Knowledge Best Defense for Fire Risks

Biodiesel production has no more or less fire hazard than an oil refinery. Plant personnel at all levels need to be vigilant and totally informed, realizing that what they do not know can lead to disaster.

In my background of 23 years of operations and 41 years of plant design and construction, I have witnessed seven serious fires and two explosions. One victim spent four hours in brain surgery but survived. Another missed death by minutes. Each accident involved ignorance. Biodiesel plants and oil refineries are not for the ignorant—or the inattentive.

Is there any operator who does not know that all materials in a biodiesel plant can burn and also explode? Still, fires occur too often and the results are often disastrous. Costs, legal issues and lost time after a fire are very difficult to overcome. As educator Derek Bok once said, “If you think education is expensive, try ignorance.” Are there ways to reduce the risk and severity of fires? For starters, a plant safety manual is critical.

We all know the fire triangle’s three sides: oxygen, fuel and ignition.

Oxygen is as important for us as carbon dioxide is for plant life; it cannot be removed or reduced except inside tanks using inert gas.

The fuel leg of the triangle is a fertile field for wise decisions. The most hazardous material is obviously alcohol. A good source of information on safety in handling this volatile chemical is the Methanol Institute and its manual (www.methanol.org). A listing of methanol incidents, pages 133-147, deserves study; note the number of fires and causes—often maintenance and hot work—and spills, which could have led to disastrous fires. The number of fatalities is shocking.

Extinguishing methanol fires requires understanding methanol-water solubility. Methanol mixtures of 25 percent or higher will burn. Dikes around methanol tanks need to be higher to contain diluted methanol. Plant and animal oils have high flash points, but once burning they provide great energy, leading to total disaster.

Ignition sources deserve our total attention during design and construction—and daily thereafter. Building, electrical and fire codes are an excellent foundation. One topic of concern is electrical classification of areas. Methanol is a Class 1 Group D material. In addition to explosion-proof equipment, instruments and lighting, I wonder if the dimensions for sumps are adequate. Visualize what would happen to vapor in a serious overflow. Static sparks and lightning are often a surprise. Grounding of all equipment and piping is vital, requiring special care in daily grounding to trucks and rail cars. Discharge from air hoses and even steam can create static discharges. Written hot work permit forms and procedures that require sign-off by both operations supervision and technical personnel are vital.

During the phase of process selection and design, there are opportunities to reduce events and inventory, especially inside buildings, and to reduce the dangers inherent in transfers. A continuous process will normally have less inventory and reduce transfer errors (e.g., overfill, transfer to a wrong tank, pump against closed valves, etc.). Product can be more uniform and reduce rework of off-spec material. This safety feature has led me to develop a very small continuous reactor and an accurate, pulse-free methanol feed pump. Less inventory and risk mean less fire-extinguishing equipment, less insurance premiums and lower operating expenses.

Equipment selection can also reduce fire risk. Vapor pressure of methanol is high (280 psig at 100 psi steam temperature) and can cause overpressure. It is probably best not to heat pure methanol. Heating methanol between closed valves (a blocked-in pump) without a relief valve can rupture pressure gauges. I have witnessed an instrument technician use “autotune mode” in a temperature controller during start-up; it overshot and filled one end of the building with methanol vapors. Fortunately, it did not explode. It is also good practice to minimize connections, flanges and valves to help reduce leaks and errors.

Choices made in construction materials can cause—or help prevent—future fires. The Methanol Institute has a section on corrosion of metals and suitability of gaskets. It is not complete, however, and does not cover corrosion of mixtures of methanol and catalysts (sodium hydroxide, sodium methoxide or sulfuric acid), nor are there charts of corrosion rate as mils/year versus concentration and temperature. Charts for methanol, glycerin, sodium hydroxide and sodium salts can be found in Perry’s Chemical Engineers’ Handbook.

Biodiesel production has no more or less fire hazard than an oil refinery. Plant personnel at all levels need to be vigilant and totally informed, realizing that what they do not know can lead to disaster.

Advertisements

Avoid Pump and Compressor Errors

Don’t assume sinusoidal flow for piston-type devices.

Piston-type pumps, compressors and engines usually are driven by a crank and connecting rod. During the stroke, flow rates don’t follow the sinusoidal curve that textbooks and manufacturers’ literature often cite. Peak rates are up to 10% higher than predicted by a sinusoidal curve and don’t occur at 90° crank rotation. These higher rates increase values of net positive suction head required (NPSHR) by as much as 20% and cause greater stresses and possibly added maintenance. The actual curve is a distorted bell curve controlled by the length ratio of the stroke and connecting rod. Rods shorter than the stroke but longer than the connecting rod need special crossheads to avoid metal interference but create amazing distortion. Knowing the length ratio of a pump and its relation to flow rate provides insights on dynamic and hydraulic features that affect performance.

