GEK Wiki / Multi-fuel Run Comparison
  • If you are citizen of an European Union member nation, you may not use this service unless you are at least 16 years old.

  • Stop wasting time looking for files and revisions. Connect your Gmail, DriveDropbox, and Slack accounts and in less than 2 minutes, Dokkio will automatically organize all your file attachments. Learn more and claim your free account.


Multi-fuel Run Comparison

Page history last edited by bk 10 years, 4 months ago

Discussion and further analysis in the forum



P - stands for pressure (generally shown as inH2O, although internally handled as Pascals)

T - stands for temperature (in °C)

Q - stands for volumetric flow rate (in m3/hr, cubic meters per hour)

R - stands for resistance (currently reported in unusual units)


Instrumentation and Equipment

Pyrolysis and combustion assembly

Automatic grate rotation

PID control of P_reactor via ejector air


Fuel level detection




Experimental Design and Run Procedure

Cold Test

At the start of each run, a "cold test" was conducted: all reactor plumbing was set up as it would be for normal operation, and the GCU set the reactor vacuum in a downward ramp pattern in 30 second intervals to provide information on flow behavior when not lit.


Start Up and Warm Up

Following this, the reactor vacuum was set to 4 inH2O and the reactor was started through the V3 ignition port with a propane torch. The flare was ammended with propane over this phase to ensure complete combustion of the gases. The port was left open for a few minutes to insure that combustion established, and then was closed. The reactor was then allowed to approach thermal equilibrium at this same vacuum over approximately 40 minutes (as determined by the temperature of incoming air at the air nozzles (T_air_in).


Diagram of reactor vacuum experimental design time series (increasing vacuum upward)



After reaching thermal equilibrium, the reactor vacuum setpoint was changed based on the above pattern in 5 minute increments.

The reactor was kept at 4 inH2O, dropping to 2, 1, and 1/4 inH2O. The reactor vacuum was kept at 1/4 inH2O for 10 minutes, the first of two cycles were started, ramping up to 8 inH2O at 5 min/setpoint then falling at 10 min/setpoint. The use of two cycles was intended to test the repeatability of reactor conditions for different vacuums.


Shut Down

At the end of the run, the reactor was allowed to cool down, with a minimum flow to sustain the flare to combust any produced pyrolysis gases, and then capped after pyrolysis had stopped.


Run Descriptions

Wood Pellet Run

The vertical round plate fuel sense "plunger" worked well with pellets and required no human intervention. With other fuels, the plunger is suspected to form "plugs" of fuel undercompression which don't fall down under gravity and reset the microswitch.

The temperatures seen over the entire wood pellet run were very low.

The ejector was not able to achieve the full pull at the 8 inH2O setpoint on the first cycle due to the premix air valve being fully open.


Softwood Chip Run

After binding issues with augering softwood chips through the long auger, the short "drying bucket" auger was attached (but not connected to the gas flow for heat exchange). 


Walnut Shell Run

Startup during the walnut shell run was interrupted by GCU resets caused by interference from the electromechanical darkroom clock used for timing the tar sampling pump. 


Fuel Properties


Bulk Density


Void Space


Moisture Content

[%, dry basis]

Angle of Repose


Wood Pellets 656 42 8.9, 5.9 after 160 min  30  
Softwood Chips 195 53 12.4 40  
Walnut Shells 235 73 10.3 35  



Tar Measurement and Correlation with Reactor Conditions


Spot samples were drawn at the reactor output, BEFORE the filter train. The spot size was 0.375" dia. The sample was drawn by a small sampling pump (KPV-20A) run for 15 seconds (@ 5V) to pull 450 mL of of gas. The gas sample was drawn from the bottom filter port, through 16" of 1/8" neoprene tubing and through the filter media (ceramic wool insulation strip) sandwiched between two metal plates clamped down with wingnuts (shown above). We expect the woodgas to be much cleaner after the filter, though we are not characterizing the filter media during this run, sampling after the filter would add an additional variable that would be hard to control for (i.e. replacing filter media before each run and accounting for any change in filter effectiveness over the run).


Filter strips from the walnut shell run ("exit" side up). In sets of ten, increasing right to left, top down.

Filter strips after spot removal with Xacto knife, grayscale chart for manual interpretation, 3 mL syringe.

Set of ten vials containing a removed spot sample and 3 mL isopropyl alcohol.



Series of 4 mL vials filled with the tar/isopropyl solution from the three runs. Each vial is aligned to the corresponding time and P_reactor setpoint shown at the top of the image. Samples were generally taken at the mid-point of each setpoint, with some delays due to human error (which are accounted for in the reactor condition correlation plots below). Top to bottom: wood pellets, softwood chips, walnut shells.


The tar concentration in the highest tar sample is 40 times that of the lowest tar sample in the series above.


Correlations Between Reactor Conditions and Tar Production


A improved method for determining tar content was developed. An absorbtion photometer was developed by Al Shinn using two blue LEDs (an emitter and detector) passing a beam across a vial holder. The measurements from device where shown to follow the Beer-Lambert Law via dilution of an extra sample, implying that measurements of concentration can be derived from the measured light absorbtion.

This measurement technique is much better than the previous (quick and dirty) method and does not fall off with darker samples.

Details about this methodology will be expanded upon.


The former quantification method derived a value for each vial based on the saturation from the HSV color representation of the color at the center of the vial. This value dropped off for the darker samples.


