ESTIMATING DRYING TIME FOR A STOCK PEANUT CURING DECISION SUPPORT SYSTEM

ESTIMATING DRYING TIME FOR A STOCK PEANUT CURING DECISION SUPPORT SYSTEM C. L. Butts, J. I. Davidson, Jr., M. C. Lamb, C. V. Kandala, J. M. Troeger ABSTRACT. A decision support system to manage commercial peanut drying facilities was developed. Empirical equations to predict peanut drying time were developed using PNUTDRY, a bulk peanut curing model. Input data for the empirical model includes daily maximum/minimum ambient temperature, maximum allowable plenum temperature, maximum allowable temperature rise, initial and final pod moisture content, and airflow rate. Using kernel moisture content instead of pod moisture content, as required in the original simulation model, had minimal impact on predicted drying time. Regression equations comparing actual drying times to estimated drying times had an R 2 of 0.72. When data from loads dried with ambient air were included, the R 2 decreased to 0.48. The equations presented have been incorporated into a software module to manage peanut curing facilities, released as FarmSuite by the Peanut Foundation. Keywords. Curing, Decision support system, Drying, Expert systems, Models, Peanuts. Peanuts are usually harvested from mid August to late October depending on latitude and planting date. Peanuts must have an average kernel moisture content below 10.5% before their value can be determined and sold (all moisture contents presented are wet basis of peanut kernels unless otherwise noted). Most peanuts in the U.S. are harvested at kernel moisture contents above 10.5% and mechanically dried or cured. Curing occurs after peanuts are threshed and placed in wagons or trailers equipped with a plenum in the bottom. The loaded containers are transported to a facility where a fan unit with a heater is connected to an opening in the wagon using a flexible duct. A properly sized fan delivers 8 to 12 m 3 min 1 m 3, which is heated using propane or natural gas. Most published guidelines include recommendations to heat the air 8 C above ambient, but no higher than 35 C (Cundiff et al., 1991; Planters, 1990; Samples, 1984; Young et al., 1982). However, many of the curing systems currently in use have the capacity to heat the air 20 C above ambient and are equipped with thermostats that will allow a maximum plenum temperature of approximately 52 C. Most on farm drying systems are capable of handling 4 to 8 trailers and are usually operated by the grower or one other individual. Commercial drying facilities typically handle up to 200 trailers simultaneously. A commercial facility is Article was submitted for review in November 2003; approved for publication by the Information & Electrical Technologies Division of ASAE in March 2004. Presented at the 2003 ASAE Annual Meeting as Paper No. 036047. The authors are Christopher L. Butts, ASAE Member, Agricultural Engineer, James I. Davidson, Jr., Mechanical Engineer (Retired), Marshall C. Lamb, Agricultural Economist, Chari V. Kandala, ASAE Member, Agricultural Engineer, and John M. Troeger, Agricultural Engineer (Retired), USDA ARS National Peanut Research Laboratory, Dawson, Georgia. Corresponding author: Christopher L. Butts, USDA ARS National Peanut Research Laboratory, P.O. Box 509, Dawson, GA 39842; phone: 229 995 7431; fax: 229 995 7416; e mail: cbutts@ nprl.usda.gov. usually operated by several employees under the supervision of a manager, all operating on 8 to 12 h shifts. According to the National Peanut Buying Point Association (personal communication, Oct. 2002), there are approximately 400 peanut buying points across the peanut production regions of the U.S., most of which have commercial drying facilities. Buying points typically receive 85% to 90% of their peanuts within a 6 week period, during which time they must properly dry the peanuts to a kernel moisture content between 7% and 10%. Both the on farm and commercial drying facilities must be managed effectively by periodically obtaining samples from each wagon on the dryer, shelling those samples, and