ABSTRACT (2017-404)

Long-term challenges for the oil spill response industry include spills occurring in the open ocean or ice-infested waters, as well as spills of heavy oil. Historically, only a small percentage of the oil from major spills is ever recovered. For example, during the Deepwater Horizon incident, skimming operations only accounted for approximately 3% of the spilled oil. This low number was due to the delayed response of assets as well as the use of inefficient skimming systems and methods. Both government and private organizations seek effective response solutions that are capable of rapid deployment to existing platforms and infrastructure.

Alion Science and Technology, under contract to the Bureau of Safety and Environmental Enforcement has developed two proof-of-concept technologies; both of which were successfully tested in 2016. Both of these research projects have been pushing the edge of the envelope for oil spill recovery operations with the use of existing technologies.

The first of these concepts is the “ICEHORSE”, an oil-in-ice skimmer designed to recover surface oil pooled in broken ice fields. The system is designed to submerge and transit below the ice field where it may then be surfaced in a pocket of oil to begin skimming. This allows for remote skimming operations while the boat remains in safer waters and gives the skimmer the ability to position itself in the thickest areas of oil. The initial proof-of-concept system has seen further development with the goal being a commercially available product.

The “Autonomous Oil Skimmer” is designed to be used with any vessel-based skimming system to improve the efficiency of recovery operations. Our proprietary control algorithm collects oil thickness data from a sensor and then directs and tracks the oil recovery operation autonomously. Collection data may be remotely accessed, allowing for centralized project oversight of multiple resources. Part of this effort included the evaluation of different oil sensors and the development of algorithms to incorporate the sensor data into the control system.

This paper discusses some of the design challenges of each system as well as the results of testing, and plans for further development.

INTRODUCTION

Both of the systems described in this paper were developed under Broad Agency Announcements (BAAs) for the Bureau of Safety and Environmental Enforcement (BSEE). The ideas were proposed by Alion in white papers to BSEE that were then funded under two separate BAAs.

ICEHORSE

The goal of the ICEHORSE project was to develop an oil recovery system with the capability to skim oil from heavily ice-infested waters. To meet this goal, a system with an off-the-shelf drum skimmer was devised. The system can submerge and maneuver below the ice pack to open patches of broken ice and oil where it then resurfaces. Once surfaced the skimming operations may commence and the entire skimmer assembly is then able to be maneuvered throughout the open ice field. Upon completion of oil recovery in the ice field the system may again be submerged and driven to another area of ice/oil or back to the support vessel.

The proof of concept design of the ICEHORSE uses an Elastec MiniMax drum skimmer supported by an aluminum tube frame. The skimmer is surrounded by an aluminum ice cage that protects it from large chunks of ice. To keep small chunks of ice from entering the ice cage and disrupting the effectiveness of the drums a sacrificial plastic mesh is placed around the ice cage. The flotation tanks of the MiniMax skimmer were modified to act as pneumatic ballast tanks giving the ability to submerge and surface from a surface mounted air supply and control panel. Locomotion of the ICEHORSE platform is provided by three JW Fishers SeaLion-2 ROV’s. In order to control the three separate ROV’s Alion developed a control algorithm that allowed for simple operation from a single joystick. The resulting system maintains the desired portability needed in oil response technology and fits within a standard cargo van or covered trailer (see Figure 1).

Figure 1:

System fits in standard van (left) and front view of system with JW Fisher’s ROV’s (right).

Figure 1:

System fits in standard van (left) and front view of system with JW Fisher’s ROV’s (right).

Performance Testing

Multiple tests were constructed to measure the performance and maneuverability of the ROV system. These tests included maneuvering tests that involved the time to maneuver the system 180° (all tests with 2” discharge hose and ROV umbilical attached), the time to travel in a straight line over a set distance while surfaced and submerged, and the time to fully submerge and surface. Operational tests were conducted for two ice concentrations, 30% and 70%. In both operational scenarios, the ICEHORSE was able to easily surface in open water/oil and maneuver within the broken ice field as well as push and corral ice and oil. The ice cage and mesh surrounding the skimmer did an excellent job at keeping ice away from the drum faces (see Figure 2).

Figure 2:

ICEHORSE transiting to broken ice (top left), beginning to surface in ice field (top right and bottom left), ICEHORSE maneuvering through ice field and skimming oil (bottom right).