Lines are a linear regression to the data. Linear regression may not be valid or significant, but is supplied to aid in seeing trends.


Discussion to follow.


2009/10/01 8:00 PM: Graphs updated (fixed flaw in T_tred, T_comb data, removed 2 outliers in Walnut Shell dataset caused by TC measurement issue).

2009/11/24: Updated graphs with r2 values. Graphs also include data from a wood cube run (to be posted shortly).




Accounting for Flowmeter Resistance

The orifice flowmeter used for measuring air in produces a non-neglible pressure drop before the air cowling, reaching up to about 1.6 inH20 at the maximum flowrates for these tests. In order to better represent what behavior a user would see when monitoring the reactor manometer (P_reactor), this should be compensated. Although available estimates are limited, a figure of .75 of the orifice differental pressure remains downstream. The compensated value for P_reactor without a flowmeter (i.e. normal setup) is assumed to be: P_reactor_no_fm = P_reactor - P_air_in*.75.


Volumetric Air Flow at Given Reactor Vacuums

The volumetric flow rate of air entering the well sealed reactor when running is given by the following equation:





Q_air_in [m3/hr] +/- 1


1 2
2 3.2
3 4.2
4 5.1
5 5.9
6 6.6
7 7.4
8 8.1
9* 8.7
10* 9.4

* - extrapolated

Table of predicted flowrates given reactor pull using the equation above.


With this information, estimates of fuel consumption and power output at given reactor pull can be derived.

A graph of the data and fitted curves for the three runs while the reactor was hot.


There is much more variablity in flow when the reactor is running, but across fuels, the fits are quite consistent.


An estimate of the volumetric flow rate of air entering the well sealed reactor when cold is given by the following equation:



(roughly twice the air enters the reactor at a given vacuum when cold. rather close to the additional volume of gas created by the biomass in the gasifier (gas/air ratio is figured to be ~2:1) when hot.)


Reduction Resistance

Reduction resistance can be calculated by dividing the pressure difference across the reduction zone by the gas flow through the reactor:Formula

The three runs that were conducted rotated the grate at a constant interval. This leads to a "stair stepped" pattern in the resistance. The increase in resistance over the intervals between grate shaking can be explained by processes such as accumulation of ash and settling and filling of void space by char or ash.

R_reduction [currently in units of Pa-hr/m3] from the first experimental cycle of the wood pellet run (smoothed by 5 sample moving average)


Another method to determine the amount of packing of the reduction zone is to look at the ratio of P_comb/P_reactor, this metric does not require a flowmeter:


Temperatures versus P_reactor

Temperatures inside the reactor are dependent on how hard it is being pulled. With increased pull rate, more air feeds combustion, generally leading to higher temperatures throughout the reactor. The data below consists of the median temperatures for each seperate 5 minute component of the main two cycles (which define the reactor pull setpoint).


The cycles are plotted with seperate symbols. The second cycle is often hotter than the first, indicating that the reactor is still approaching thermal equilibrium.

Note an interesting trend with the softwood chips. At high pull rates, both combustion and top of reduction decrease, while bottom of reduction continues upward. This suggests that combustion reactions may have been pulled downward towards the bottom of reduction.

Changes in reactor vacuum lead to changes of 100°C for softwood chips, ~200°C for walnut shells, and ~350°C for wood pellets. The effect of increased pull rate generally falls off.

A difference gas out temperature for walnut shells is obvious. The cause is unknown. Comparing the bottom of reduction temperatures to gas out might be interesting.

Air input temperature (measured inside of the air nozzle risers, just before the nozzle) shows no relation to P_reactor. It ranged from 525-725°C, centering around 625°C.

Pyrolysis profile

The differing profiles may in part be due to the varied void space among fuels, allowing more convective heat transfer upward.


Side by Side Comparisions


Wood Pellets - 09/23/09

Softwood Chips - 09/16/09

Walnut Shells - 08/24/09

Contour Plot

Wood Pellets - 09/23/09

Softwood Chips - 09/16/09

Walnut Shells - 08/24/09


Wood Pellets - 09/23/09

Softwood Chips - 09/16/09

Walnut Shells - 08/24/09




Worth Investigating

Other relationships that can be explored in the data:

R_reduction relation to delta T reduction (does reduction packing lead to a larger delta T?)

R_reduction rate of change (how fast does the reduction zone pack, what would be reasonable grate shaking intervals)

Issues with determining R_reduction (largely gas flow measurment issues)

Modeling fluid behaviour:

Air flow through pre-heat lines given P at air cowling to P_comb differnetial. Quick graph shows two curves hot and cold with transition. Can the transition be explained by increased air volume (would it be a single curve when based on air velocity).

T_comb dropped rapidly when the reactor was stopped to refill the hopper at 160 min (walnut shell run). Intentionally stopping the reactor and determining the rate of cooling of combustion could indicate the degree to which fuel is impacting combustion thermally.

Does the delta T_tred - T_bred switch at a consistent temp? A quick look at a noisy fit seems to indicate tred = 750°C is the zero point. This is in line with the lower temp data from Jays Softwood Chip Run on another GEK.


Datasets and Code

The data used to produce the graphs above will be provided shortly.

tar_conditions.csv - Dataset from the runs. Run conditions are sampled at the recorded second that a sample draw was started (and drawn for 15 seconds). Data from the tar samples.


Contact for more information.

Comments (0)

You don't have permission to comment on this page.