determine the moisture content of the peanut kernels to prevent overdrying. Operators use various rules of thumb to estimate drying time and to schedule sampling. The on farm dryer may schedule sampling late at night immediately before returning home or very early in the morning before starting other farm operations. He then must either adjust the plenum temperature so that the drying will not be completed before the next sampling time, or he must remember to return to repeat the measurement in the middle of the day. The commercial facility also samples peanuts from each dryer, but it must have an effective system of recording and tracking the progress of each load of peanuts. The approximate sample time and the kernel moisture content are recorded on an individual ticket or in a log book. Many facilities place the ticket on a large process board to maintain a visual account of the status (occupied or unoccupied) of each dryer and the progress of drying. This information must be effectively communicated to the next shift of employees during the 24 h a day operation. Occasionally, due to poor communication, poor timing of sampling, poor estimates of drying time, or human error, a wagon load of peanuts is left on the dryer too long or is left waiting to be put on a dryer, causing irreparable damage to the peanuts. Drying peanuts below 9% to 10% costs as much as $6.60/t in additional fuel and labor costs and reduces peanut value by as much as $32/t (Butts, 1995). Transactions of the ASAE Vol. 47(3): 925 932 2004 American Society of Agricultural Engineers ISSN 0001 2351 925

OBJECTIVES A system to monitor commercial and on farm peanut drying facilities was developed that provides a visual representation of all wagons of peanuts currently on active peanut dryers, estimates of peanut drying time and current moisture content, and consistent sampling schedules. This information would help facility managers minimize queues for dryer space and maximize effectiveness of the labor. The objective of this research was to develop an empirical method for calculating peanut drying time and compare that to observed drying time. MODEL DEVELOPMENT At least two computer simulation models have been developed and tested over the years to estimate the time required to dry farmer stock peanuts and the potential impact on peanut quality (Troeger, 1989; Colson and Young, 1990). Both models require similar input data (Butts et al., 1989), such as ambient air temperature, plenum air temperature, airflow rate, initial peanut moisture, and peanut volume. The primary difference between the models is the form of the input data. PNUTDRY (Troeger, 1989) requires ambient temperature data in the form of daily maximum and minimum temperatures (T DMX and T DMN, respectively). Wet bulb and dry bulb temperatures are then calculated on a time scale to match the time step of the model. PEADRY (Colson and Young, 1990) requires wet and dry bulb temperatures of the air after heating. PNUTDRY calculates plenum temperature based on the simulated ambient temperature, the user specified temperature rise (T PRS ), and the maximum allowable plenum temperature (T PMX ). The other major difference between the models is that PNUTDRY uses whole pod moisture, while PEADRY uses moisture content of the hulls and kernels as separate components. Both models accurately simulate peanut drying times, but considerable programming would be required to incorporate either of these models, and the response time would be unacceptable in a real time decision support system. Butts and Hall (1994) developed a decision support system that requires only initial peanut moisture as input data. A quadratic equation developed using data from a single drying season was used to estimate total peanut drying time with no regard to plenum temperature or ambient conditions. Comparisons of actual and predicted total drying time indicated errors as high as 200%. Troeger (1995, unpublished data) used the PNUTDRY model to develop empirical equations that could be used to predict total