Figure 2:

ICEHORSE transiting to broken ice (top left), beginning to surface in ice field (top right and bottom left), ICEHORSE maneuvering through ice field and skimming oil (bottom right).

ICEHORSE Conclusions

The ICEHORSE prototype proved to be a great success during the testing performed at Ohmsett. The system had the ability to be quickly deployed and was easily maneuvered around the test bed. It successfully demonstrated the concept of submerging an oil skimmer to go under an ice pack in order to skim oil from ice-infested waters.

Alion is continuing to advance this technology (again with BSEE funding) by developing a new system that will make the ICEHORSE more commercially viable. Rather than the frame of the system being driven by ROV’s, the underlying frame itself will become the ROV with integrated thrusters and an advanced ballasting system. This should allow for even better control above and below the water surface and allow for any skimmer with a similar form factor to the initial design to be mounted to the frame. Future testing will again be conducted at Ohmsett and also seek to gather accurate oil recovery data for different recovery heads with a variety of oils in ice-infested waters.

AUTONOMOUS OIL SKIMMER

Under this project, Alion has developed an automated skimmer system that can autonomously maneuver and skim the oil from a given area with automatic tracking and reporting of progress and performance. This Autonomous Oil Skimmer (AOS) consists of a commercial-off-the-shelf (COTS) skimmer and vessel, a COTS autopilot system, a high precision navigation package, oil sensors, and custom control software (see Figure 3). The result is a strap-on navigation, sensor, and computer control system that can be used with a variety of skimmer and vessel systems for maximum oil spill response flexibility. Custom control software communicates with the navigation instruments to control and track the location of the vessel. The control software seeks to optimize the oil recovery operation by attempting to steer the vessel through the thickest areas of oil the system has previously encountered. The thicker areas of oil are measured using an oil thickness sensor combined with an oil-in-water discharge sensor that also allows the mapping of the encountered spill when combined with the GPS coordinates.

Figure 3:

AOS system flowchart

Figure 3:

AOS system flowchart

As previously stated, the AOS package is designed so it may be delivered to an existing Oil Spill Response Organization (OSRO) support vessel that is outfitted to conduct underway oil skimming. The system is then connected to existing auxiliary ports on the vessel’s autopilot and navigation system. The oil thickness sensor is then mounted to the front of the skimmer or other suitable area where it may come in contact with the same oil the skimmer will interact with. The oil-in-water sensor is installed on the discharge hose of the skimmer and measures the amount of oil actually captured by the skimmer. The data from the thickness sensor is used to map and direct the oil skimming operations while the discharge sensor attempts to track the actual amount of oil recovered in real-time. This data may be also be used to improve the efficiency of the oil collection by adjusting recovery rates of the skimmer, etc. and also allows a supervisor to direct operations from a remote location.

AOS Testing

Testing was performed in multiple stages to allow the testing of different portions of the system in controlled and uncontrolled environments. The first stage of testing involved identifying the sensors needed to capture thickness and oil percentage data. Sensors of interest were tested at the Ohmsett facility in Leonardo, New Jersey. To test the sensors ability to track oil thickness while moving, each sensor was passed through oil patches of known thicknesses at varying speeds. The two sensors of interest, an Arjay 2852-HCF and GE Leakwise ID-227, were both able to detect the oil and water interface but struggled to generate an accurate estimate of thickness. It was determined this was mostly due to hydrodynamic issues (the commercial oil sensors identified for the initial testing are intended for static applications). This led to Alion designing an articulated sensor mount that allows the thickness sensors to better track the surface of the oil/water (see Figure 4).

Figure 4:

Articulated floating sensor mount shown with Arjay 2852-HCF.

Figure 4:

Articulated floating sensor mount shown with Arjay 2852-HCF.

Testing of the control system was performed at Gardner’s Lake in Connecticut. This location was chosen to provide a semi-controlled environment where performance was less likely to be influenced by waves, vessel traffic, or current. A COTS autopilot system was installed on the support vessel to allow the control system to autonomously steer the vessel in the desired manner. The control system software was modified to read simulated oil thickness readings from a simulated spill based upon current position instead of actual oil thickness readings, but the remainder of the control software was the same as would be used for a real spill (see Figure 5). Although no oil was used, the full system was set up and deployed on a commercial oil skimming vessel. See Figure 7 for a schematic of the equipment loaded on the vessel and Figure 6 for a picture of the vessel at Ohmsett. This round of testing quantified the ability of the support vessel to follow the algorithm’s desired commands. This data was then used to further develop the oil recovery algorithm by ensuring steering maneuvers that are realistic and would not force oil under or over the booms when driving the oil skimming vessel.

