.0126 0.0322 0.0192 0.0080 | 0.072 13.9 Mote: 1 pet = 6-9kN/a” PARAMETRIC STUDIES To show the cffect of scvcral important factors on the design of HMA trackbeds, parametric studies were made on both overlayment and under- layment by varying one parameter but keeping the others unchanged. Fig- ure 6 shows two standard cases of HMA trackbeds, one for overlayment and the other for underlayment. The dimensions and properties of rails and ties used for the parametric study are the same as those for the Ravenna Yard, except that the tie spacing is changed to 20 in, (510 mm). ‘The elastic modulus and Poisson’s ratios shown in the figure are typical for use in track construction. In the standard case of underlayment, no base material is placed below the asphalt layer. If a base course exists, an elastic modulus of 15,000 psi (104 MN/m?) is used because the stress level in underlayment is smaller than that in overlayment due to the presence of the extra ballast layer. Effcct of Hot Mix Asphalt and Base Thicknesses Figure 7 shows the effect of hot mix asphalt and base thicknesses on tensile strains in both overlayment and underlayment. In the case of over- layment, when the hot mix asphalt is in the practical range of 4 to 8 in. (100 to 200 mm) thick, the thickness of base course has a significant effect on the tensile strain, but the thickness of hot mix asphalt has very little ef- fect. In the case of underlayment, the effect of base thickness on the ten- sile strain is relatively small. Figure 8 shows the effect of hot mix asphalt and base thicknesses on compressive stresses in both overlayment and underlayment. It can be seen RAILROAD TRACKBEDS 489 that the thickness of base course has a much greater effect on compressive stresses in overlayment than in underlayment. It is therefore recommended that base course be used for overlayment but not for underlayment. In the following analysis of underlayment, the asphalt concrete is placed directly on the subgrade and no base course is used. Effect of Ballast and Hot Mix Asphalt Thicknesses Figure 9 shows the effect of ballast and hot mix asphalt thicknesses on tensile strains and compressive stresses in underlayment. The thickness of fp — win, 4 mE 4 in, _ HOT MIX ASPHALT E = 600,000 psi_v 10 in. BASE MATERIAL Es 30,000 psi v = 0.35 = SUBGRADE SoTL B= 6,000 psi v = 0.40 (a) OVERLATIENT be win 4 TIE TE 60,000 psi v= 0.35 8 in, HOT MIX ASPHALT 600,000 psi v = 0.45 (2 SUBGRADE SoTL E+ 6,000 psi v = 0.40 (b)_UNDERLAYHENT. Fig. 6. Standard Cases for Parametric Studies. (1 in. = 25.4 mm; 1 psi = 6.9 kN/m?). hot mix asphalt has more effect on the compressive stress than on the ten- sile strain, The effect of ballast thickness on both tensile strains and com- pressive stresses is quite significant. It is therefore recommended in the design of underlayment that both ballast and asphalt concrete thicknesses be varied to keep these stress and strain within allowable limits. 490 HUANG, LIN AND DENG HOT MIX ASPHALT THICKNESS (m=) HOT MIX ASPHALT THICKNESS (em) 350, 280350450 150, 250 350 as0 3.0 3.071 1 2s 2s 7 2.0 7 2.0) 7 ee Ld as Saf Sos z tin easem PQ Bie Line 25.8 tot | ao. : 8 BW HoT NUK ASPHALT THICRNESS.Cin-1 HoT MT ASPHALT THTCRNESS Cin.) (1) ovemcarinr (oy emencanert Fig. 7. Effect of Hot Mix Asphalt and Base Thicknesses on Tensile Strains. Effect of Hot Mix Asphalt Moduli. Figure 10 shows the effect of hot mix asphalt moduli on normalized tensile suains and strain repetitions in both overlayment and underlayment. HOT-MIX ASPHALT THUCKKESS (00) HOT MIX ASPHALT THICKNESS (am) 15025035050, 15025035080 MY ie a oP |. ; 2 R 7 20,0 wot eno bos 5 0 | n¢ 5 — E 5 2 lio & oo Be BE s 5 = bo 2 f 30 B 5 ee PQ z 8 we | — E, 8 6 : Tab ase Lao ; 0 in 4 in 4 in 1 ine 28.4 ma 0 in Line 25.4 oa # inbso \ 13 inf 50 ade {jf ‘ t — a s uw 1 4 % 2 6 HOT MIX ASPHALT THICKNESS (in.) HOT MIX ASPHALT THICKNESS (in.) (2) OVERLAWENT () UNDERLAYIENT Fig. 8. Effect of Hot Mix Asphalt and Base Thicknesses on Compressive Stresses. RAILROAD TRACKBEDS 491 WOT MIX ASPHALT THICKNESS (9) oT MIX ASPHALT THICINESS (am) 1030 aot tgo 3800 ag 2s 1 et 10 , Les 20 Ss se * & z £ TN g Las ¥ gts E ost g 6 a Las 8 g ——~ 8 in] a g 5 12 in| g g B10 g¢ e é T 16 in = 4 bas 7 ee Lin g ost ff | af | 4 8 2 16 4 8 2 6 MOT MIX ASPHALT THICKNESS (in) HOT MIX ASPHALT THICKNESS Cin.) (a) TENSILE STRAIN, (0) COMPRESSIVE sTRESS Fig. 9. Effect of Ballast and Hot Mix Asphalt Thicknesses on Tensile Strains and Compressive Stresses. For the standard cases shown in Figure 6, the critical tensile strain is 1.90 x 10° for overlayment and 1.58 X 10% for underlayment, and the corre- sponding number of strain repetitions according to Equation [3] is 1.93 X 10° and 5.49 X 10°, respectively. The normalized tensile strain and strain tepetition are the strain and repetition ratios with respect to the standard case. It can be seen that an increase in the clastic modulus of hot mix as- phalt decreases the strain in underlayment more significantly than that in HOT MIX ASPHALT MODULUS (60/n?) HOT MIX ASPHALT MODULUS (81/m2) 4 4 10 2 6 8 wo. no 2 63 TENSILE STRAIN = REPETITION TENSILE STRAIN REPETITION 1.0 NORMALIZED STRAIN REPETITION NORMALIZED TENSILE STRAIN NORMALIZED STRAIN REPETITION NORMALIZED TENSILE STRAIN 6 10 HOT MIX ASPHALT MODULUS (x10Spsi) HOT MIX ASPHALT HooULUS (x108psi) (3) OVERLAYMENT ()UNDERLAYMENT Fig. 10, Effect of Hot Mix Asphalt Moduli on Normalized Strains and Repetitions. 492 HUANG, LIN AND DENG the overlayment. Asa result, the allowable number of repetitions for over- layment decreases significantly with the increase in hot mix asphalt moduli. The effect of hot mix asphalt moduli on the number of repetitions is not significant in underlayment because the decrease in tensile strain compen- sates for the increase in modulus. This indicates that stiffer asphalt may be used in underlayment to reduce the compressive stress on the surface of subgrade. However, to increase the fatigue life during cold weather, more flexible asphalt should be used for overlayment. SUMMARY AND CONCLUSIONS A review was made of several computer models, which could be used for the structural analysis of HMA trackbeds. Based on this review, a new model called KENTRACK was developed. KENTRACK is a very efficient computer model for the design of HMA trackbeds and possesses many ca- pabilities which cannot be found in other modcls. Results obtained from KENTRACK were compared with those from other models as well as with the measured deflection profile at the Ravenna Yard test site in Kentucky and found to be in good agreement. In the design of asphalt highway pavements, two failure criteria have fre- quently been used: one is the maximum horizontal tensile strain at the bottom of asphalt layer for fatigue cracking, and the other is the maximum vertical compressive strain on the surfacc of the subgrade for permanent deformation, These criteria developed by the Shell International Petroleum Company were used for the design of the HMA trackbed at Ravenna Yard and for parametric studies. Due to the difference in highway and railroad loadings, the compressive strain criterion was converted to a compressive stress criterion. The conversion procedures, based on the AASHTO equa- tion, are briefly described. One major problem in the design of IIMA trackbeds is the lack of failure criteria. Equations (3] and [5] are based on highway loadings and may not be applicable to railroad loadings, because the materials under railroad tracks are in a more confined state as compared to under highway loadings. The allowable number of repetitions for the standard cases shown in Fig- ure 6 is in the order of 108, which is considered too small if one car ina train is considered as one repetition. When a train travels at a high speed, the stress or strain undcr onc carload is not completely released before the application of the next carload. Therefore, it is doubtful that one car should be considered as one repetition. It is recommended that tentative criteria be established and more experimental trackbeds be constructed. These criteria can then be modified as more experience is gained through their use. A parametric study was made by KENTRACK and results in the follow- ing conclusions: (a) To reduce the critical stress and strain, a high quality base course should be used in overlayment. A base course is not effective in underlay- ment and can be omitted. RAILROAD TRACKBEDS. 