drying time, auxiliary energy requirements, and the percent split kernels. The model was first run using a standard airflow rate of 20 m 3 min 1 m 3 and a standard initial pod moisture content of 20% over a range of ambient temperature conditions that normally occur during the fall peanut harvest. Simulations were run covering a range of dryer operating parameters such as the maximum plenum temperature and the temperature rise above ambient (see table 1 for variable definitions). The drying times obtained from these simulations at standard moisture and airflow were tabulated, and a regression was developed to estimate the reciprocal of the time required to reduce the pod moisture content to 7%, as shown in equation 1: 2 RTIME = A0 + A1 TPRS + A2 DTPR + A3 DTPR (1) The coefficients used in the equation for RTIME depend on T DMN and the range in daily temperature (T DRN ), as shown in table 2. Similarly, the simulations were run using non standard initial moisture contents and non standard airflow rates. Regression equations of the form shown in equation 2 were developed to estimate the natural logarithm (LNRTIM) of the ratio of drying time at non standard conditions to that at standard conditions (t rel ): 2 LNRTIM = B0 + B1 RSMC + B2 RSMC + B3 SMSF (2) The coefficients used in the equation to calculate LNRTIM vary with T DMN and the allowable rise in temperature above ambient (T PRS ) and are tabulated in table 3. Table 1. Variable names, definitions, and dimensions of variables used in the decision support system for curing peanuts. Variable Definition DTPR = (T PMX T DMN ) T PRS, difference in maximum rise needed to achieve T PMX and T PRS ( C) FMC k Desired final moisture content of peanut kernels (% wet basis) FMC p Desired final moisture content of whole peanut pods (% wet basis) IMC k Initial moisture content of peanut kernels (% wet basis) IMC p Initial moisture content of whole peanut pods (% wet basis) LNRTIM Natural logarithm of the relative time ratio (t rel ) MC k Moisture content of peanut kernels (% wet basis) MC p Moisture content of whole peanut pods (% wet basis) RSMC = 20/MC p, standard moisture content ratio (dimensionless) SMSF = (MC p /20) (FLOW/20), standard moisture, standard airflow ratio (dimensionless) RTIME = 1/t ref t Time required to dry peanut pods from IMC p to the desired final moisture content (h) t i Time required to dry peanut pods from IMC p to the desired final moisture content (h) t f Time required to dry peanut pods from FMC p to 7% w.b. final pod moisture content (h) t ref Time required (h) to dry peanut pods from standard MC (20%) to 7% w.b. using a standard airflow (20 m 3 min 1 m 3 ) t rel Ratio of actual time to dry peanut pods from a given peanut pod moisture content to 7% w.b. to t ref T DMN Minimum daily temperature of the ambient air ( C) T DMX Maximum daily temperature of the ambient air ( C) T DRN = T DMX T DMN, daily temperature range ( C) T PMX Maximum allowable plenum temperature of the drying air ( C) 926 TRANSACTIONS OF THE ASAE

Table 2. Coefficients used in equation 1 to calculate the reciprocal of the reference time (RTIME = A 0 + A 1 T PRS + A 2 DTPR + A 3 DTPR 2 ) as a function of the minimum daily temperature (T DMN ) and the range of daily temperature (T DRN ) (source: Troeger, 1995, unpublished data). T DMN T DRN A 0 A 1 A 2 A 3 4 4 1.852156E 03 9.953614E 04 8.911448E 04 1.072162E 04 4 8 1.317054E 03 9.672455E 04 7.916016E 04 4.94704E 05 4 12 4.241914E 04 9.180098E 04 7.377416E 04 3.36839E 05 4 16 1.413567E 03 8.19017E 04 6.203818E 04 2.41878E 05 4 20 3.589781E 03 7.074627E 04 5.192666E 04 1.85408E 05 8 4 1.553843E 03 1.196871E 03 1.139335E 03 1.373213E 04 8 8 1.317179E 03 1.184897E 03 1.038736E 03 6.69513E 05 8 12 4.425663E 04 1.138148E 03 9.502175E 04 4.33753E 05 8 16 1.713295E 03 1.02154E 03 8.623724E 04 3.8464E 05 8 20 4.174396E 03 8.97215E 04 7.874638E 04 3.82802E 05 12 4 3.763067E 04 1.373741E 03 1.30281E 03 1.582728E 04 12 8 4.419897E 04 1.379923E 03 1.218081E 03 7.79914E 05 12 12 5.985383E 04 1.325167E 03 1.033598E 03 4.00115E 05 12 16 