I have dug deep into my memory to add 21 more items to the list of process, mechanical, and piping design mistakes that have made our job as automation engineers more challenging and in some cases impossible. We learn the most by our mistakes. Besides preventing the reoccurrence, this list should provide insight to prevent conceptually similar problems.
(10) Radial rather than axial mixing. The dead zone in the upper portion of the vessel creates a large and unknown dead time for concentration, pH, and temperature control.
(11) Horizontal vessel or sump. There are dead zones in most of the vessel affecting concentration, pH, and temperature control. The only well mixed zone is the cylindrical area with a diameter about the same as the vessel height but even here the results are consistent from exchange of material with the other zones in vessel. The sump is worse.
(12) Feed entry point near vessel discharge nozzle. Feed entering the vessel short circuits the mixing zone prematurely ending up in the discharge flow. The result is an erratic, noisy, and inconsistent concentration, pH, and temperature response. For reactions, the shortening of the residence time of a reactant feed to less than the reaction time causes reactant to appear in the discharge reducing yield.
(13) Horizontal distillate receiver. The level change from a change in inventory for an upset is less than sensor noise or threshold sensitivity. For the most popular distillation column control scheme where column temperature is controlled by the manipulation of distillate flow, the correction for a disturbance does not occur until the distillate level controller sees the effect of the change in the distillate flow as an inventory change. The temperature controller subsequently has no effect on the column until it overcorrects the distillate flow and even then there is a limit cycle from the threshold sensitivity limit.
(14) Non-uniform packing distribution in column. Channeling of vapor flow causes inconsistent concentration and temperature control and loss of separation efficiency in packed columns.
(15) Temperature sensor not extending into representative point in a column. The sensor should extend at least 5 thermowell diameters past the wall or heat conduction loss will cause the measured temperature to be a function of the ambient temperature and rainstorms especially for un-insulated columns and sensor nozzles. For columns with packing, the sensor installation should not damage packing or disturb major flow patterns. For columns with trays, the sensor should not block downcomers or bubble caps. The best location may be in the froth at the top of the liquid on the tray.
(16) Temperature sensor or electrode not extending into representative point in an agitated vessel. The sensor should extend at least 5 thermowell diameters past the baffle into the well mixed zone of an agitated vessel. Otherwise the concentration and temperature response will be delayed and inconsistent.
(17) Single temperature location in fluidized bed reactor. There are hot and cold spots in these reactors due to uneven catalyst and flow distribution. Multiple sensors are needed for each traverse to detect hot and cold spots and to offer the computation of an average temperature. Several traverses at various distances from feed entry are needed to see the conversion for various residence times.
(18) Low flow causing a slower heat transfer or mass transfer rate and a higher fouling rate. Velocities less than 1 fps for temperature and less than 5 fps for pH can cause a significant slowing of the response from a decrease in the heat transfer coefficient and mass transfer coefficient, respectively. Even more dramatic is the slowing of the response from an increase in the fouling rate at low velocities.
(19) Variable jacket or coil flow. The throttling of jacket or coil flow for temperature control causes an increase in the process gain and dead time for low coolant demand that can result in severe cycling. For batch vessels and for production rate changes to continuous vessels, the nonlinearity of the process gain and dead time in the jacket or coil temperature loop increases process variability.
(20) The improper mixing of steam and water in a jacket or coil. The transition between coolant water to steam in split range control causes multiple phases to exist in jackets and coils. Localized collections of steam bubbles create hot spots. Bubbles hitting temperature sensors cause erratic readings. Tighter temperature control is achieved by the use of a steam injector to create a cold to hot water transition with no bubbles.
(21) Insufficient surge tank volume. Variability from changes in incoming feed flows cannot be adequately absorbed and is transferred to the tank discharge flow upsetting downstream users. The change in discharge flow tends to become more severe as the level approaches or hits the low or high level limits. For the implication in terms of tuning level controls see the 3/19/2013 blog Processes with No Steady State ... (Conclusion).
(22) Pump head not high enough for high destination static head or high flow. For high static head the flow becomes erratic sensitive to slight pressure disturbances and can even reverse unless there is a check valve. This is a particular problem for variable speed drives as the speed and consequently the pump head is lowered for a low flow demand. For a high flow there may not enough head to match the frictional pressure losses in the system and the control valve. The operating point is on the flat part of the installed curve of a control valve causing a loss of sensitivity and a wandering of the flow. This is a particular problem for conventional butterfly valves due to greater loss of sensitivity from flattening of the inherent flow characteristic at rotations > 45 degrees.
(23) Valve to system drop ratio too low. The valve runs out of available valve drop at flows above design flow and the installed flow characteristic becomes too steep at low flows causing a greater sensitivity to stiction near the closed position. The increase in nonlinearity and the loss in rangeability prevent the valve from doing its job. In an attempt to save energy, inadequate valve to system drop ratios are becoming common.
(24) Sparger ring near electrode. The bubbles hitting or momentarily attaching to the electrode cause a noisy pH or dissolved oxygen measurement.
(25) Axial compressor shaft inertia too low. The unloading of the impeller when the compressor goes into surge and flow reverses can cause the compressor to accelerate. In one specially designed high efficiency compressor, a speed acceleration measurement and shutdown was needed to prevent the compressor from reaching a damaging high speed in less than a second.
(26) Heat exchanger in recirculation line with low heat transfer area or coefficient. If the heat transfer area is low, the self-regulating initial response to a change in coolant flow is complicated by an integrating response from the recycle effect of vessel contents.
(27) Dip tubes too large or long for low reagent flow. The use of a conventional dip tube size and length results in huge injection delay of reagents. For low reagent flows in neutralization processes, the delay can be an hour or more when the valve is opened after being closed for a long time. Also, the slow draining of reagent in the dip tube after closing can cause the pH to drift for hours upsetting batch operations.
(28) Gravity feed of reagents. The flow depends upon the height of reagent and distribution in the supplying vessel. Changes in flow at the vessel from a rotary airlock valve or slide gate valve can have a significant transportation delay to the destination.
(29) Solid or gaseous reagents. The time required for the solid or gas to dissolve may exceed the residence time of the equipment. This is true for flashing ammonia in static mixers and magnesium hydroxide or lime in vessels.
(30) Multiple liquid volumes between a manipulated flow and a sensor. The residence time of each well mixed volume can be considered to be a time constant. A single process time constant is beneficial. Multiple process time constants are detrimental because time constants in series create dead time. The manipulation of a feed or heat transfer fluid to a volume upstream of the volume whose concentration or temperature is being measured and controlled can have a horrendous process dead time. The dead time problem increases with the number and size of the intervening volumes.