Figure 5:

Example view of the estimated oil map, recovered oil, sensor data, and algorithm navigation information.

Figure 5:

Example view of the estimated oil map, recovered oil, sensor data, and algorithm navigation information.

Figure 6:

Scaled support vessel used for testing at Ohmsett and Gardner’s Lake showing an Elastec TDS-118 rotary drum skimmer .

Figure 6:

Scaled support vessel used for testing at Ohmsett and Gardner’s Lake showing an Elastec TDS-118 rotary drum skimmer .

Figure 7:

Layout of scaled support vessel for Ohmsett and Gardner Lake testing.

Figure 7:

Layout of scaled support vessel for Ohmsett and Gardner Lake testing.

The final round of testing was again performed at the Ohmsett facility. Since the control system’s ability to steer the vessel was thoroughly tested at Gardner Lake these tests were aimed at measuring the redesigned floating sensor mounts ability to maintain contact with the surface of the oil/water with varying oil thicknesses, wave heights, and speeds. The support vessel was towed by the Ohmsett’s movable bridges over the tank for safety and to maintain a constant speed. A channel constructed of floating booms was constructed so an estimated oil thickness could be generated based on oil volume. Controlled testing of the oil-in-water discharge sensor was also performed at this time. See Figure 8 for a diagram of the test set-up at Ohmsett and Figure 9 for a photo of a test in progress.

Figure 8:

Actual test setup of Ohmsett oil spill tank.

Figure 8:

Actual test setup of Ohmsett oil spill tank.

Figure 9:

Support vessel and skimmer being towed through an oil slick.

Figure 9:

Support vessel and skimmer being towed through an oil slick.

Alion prepared two oil slick thickness sensors for testing: an Arjay 2852-HCF and a GE Leakwise ID-227. One day of testing was allotted for each of the two sensors. Using new Hydrocal 300 test oil, the bridges traveled in the southerly direction at speeds ranging from 0.75 to 1.5 knots. As the system traveled along the length of the basin, multiple oil slicks of controlled thicknesses ranging from 5 mm to 15 mm were produced. Each test run consisted of two to three individual oil slicks approximately 50 feet in length with approximately 75 feet of open water separating each. At the conclusion of each run, the remaining oil was cleared to the North of the test area in order to minimize any residual surface oil that might interfere with subsequent readings. Testing was performed in calm surface conditions and waves with amplitudes ranging from 4 to 10 inches and wavelengths of 25 to 42 feet. A series of 12 tests using defined parameters were repeated for each sensor. Figures 10–13 show sensor readings overlaid with a concurrent photo for four of the tests.

Figure 10:

Example of Arjay sensor data upon leaving a slick in calm water.

Figure 10:

Example of Arjay sensor data upon leaving a slick in calm water.

Figure 11:

Example of Arjay sensor data after leaving a slick in wave action (boom is being pushed to one side by wind).

Figure 11:

Example of Arjay sensor data after leaving a slick in wave action (boom is being pushed to one side by wind).

Figure 12:

Example of GE sensor data leaving an oil slick in calm water.

Figure 12:

Example of GE sensor data leaving an oil slick in calm water.

Figure 13:

Example of GE sensor leaving an oil slick in wave action.

Figure 13:

Example of GE sensor leaving an oil slick in wave action.

Oil-in-Water Sensor Setup

In order to evaluate the performance of the oil-in-water sensor, the objective was to provide conditions of varying oil/water concentration in a flow stream for measurement purposes. To accomplish this, a stationary skimmer was set up in a test area of approximately 15 feet by 15 feet that provided a controlled environment to recover the oil and water to be sent through the discharge hose. An Elastec TDS118G grooved drum skimmer was used as the collection device with one wiper removed in order to reduce the recovery rate of the skimmer, achieve longer steady state conditions, and reduce oil consumption during each test (Figure 14). A cargo hose was routed from the skimmer’s recovery pump to the Arjay 2852-OWI sensor, which was installed in a hard pipe and rigidly mounted to a forklift mast for stability. Immediately after the sensor location, a three-way valve was installed, which allowed for flow to be directed into a collection tank or to a second exit port for sampling.

Figure 14.

Actual discharge sensor test setup.