2.995638E 03 1.196853E 03 8.824946E 04 2.64732E 05 12 20 6.18092E 03 1.03012E 03 8.408308E 04 3.18336E 05 16 4 7.660201E 04 1.598182E 03 1.559735E 03 1.924761E 04 16 8 7.740266E 04 1.593390E 03 1.48485E 03 9.82715E 05 16 12 1.514611E 03 1.561234E 03 1.33715E 03 6.31933E 05 16 16 3.902024E 03 1.435324E 03 1.225658E 03 5.66537E 05 16 20 7.543677E 03 1.246509E 03 1.107718E 03 5.68876E 05 20 4 2.333728E 03 1.839484E 03 1.746028E 03 1.925707E 04 20 8 1.480240E 03 1.885144E 03 1.847576E 03 1.221702E 04 20 12 2.254566E 03 1.846800E 03 1.745967E 03 8.88771E 05 20 16 4.897435E 03 1.710860E 03 1.491594E 03 5.84087E 05 24 4 3.039418E 03 2.244047E 03 1.904825E 03 1.855586E 04 24 8 2.264362E 03 2.260189E 03 2.305136E 03 1.523335E 04 24 12 4.218008E 03 2.120024E 03 1.994594E 03 8.27993E 05 Table 3. Coefficients used in equation 2 to calculate the natural logarithm of the relative time (LNRTIM = B 0 + B 1 RSMC 2 + B 2 RSMC + B 3 SMSF) as affected by the minimum daily temperature (T DMN ) and the maximum allowable temperature rise above ambient (T PRS ) (source: Troeger, 1995, unpublished data). T DMN T PRS B 0 B 1 B 2 B 3 4 8 0.5892391 1.044269 0.5054639 5.043397E 02 4 12 0.6724246 1.021435 0.3900988 4.108837E 02 4 16 0.6492546 1.098641 0.4809381 3.155203E 02 4 20 0.7134042 1.053481 0.3708081 3.073103E 02 8 8 0.2598 1.117875 0.9266042 0.068529 8 12 0.3673521 1.152978 0.8413474 5.572158E 02 8 16 0.4322834 1.160488 0.7774493 4.924418E 02 8 20 0.4583919 1.142093 0.7260628 4.236175E 02 12 8 8.307089E 02 1.119702 1.126191 8.955975E 02 12 12 0.2026304 1.20579 1.081945 7.878532E 02 12 16 0.2662691 1.208358 1.01971 7.762131E 02 12 20 0.3279044 1.161435 0.9086008 7.506986E 02 16 8 8.371187E 02 1.065448 1.118189 0.1364531 16 12 0.1121507 1.230387 1.241069 0.1228331 16 16 0.2336255 1.171985 1.052741 0.1143812 16 20 0.2773098 1.129945.9527559 0.1001203 20 8 0.2676973 0.9751374 0.9265146 0.2190745 20 12 0.317797 1.090478 0.9679775 0.1952968 20 16 0.3226129 1.101325 0.9558112.01770995 20 20 0.3224531 1.069547 0.8963723 0.1492784 24 8 0.7998526 0.704495 0.2393687 0.3347263 24 12 0.6229932 0.9213273 0.5926845 0.2943504 24 16 0.5485932 0.9326289 0.6262487 0.242213 To calculate the time required to dry from the desired moisture to 7%, RTIME must be calculated using the appropriate values of A 0, A 1, A 2, and A 3. The time required to dry peanuts at reference conditions is calculated using equation 3: 1 t ref = (3) RTIME Vol. 47(3): 925 932 927

Figure 1. Flowchart for calculating peanut drying time. Next, LNRTIM is calculated using the appropriate values of B 0, B 1, B 2, and B 3 based on values of T DMN and T PRS. Based on the definition LNRTIM, t rel is calculated using equation 4: LNRTIM t rel = e (4) The time required to dry peanuts from the stated initial moisture content to 7% is then calculated by: t i = tref trel (5) The process is repeated using the desired final pod moisture content to determine the time required to dry from the final moisture content to 7% (t f ). The time required to dry peanuts from the initial to final pod moisture can then be determined by equation 6: t = t i t f (6) The selection of coefficients to calculate LNRTIM and RTIME must be based on the values of T DMN and T DRN and on the values of T DMN and T PRS, respectively. To accommo- 928 TRANSACTIONS OF THE ASAE

date actual conditions experienced during the peanut harvest season, t rel and t ref were calculated four different times using the tabulated coefficients. Linear interpolation was then used to calculate the total drying time (fig. 1). An example follows for calculating the time required to dry peanuts from an initial pod moisture (IMC p ) of 18% to a final pod moisture (FMC p ) of 10% using an airflow of 12 m 3 min 1 m 3. The ambient temperature ranges from 15 C daily minimum (T DMN ) to 29 C daily maximum (T DMX ). The thermostat for the dryer limits the plenum temperature (T PMX ) to 35 C, while the burner has a capacity to heat the air only 8 C above ambient (T PRS ). The first step is to calculate the temperature parameters required to determine