In this case study we discuss how TRACLabs Inc. has used a layered, intelligent control system to operate the Cascade Distiller System (CDS), a water processing system (see Figure 1) that, through a joint NASA-industry rapid development program, is being advanced from proven feasibility (technology readiness level 3) to a technology wpnstration (technology readiness level 6). TRACLabs Inc.’s layered control software, known as 3T, separates the general intelligent control problem into three interacting pieces

• A reactive bottom layer consisting of a set of hardware-specific modules, for example, switchers, low level control algorithms, or data analyzers, which are tightly bound to the specific hardware and must interact with that hardware in real-time. The software modules in this layer are called skills and interface to the middle tier via skill managers.
• A sequencing middle layer which differentially activates the reactive modules in the bottom layer in order to direct changes in the state of the world and accomplish specific tasks. The skills connect to the sequencer via Ethernet.
• A planning/scheduling top layer which reasons in depth about goals, resources and the future effects of current actions.

A total of thirty-six skills and eleven skills were developed for the CDS and facilities low-level controllers respectively. The skills used for the CDS are relatively simple — opening and closing valves, throwing relays and watching for tanks to be filled or emptied and set points to be achieved. The skills level gives the fastest response, so we implemented a timed-valve toggle skill, for example, that was useful for some of the draining procedures (see sequencer section). In the same vein the skills controlled the facility valves that introduce de-ionized (DI) water into the CDS so as to prevent overfilling if communications with the sequencer level were lost. We also developed a skill to compute average consumption and production rates and specific power requirements.

Operating the CDS

The distiller drum was spun up to 1200 rpm and the system was placed under vacuum, ≤ 0.3 psia. Then the product (cold) loop was filled with DI water. For continuous operation (as in mission testing), product water remained in the product loop so this priming was needed only for the first batch. A flow switch indicated when the loop was full, at which time the feed water was allowed to flow to the distiller. The feed was pulled from the feed tank into the distiller across a regulator valve, based on the action of pitot tubes in the brine (hot) side of the distiller chambers. A flow switch in the hot loop indicated when the loop was full at which time the thermal electric heat pump (THP) was activated and distillation began. Product water was pushed into the product tank across a check valve based on the action of pitot tubes in the cold side of the distiller chambers. Once the hot loop temperature reached 40°C, chilled water was allowed to flow through the heat exchanger on the cold loop to regulate the temperature. The sequencing layer of the control system automatically executed this particular sequence of activities to start the CDS. Once the CDS was in operation, the control system monitored for several off-nominal situations:

• Loss of flow in either loop
• Loss of power to the THP or the distiller
• Change in distiller speed of more than 50 RPM
• Hot loop temperatures higher than 55°C
• Uninterruptible power system switching to battery for either the facility or CDS controls

Any of these conditions would cause the system to shutdown. Shutdown consisted of closing off the feed, and waiting two minutes to process the residual feed in the feed line. Then the THP was turned off and a delay of three minutes was instigated for the chiller to equalize the temperatures in the loops, thus halting distillation. Then the chiller was turned off and both loops were drained to the tanks. The CDS was then re-pressurized and the distiller drum spun down. At this point the staff usually took samples from each of the tanks. Then the product and brine tanks were drained. A ten-liter batch of feed produced nine liters of product water (a 90% recovery) and one liter of brine. To avoid damaging the motor, applying vacuum and re-pressurizing are conducted in an order that insures any residual fluid is always being directed away from the distiller into the tanks.

Results

The CDS control system ran successfully for over 1600 hours during the 50 week test. The software provided for safer CDS operations with a smaller operating staff and it allowed for more efficient operation.

Key Publications

Bonasso, R.P. A Distributed 3T Control System to Manage Readiness Testing of a Cascade Distiller System, in Proceedings of International Conference on Environmental Systems, San Francisco, 2008.