Figure 14.

Actual discharge sensor test setup.

Oil-in-Water Sensor Test performed and Results

One day of testing was allotted for the oil-in-water sensor. Prior to test start, the area was preloaded with used Hydrocal test oil to create a ~3-inch slick. Testing began by operating the skimmer at a slow drum speed to achieve a high oil-to-water ratio for the sensor measurement. Operational speeds of the skimmer were varied until the desired sensor output reading was achieved and reasonably steady. Flow was then quickly diverted from the discharge hose to sample containers (5 gallon bucket or 1 quart jar). Two samples were collected in sequence for each of the oil/water concentrations. Skimmer efficiency was intentionally increased or decreased in small increments in order to cover the full signal range of the sensor, which spanned 7 mA to 20 mA based on oil/water concentration.

The sensor data was recorded continuously. Time markers were inserted into the data stream at the start and stop of each sample in order to compare fluid samples representative of the fluid stream passing the sensor as data was recorded. The point for each sample collection plotted in Figure 16 is the averaged sensor readings with standard deviation between the recorded start and stop marker. An example graph of the recorded sensor data is included in Figure 15. It was found to be difficult to keep a constant oil/water concentration, which led to constantly varying sensor output that can be observed in this graph. The negative slope of sample 3 results in a larger standard deviation than sample 4. Samples collected were later analyzed in the lab for BS&W.

Figure 15.

Example Oil-In-Water sensor output.

Figure 15.

Example Oil-In-Water sensor output.

Figure 16.

Oil-In-Water sensor results.

Figure 16.

Oil-In-Water sensor results.

In an effort to cover the entire sensor range and conserve oil, the discharge hose was routed back to the test area for several tests. During these tests (labeled tests 33–36), increasing emulsification of the test oil was observed. The test area was skimmed and cleared before adding one inch of Hydrocal prior to the start of test 37. Four tests were performed in this configuration with samples retrieved for analysis.

Oil-in-Water Sensor – Results

The results of the Oil-In-Water sensor test are shown in Figure 16. The sensor appears to do a fairly good job of tracking instances of high oil concentration. There were four anomalous data points (circled); it is not clear why the sensor reading was so high for those particular points. As the sensor readings drop below 70% the readings become fairly erratic. There are two possible causes for this. First, the sensor is primarily intended for detecting low concentrations of water in oil and the manufacturer only claims good performance at higher oil concentrations. Second, it was difficult to use the skimmer to get higher water concentrations as it is an oleophilic skimmer that is designed to pick up oil and not water. To increase the water content a hose was run into the skimmer sump and water added; this water may not have been well mixed with the oil prior to being pumped out of the sump and past the sensor.

ALGORITHM PERFORMANCE

During the Gardner Lake test, it was determined that the original “lawn mower” skimming pattern is not ideal for oil skimming due to the limited turn rate of the skimming vessel. Using a modification of the original MATLAB program and algorithm, several new potential oil skimming algorithms were implemented taking this slower turn rate into account. These algorithms were implemented and tested on several different spill shapes of approximately the same area in order to compare their performance relative to each other. The three oil spill patterns used were: oblong, round, and narrow (see Figure 17).

Figure 17.

Oil spill patterns; oblong (top left), round (top right), and narrow (bottom).

Figure 17.

Oil spill patterns; oblong (top left), round (top right), and narrow (bottom).

The original MATLAB program was restructured and simplified to make the implementation of new algorithms easier. Simple vessel movements were placed into functions to create a cleaner and simpler implementation interface. The new program was also modified to allow for standard spill shapes to be selected for consistency between algorithm simulations. Geographic coordinates were replaced with Cartesian coordinates for simplicity and clarity.

Several features were also added to the simulation program to make the simulations more realistic and the algorithms more efficient. The simple averaging spill-smoothing function was replaced with a gradual multi-step vessel path smoothing function in addition to periodic global Gaussian spill smoothing to more accurately represent oil slick diffusion. To increase skimming efficiency, an optional geometric virtual boundary was added that can restrict the vessel’s movement to the more concentrated areas within the spill. Also, the oil pickup method itself was changed to more realistically model oil skimming, which is largely dependent on the maximum pickup rate of the skimmer.

Seven algorithms that produced different patterns to recover oil were tested. The algorithms were tested using different oil spill patterns and different vessel speeds. The looping (with flip) pattern (see Figure 18) seemed to perform slightly better across all tests (see Figure 19). The looping algorithm consisted of the following steps.