Crankshaft and Crank

Figure 1. This basic design is used for piston pumps, compressors and engines.

In reciprocating pumps, compressors, etc., the crankshaft and crank move a connecting rod and piston in a cylinder (Figure 1). The crankshaft center is placed at 90° on the 0°–180° x axis; the crank’s rod bearing is shown at 45°. Clockwise rotation of the shaft will cause the crank bearing to generate a sinusoidal x-y curve between 0° and 360°. Figure 2 shows this sinusoidal velocity curve between 0° and 180°. It also shows the actual velocity curve for a piston when the rod length divided by crank length is 2.1. It is not sinusoidal.

Flow recorders could provide a pump flow profile but it’s much easier to use simple trigonometry to calculate piston positions versus crank angles, then calculate piston travel per degree of crank rotation and plot the results. They are not sinusoidal. (Compressors and piston engines would follow similar curves.)

Piston Velocity versus Crank Angle

Figure 2. Actual curve peaks at a different location and has higher peak.

I’ve observed the following:

  • As already mentioned, peak flow rates are up to 10% higher than a sinusoidal curve predicts.
  • The curve shape depends on the ratio of rod length to crank length.
  • Peak rate doesn’t occur at the 90° point but rather at 95°–120° depending on the ratio.
  • From 180° to 360° (the suction portion), the curve is a mirror image of the 0° to 180° discharge portion.
  • During the suction portion of the curve, flow rates also are higher and peak earlier than the 270° point.
  • The curve also will change if the centerline of the cylinder doesn’t pass through the center of the crankshaft.
  • Multi-piston pumps and compressors provide less “smoothing” effect than predicted because the bell-shaped curve has a sharper peak.

THE IMPLICATIONS

The difference between actual and sinusoidal curves affects many facets of operation and mechanical design:

  • Check valves and passages will have higher-than-predicted peak flow rates, and pressure drop will be higher — by the square of flow rate (20%).
  • This will impact NPSHR and possibly induce vaporization as the column of liquid in the suction pipeline is accelerated through valves, etc.
  • Even a small amount of vaporization or release of dissolved gas will reduce pump discharge rates substantially. Because flow meters in pulsation flow aren’t practical, difficulties often first appear downstream as corrosion or chemical problems. Plant engineers with any indication of such problems should check the many NPSH factors such as air pockets in the pump, supply tank level error, plugs in valves or piping and check-valve backflow.
  • Maximum rotation per minute (rpm) will be lower than indicated by the sinusoidal curve. If a pump is running near the manufacturer’s maximum recommended speed, starved suction is a possibility.
  • Loads on bearings will rise somewhat, especially in high-speed compressors.
  • Stresses in connecting rods will increase but rod failure is rare.
  • Manufacturers of pulsation dampeners and surge suppressors use the sinusoidal curves in their literature and possibly sizing formulas. Yet, surge dampeners must handle the sharper peak of a bell curve compared to a sinusoidal curve. While these devices can’t provide constant or pulse-less flow, they do make a major contribution unless terribly undersized, improperly installed or the gas cushion is lost.

Calculate Actual Flow Rates

Refer to Figure 1.(A spread sheet will aid calculations.)

  1. Crank length = OC; piston rod length = CP.
  2. Assume an angle a = SOC.
  3. Line AC = OC × sin a.
  4. Line SA = OC – OC × cos a.
  5. Line AP = (CP2 – AC2)0.5.
  6. Line SP = AP + SA (piston travel from 0°).
  7. To find flow rate at any crank angle:

a. Calculate piston position and travel for two crank angles, perhaps 2° apart.

b. The difference in piston positions equals piston displacement over the time interval between the two crank angles. It is an average over the two readings, not an instantaneous rate. As the step size approaches zero, displacement nears true velocity.
c. This can be converted into flow rates of in3/min, gpm or other units, if piston diameter and rpm are known.

ASSESS ASSUMPTIONS

The above comparisons should help you understand the real flow environment and thus avoid NPSH and pipe-sizing errors. Calculating stresses due to velocity and acceleration, possibly with simple algebra (see sidebar), can aid in optimizing design of pumps, compressors and engines.

It’s all too easy, even in engineering, to accept what appears to be logical. Unfortunately, incomplete or inaccurate analysis leads to less-than-optimum results in many different fields. Another common failing is inappropriately extrapolating data — see, for instance, “Show Some Skepticism.”