the coefficients of equations 1 and 2. The first is range in daily ambient temperatures (T DRN ) and is defined in table 1: T DRN = T DMX T DMN = 14 C (7) The variable DTPR represents the difference in the maximum temperature rise required to achieve the desired plenum temperature (T PMX ) and the temperature rise that can be generated by the dryer s burner (T PRS ). It is a measure of the dryer s ability to reach the maximum allowable plenum temperature. For this example: DTPR = (T PMX T DMN ) T PRS = 12 C (8) Since the coefficients (A 0, A 1, A 2, and A 3 ) used to calculate RTIME depend on the value of T DMN and T DRN, the value of RTIME must be determined by interpolation. Calculate RTIME using equation 1 at the tabulated values of T DMN and T DRN that are closest to the desired values. In this example, the values of RTIME when T PRS is 8 and DTPR is 12 are: T DMN = 12, T DRN = 12; RTIME 1 = 0.017841 T DMN = 12, T DRN = 16: RTIME 2 = 0.019348 T DMN = 16, T DRN = 12; RTIME 3 = 0.020950 T DMN = 16, T DRN = 16; RTIME 4 = 0.021934 Linear interpolation is then used to calculate the value of RTIME for T DMN = 15 and T DRN = 14 by the following set of equations: T 16 RTIME RTIME ( RTIME RTIME ) DRN 5 = 2 + 1 2 (9) 12 16 T 16 RTIME RTIME ( RTIME RTIME ) DRN 6 = 4 + 3 4 (10) 12 16 RTIME = RTIME 6 + = 0.02073 ( RTIME RTIME ) 5 TDMN 16 6 12 16 (11) Using the definition of RTIME as the reciprocal of the time required to dry peanuts at reference conditions, calculate t ref : 1 t ref = = 48.2 h (12) RTIME The relative time required to dry from 18% to the standard cutoff moisture of 7% and from the desired final moisture of 10% to the standard cutoff moisture is similarly calculated. Since the value of T PRS is one of the values tabulated for the coefficients (B 0, B 1, B 2, and B 3 ), LNRTIME will only have to be calculated twice each for the initial and final moisture content. The calculations for the relative time for the initial moisture are shown below. The values of RSMC and SMSF must first be calculated: 20 20 RSMC = = = 1.11 (13) IMC p 18 IMC p FLOW SMSF = = 0.54 (14) 20 20 Using the values of RSMC and SMSF calculated above, and the coefficient values from table 3, the values of LNRTIME are: T DMN = 12, T PRS = 8; LNRTIME = 0.09632 T DMN = 16, T PRS = 8; LNRTIME = 0.06291 Interpolating between the two values to obtain the value of LNRTIME corresponding to T DMN = 15 and T PRS = 8 results in LNRTIME = 0.07126. Substitution into equations 4 and 5 yields the time required to dry peanuts from the initial pod moisture content to 7%: t i = 44.9 h. Similar calculations for the final pod moisture results in the time required to dry from the final pod moisture to 7%: t f = 6.3 h. Substitution of the values of t i and t f into equation 6 results in the time required to dry peanuts from the initial to the final moisture content: t = 38.6 h. MODEL VALIDATION Data collected during the 1988, 1989, and 1992 harvests were used to validate the estimates of total drying time. Temperature and dryer performance data collected in previous studies were used to compare actual drying times to predicted drying times. Dryer performance data included initial and final kernel moisture contents and dryer start and stop times. Temperature data included ambient and plenum temperature monitored on 60 s intervals and recorded on 15 min intervals during the studies. A microprocessor was used to control plenum temperature during these studies. Kernel moisture contents were determined during all three years using the oven method described in ASAE Standard S410.1 (ASAE Standards, 1997). Hull moistures were determined only during the 1988 harvest. Pod moisture content was calculated during the 1988 study using a mass weighted average of the hull and kernel moisture content. Total drying time was calculated using initial and final kernel and pod moisture contents, and then compared to actual drying time using regression and analysis of variance. RESULTS AND DISCUSSION Drying data were collected for 106 loads during the 1988, 1989, and 1992 harvests and are summarized in table 4. Initial kernel moisture ranged from 9.3% to 35.7%, while the final kernel moisture ranged from 5.4% to 13.2% during these tests. Weather data covered a range of conditions normally Vol. 47(3): 925 932 929