  1. Starting at a specified position, travel perpendicular to spill direction until edge is reached.

  2. Turn around 180 degrees at edge and travel through spill, parallel to original path.

  3. Turn at the next edge, the same direction as before and slightly more than 180 degrees, to increment the looping along the length of the spill.

  4. Repeat steps 2–3 until the conditions for step 5 or 6 have been met.

  5. (optional) When the pattern begins to overlap the previously-covered area, change the turn direction to “flip” to the next pattern segment and again repeat steps 2–3.

  6. When the edge is reached, return to specified starting position and repeat steps 1–4 in another orientation relative to the spill.

Figure 18.

Looping algorithm pattern.

Figure 18.

Looping algorithm pattern.

Figure 19.

Spill 1, oblong pattern results.

Figure 19.

Spill 1, oblong pattern results.

CONCLUSIONS

Thickness Sensors

Neither of the two thickness sensors worked as well as desired. Although both easily detect the transitions from oil to water and vice versa, there is no solid thickness assessment. The sensors appear to be able to give some relative indication of thickness, but only in a gross sense, ot small differences in thickness. It is not clear whether this is a fundamental limitation of the sensor technology or the hydrodynamic packaging. One flaw with any sensor that must be in contact with the oil surface is that the act of making the measurement disturbs the surface being measured. Also, if it does not remain exactly on the surface, then the measurement is also inaccurate. It is difficult to ensure a contact sensor remains exactly on the surface given the interactions of the waves (swell and chop) with the movement of the sensor float and supporting structure (skimmer and vessel). Attempts to improve the performance will just make the entire structure more complex and thus less supportable in the field (both installation and maintenance as well as use by non-technical vessel operators).

Inline Sensor

The inline sensor worked reasonably well for high oil concentrations. Although the sensor was erratic below about 70% oil, this should not be a problem for typical oil spill recovery operations as most skimmers have efficiency much better than this. Coupled with a flow sensor, this could be used to provide real-time estimates of recovery rate and efficiency.

Autopilot / Vessel Integration

The Alion AOS concept is for the system to be mounted onto a vessel of opportunity, mainly (but not exclusively) vessels belonging to OSROs. To this end, partnering with an OSRO would be extremely beneficial for further open water vessel testing. Many of the OSRO’s have vessels outfitted with commercial autopilot systems, the latest in underway skimming technologies, and advanced sensor packages. This partnership would aid in the integration of the system to existing platforms as well as gain access to a wealth of knowledge from professional oil spill management personnel.

If a standardized system is developed and implemented, an oil recovery workboat could have the modifications already installed to ease the installation of the AOS. Instead of the field technicians “tapping” into the autopilot a pre-installed access port would allow a quick installation of the AOS. This would be analogous to a harbor pilot boarding a vessel for pilotage and plugging his navigation computer into the existing “pilot port” all large vessels are required to have.

For vessels that do not have an existing autopilot system or where safety concerns necessitate manual steerage, the system may still be used to direct the operations by providing heading commands to an operator using the display. This is analogous to the way people receive directions from the GPS map system in their car, which provides turn-by-turn driving instructions, but does not actually steer the car.

For both applications, a monitor for the crew should be designed to aid in improving efficiency of the skimming operations. Information such as current heading and position may be showed as well as the estimated time of arrival to the next waypoint and the intended new course. A portion of the screen may also be dedicated to the oil collection statistics like estimated quantity of oil collected to aid with scheduling tank offloads. For manual steering operations this information can be augmented with audible cues much like current automotive navigation systems.

Algorithm Development

Because the skimming algorithm simulations were unable to accurately model many of the intricacies present in real world oil skimming, it would be valuable to perform tests of algorithms using an actual oil-skimming vessel. Different factors such as the safe turning radius, optimal speed, and actual skimming efficiency can be tested by running several algorithms using different sizes of vessels and different types of oil skimmers.

The biggest challenge in further development of this autonomous skimming system is capturing accurate data of both oil thickness and the percentage of captured oil. Future work for the efficacy of this system should involve developing or identifying better technologies for the estimation of the oil thickness and percentage in water in a wide variety of conditions.

ACKNOWLEDGEMENTS

This work was funded by BSEE, U.S. Department of the Interior, Washington, D.C., under Contract E14PC00031 and E14PC00035.