493 (b) In the design of underlayment, ballast thickness has a significant ef- fect on the critical stress and strain and should be varied according to sub- grade conditions. The thickness of asphalt concrete is effective in reducing the compressive stress but not so effective in reducing the tensile strain, especially when the asphalt concrete is less than 8 in. (200 mm) thick. (c) A stiffer asphalt concrete may be used in underlayment to decrease permanent deformation, but a softer asphalt concrete should be used in overlayment to reduce fatigue. ACKNOWLEDGMENTS The financial support given by the National Asphalt Pavement Associa- tion and The Asphalt Institute for the conduct of this research is appre- ciated. Thanks are due Richard W. Smith, NAPA’s former Director of Re- search and Development; James F. Shook, TAI’s Manager of Engineering Services. Jerry G. Rose, Principal Investigator of this project; and Vincent P. Drnevich, Chairman of Civil Engineering Department, University of Ken- tucky, for their assistance in this study. The dynamic moduli and other properties of the cored samples were tested in the TAI’s laboratory under the direction of Ben F, Kallas, Principal Engineer for Special Investigations. LITERATURE CITED 1. J. G. Rose, C. Lin, and V. P, Drnevich, “Hot Mix Asphalt for Railroad Trackbeds— Construction, Performance and Overview,” Proceedings of the Association of As- pbalt Paving Technologists, Vol. 53 (1984), 19-50. Y. H. Huang. C. Lin, X. Deng. and J. G. Rose. “KENTRACK. A Computer Pro- gram for Hot Mix Asphalt and Conventional Ballast Railway Trackbeds,” Research Report, The Asphalt Institute (to be published in 1984). S. Fateen, “A Finite Element Analysis of Full Depth Asphalt Railroad Track,” Master's Thesis, Department of Civil Engineering, University of Maryland, 1972. $. D. Tarabji and M, R. Thompson, “Finite Element Analysis of a Railway Track Support System, User’s Manual,” Report No. FRA-OR & D-76-257, U.S. Depart- ment of Transportation, 1976. C. S. Chang, C. W. Adegoke, and E, T, Selig, “GEOTRACK Model for Railrvad Track Performance,” Journal of the Geotechnical Engineering Division, American Society of Civil Engineers, Vol. 106, NO. GT11 (November 1980), 1201-1218. . J. Colville, “Design Criteria for Full-Depth Hot-Mix Asphalt Railway Roadbeds,” Report NO. QIP-99, National Asphalt Pavement Association, or NO, RR-78-2, The Asphalt Institute, 1978. Y.H. Huang, C. Lin, and J. G. Rose, “Asphalt Pavement Design: Highway Versus Railroad,” Journal of Transportation Engineering, American Society of Civil Engi- neers, Vol. 110, No. 2 (March 1984), 276-282. D. Hwang and M. W. Witczak, “Program DAMA (Chevron), User's Manual,” Depart- ment of Civil Engineeting, University of Maryland, 1979. J. G. Rose, Y. H. Huang, V. P. Drnevich, and C. Lin, “Hot Mix Asphalt Railroad Trackbed, Construction and Performance,” Research Report NO, QIP 104, National Asphalt Pavement Association, or RR-83-2, The Asphalt Institute, 1983. - 2 x = ° 494 HUANG, LIN AND DENG 10. TAL, “Research and Development of The Asphalt Institute's Thickness Design Manual (MS-1) Ninth Fdition,” Research Report NO. RR-82-2, The Asphalt Insti- ture, 1982. 11, J. F. Shook, F. N. Finn, M. W. Witezak, and C. L. Monismith, “Thickness Design of ‘Asphalt Pavements—The Asphalt Institute Method,” Proceedings of the Sth Interna- tional Conference on Structural Design of Asphalt Pavements, Vol. 1 (1982), 17-44. 12, A. I, M. Claussen, J. M. Edwards, P. Sommer, and P. Uge, “Asphalt Pavement De- sign—The Shell Method,” Proceedings of the 4th International Conference on the Structural Design of Asphalt Pavements, Vol. 1 (1977), 39-74. 13. L. KE. Santucci, “Thickness Design Procedure for Asphalt and kmulsified Asphalt Mixes,” Proceedings of the 4th International Conference on the Structural Design of Asphalt Pavements, Vol. 1, (1977), 424-456. 14. E. J. Yoder and M. W. Witezak, Principles of Pavement Design, 2nd Edition. John Wiley & Sons, Inc., 1975.

You might also like