Table 4. Summary of peanut curing parameters used as input data to validate model to estimate peanut drying time. Year Parameter Average Standard Deviation Maximum Minimum 1988 Initial kernel moisture (IMC k, %) 19.4 6.1 32.8 11.5 (n = 28) Final kernel moisture (FMC k,%) 9.9 0.5 10.7 8.6 Daily minimum temperature (T DMN, C) 8.7 4.6 16.8 3.2 Daily maximum temperature (T DMX, C) 27.8 2.6 30.9 24.3 Maximum plenum temperature (T PMX, C) 27.9 2.7 30.9 24.3 Drying time (t, h) 24.7 14.1 51.5 3.0 1989 Initial kernel moisture (IMC k, %) 18.2 5.9 28.3 11.8 (n = 40) Final kernel moisture (FMC k,%) 10.8 1.2 13.2 6.8 Daily minimum temperature (T DMN, C) 21.5 2.8 23.9 14.1 Daily maximum temperature (T DMX, C) 31.0 2.4 34.3 27.2 Maximum plenum temperature (T PMX, C) 34.6 2.4 37.2 28.8 Drying time (t, h) 20.3 14.9 51.3 4.0 1992 Initial kernel moisture (IMC k, %) 17.2 7.0 35.7 9.3 (n = 38) Final kernel moisture (FMC k,%) 9.8 1.0 10.6 5.4 Daily minimum temperature (T DMN, C) 15.7 5.5 24.0 3.1 Daily maximum temperature (T DMX, C) 27.2 3.6 34.4 17.2 Maximum plenum temperature (T PMX, C) 36.0 2.4 37.7 26.4 Drying time (t, h) 19.5 10.8 38.3 3.0 3 year Initial kernel moisture (IMC k, %) 18.2 6.4 35.7 9.3 average Final kernel moisture ((FMC k,%) 10.2 1.1 13.2 5.4 (n = 106) Daily minimum temperature (T DMN, C) 16.3 6.6 24.0 3.1 Daily maximum temperature (T DMX, C) 28.8 3.4 34.4 17.2 Maximum plenum temperature (T PMX, C) 33.5 4.0 37.7 24.3 Drying time (t, h) 21.1 13.4 51.5 3.0 encountered during peanut harvest. Daily minimum temperature ranged from 3.1 C to 24.0 C, while the daily maximum ranged from 17.2 C to 34.4 C. The maximum plenum temperature represents the thermostat setting of the dryer and averaged 33.6 C during the 106 loads cured during the 3 year study. Depending on initial and final moisture contents and dryer operation, drying times averaged 21 h and ranged from 3 to 52 h. Since pod moisture content was not available for all loads used in this study, previously unpublished data (Butts, 1988, 1990, 1991) in which kernel and hull moisture contents were determined throughout the drying process were used to generate a relationship between pod and kernel moisture content. A linear regression (fig. 2) with an R 2 of 0.988 was developed to estimate pod moisture content given the kernel moisture as follows: MC p = 0.8929 MC k + 1.2623 (15) This equation was used to estimate pod moisture from the kernel moisture when no pod moisture data was available. Using the equations and procedure previously described, the drying time was estimated using the temperature and moisture data for each load. Drying times were estimated using the kernel and pod moisture contents. Linear regression was used to compare actual drying time to the predicted drying time (figs. 3 and 4). Regression results are shown in table 5. Regression analysis using data from all three years and the kernel moisture content had an R 2 of 0.484 with a slope near unity and an intercept of 3.67 h (fig. 3). The tests conducted during the 1988 harvest resulted in an average temperature rise of less than 8 C, which is outside the range of data used to develop the original prediction equations. Regression analysis using only data from 1989 and 1992 resulted in an R 2 of 0.726. The intercept was less than 1 h (fig. 4). Further regression analysis indicated that a slight improvement in the prediction of drying time was achieved if Pod Moisture Content (% w.b.) 50 40 30 20 10 Observed Data MC p = 0.8929*MC k + 1.2623 R 2 =0.988 99% Confidence Interval 0 0 10 20 30 40 50 Kernel Moisture Content (% w.b.) Figure 2. Relationship between moisture content of whole peanut pods (MC p ) and peanut kernels (MC k ) while curing. the pod moisture content was used in the model. The R 2 increased by about 0.015 when the pod moisture was used, compared to using the kernel moisture to predict drying time (table 5). This was true regardless of whether or not the data from 1988 were included in the analysis. Analysis of variance showed that drying time predicted using the pod moisture content was not significantly different from the actual drying time or the time predicted using the kernel moisture content when all three years were used in the analysis (table 6). However, the actual drying time was significantly lower than the drying time predicted using the kernel moisture content. If the 1988 data were excluded from the analysis, then there were no significant differences among the mean drying times. Commercial drying facilities use at least an 8 C temperature rise, with most not controlling the temperature rise at all. 930 TRANSACTIONS OF THE ASAE

Predicted Drying Time (h) 100 80 60 40 20 0 0 10 20 30 40 50 60 Actual Drying Time (h) t pred = 3.67 + 1.1526*t actual R 2 =0.484 Figure 3. Linear regression of actual drying time compared to predicted drying time using kernel moisture content and data from 1988, 1989, and 1992. The temperature rise in commercial drying facilities is most often limited by the original dryer specifications. Most facilities use the thermostat to maintain a relatively constant plenum temperature. Therefore, the data from 1989 and 1992 are more representative of actual conditions at a commercial peanut drying facility. A second mitigating factor for excluding the 1988 data from the analysis is the fact that the dryers did not operate reliably. Numerous flame and power failures occurred during the collection of the data, making it difficult to determine a representative initial moisture and an accurate drying time. These analyses indicate that reasonable estimates of drying time can be achieved if kernel moisture content is used as an input to the model, with only a slight increase in prediction accuracy when the estimated pod moisture content is used. A custom interface that looks like a tabular notebook was developed to present data and drying time estimates for each load (fig. 5). As the load of peanuts arrives at the drying facility, a sample is retrieved, shelled, and the moisture content determined using an electronic moisture meter. The Predicted Drying Time (h) 100 80 60 40 20 t pred = 0.11 + 1.0829*t actual R 2 =0.726 0 0 10 20 30 40 50 60 Actual Drying Time (h) Figure 4. Linear regression of actual drying time compared to predicted drying time using kernel moisture and data from 1989 and 1992. MC Used in Model Table 5. Results of linear regression comparing predicted (t pred ) and actual (t actual )drying times. Crop Years Used in Analysis Regression, t pred = R 2 Kernel 1988, 1989, 1992 3.667 + 1.153 t actual 0.484 1989, 1992 0.115 + 1.083 t actual 0.726 Pod 1988, 1989, 1992 3.077 + 1.109 t actual 0.499 1989, 1992 0.504 + 1.051 t actual 0.741 operator selects the Add feature on the screen and enters owner information and initial moisture content. The software uses previously entered operating data such as ambient temperature (T DMX, T DMN ), plenum temperature (T PMX ), temperature rise (T PRS ), and preferred cutoff moisture content (FMC p ) and the individual load data to calculate drying time, time on the dryer, time remaining, and the next scheduled sampling time. Sampling times are scheduled based on the last measured moisture content and range from hourly to 12 h. Once the time of day reaches the desired sampling time, the status of the load changes from Okay to Check. If the time of day is greater than the predicted end time, or if the estimated moisture content decreases below the desired final moisture content, then the status changes to Remove. The variables of time on dryer, remaining time, estimated moisture content, and status are recalculated and displayed on a 15 min interval. The moisture content is estimated by subtracting the change in moisture (Butts and Hall, 1994) from the last measured moisture content: 2 MC = 0.0022t elapsed + 0.2382t elapsed + 0.2497MCk 3.4658 (16) where MC k is the most recently measured moisture content, and t elapsed is the time (h) since the last measurement was made. After the status of the load changes to Check, the operator retrieves, shells, and determines the moisture of the peanuts. The new moisture is updated, a new drying time is calculated using equations 1 through 6, and a new sampling time is scheduled. When the status of the load changes to Remove, the operator should obtain another moisture sample to verify that the load has reached its desired level and then use the Remove feature of the program to enter the final moisture content and delete the load from the system. Each time a load of peanuts is added, updated, or removed from the system, an entry is recorded into a log file. The log file contains the load identity, the dryer information, and the load s progress throughout the curing process. The log file can be imported into a database for analysis and improvements in managing the system. [a] Table 6. Comparison of predicted and actual drying times using kernel and pod moisture content as input. Average Drying Time (h) [a] Drying Time Prediction All Data (1988, 1989, 1992) Excluding 1988 Actual 21.2 a 19.9 a Kernel MC 27.3 b 21.4 a Pod MC 25.8 ab 20.1 a Means in the same column followed by the same letter are not significantly different (P > 0.05). Vol. 47(3): 925 932 931

Figure 5. Example display screen of PECMAN, the decision support system for managing a peanut curing facility. CONCLUSION A system of equations was developed to predict the time required to dry peanuts in commercial peanut drying operations. Coefficients of the equations varied with temperature parameters including daily minimum temperature, daily temperature range, and the temperature rise imparted to the drying air by the dryer heating unit. Using data recorded during three different drying seasons, drying times were predicted by this system of equations and compared to drying times actually recorded. The predicted drying time correlated very well with the actual drying time, as indicated by R 2 values as high as 0.73 when temperature conditions were within the range used in the original model development. Kernel moisture content could be used to predict drying time reasonably well, even though the original model was developed using the moisture content of the whole pod. These equations have been incorporated into a revised decision support system for managing peanut curing systems called PECMAN and released as part of a larger peanut production decision support system, called FarmSuite, by the Peanut Foundation. REFERENCES ASAE Standards. 1997. S410.1: Moisture measurement Peanuts. St. Joseph, Mich.: ASAE. Butts, C. L. 1995. Incremental cost of overdrying farmers stock peanuts. Applied Eng. in Agric. 11(5): 671 675. Butts, C. L., and M. J. Hall, III. 1994. Development of a graphical decision support system for a peanut curing operation. ASAE Paper No. 946515. St. Joseph, Mich.: ASAE. Butts, C. L., J. M. Troeger, and J. H. Young. 1989. Comparison of two peanut bulk curing models. ASAE Paper No. 896599. St. Joseph, Mich.: ASAE. Colson, K. H., and J. H. Young. 1990. Two component thin layer drying model for unshelled peanuts. Trans. ASAE 33(1): 241 246. Cundiff, J. S., K. D. Baker, F. S. Wright, and D. H. Vaughan. 1991. Curing quality peanuts in Virginia. Virginia Cooperative Extension Service Publication 442 062. Blacksburg, Va.: Virginia Tech. Planters. 1990. A grower s guide to quality. Winston Salem, N.C.: Planters Lifesavers Co. Samples, L. E. 1984. A curing guide for Georgia peanut growers. University of Georgia Cooperative Extension Service Leaflet 355. Athens, Ga.: University of Georgia, College of Agriculture. Troeger, J. M. 1989. Modeling quality in bulk peanut curing. Peanut Science 16(2): 105 108. Young, J. H., N. K. Person, J. O. Donald, and W. D. Mayfield. 1982. Harvesting, curing, and energy utilization. In Peanut Science and Technology, 458 485. H. E. Pattee and C. T. Young, eds. Yoakum, Texas: American Peanut Research and Education Society. 932 TRANSACTIONS OF THE ASAE