without an artificially induced notch at horizon for situations where fault infor- tem operation is off nominal.
nominal. The re-
different rotational loading speed levels. mation is available. Further, this work sults obtained showed that the detection Health monitoring verification was per- presents a select evaluation of an online algorithms are capable of predicting formed by integrating three different health monitoring scheme of a rotating anomalies in the rotor disk with very advanced machine-learning algorithms disk using a combination of high-cal- good accuracy. Each detection scheme for anomaly detection in continuous iber sensor technology, high-precision performed differently under the same data streams from spinning tests of a in-house spin test system facilities, and experimental conditions, and each de- subscale turbine engine-like rotor disk unprecedented data-driven fault detec- livered a different level of precision in up to a speed of 10,000 rpm. tion methodologies. terms of detecting a fault in the rotor. This study compares an outlier detec- The methodologies applied in this Overall rating showed that both the tion algorithm (Orca), one-class sup- study can be considered as a model- Orca and OCVSM performed better port vector machines (OCSVM), and based reasoning approach to engine than the IMS technique. the Inductive Monitoring System (IMS) health monitoring. Typical model-based This work was done by Ali Abdul-Aziz, for anomaly detection on the data reasoning techniques compare a system Mark R. Woike, Nikunj C. Oza, and Bryan streams. These techniques were used to model or simulation with system sensor L. Matthews of Glenn Research Center. Fur- inspect the experimental data under data to detect deviations between values ther information is contained in a TSP (see the same operating conditions em- predicted by the model and those pro- page 1). ployed in the tests, and using the meas- duced by the actual system. In fact, a Inquiries concerning rights for the commer- ured vibration response (blade tip model-based reasoner uses the collected cial use of this invention should be addressed clearance) as a key input to check the system parameter values as input to a to NASA Glenn Research Center, Innovative viability of these techniques on detect- simulation and determines if a particu- Partnerships Office, Attn: Steven Fedor, Mail ing the disk anomalies and to evaluate lar set of input values is consistent with Stop 4–8, 21000 Brookpark Road, Cleve- the performance of each methodology. the simulation model. When the values land, Ohio 44135. Refer to LEW-18758-1. The performance of the algorithm is are not consistent with the model, a measured with respect to the detection “conflict” occurs, indicating that the sys-
Measuring Air Leaks Into the Vacuum Space of Large Liquid
Hydrogen Tanks John F. Kennedy Space Center, Florida Large cryogenic liquid hydrogen These tanks are roughly 70 ft (≈21 m) in weigh the remaining liquid hydrogen. tanks are composed of inner and outer diameter (outer shell) and the inner This upgrade, allowing boil-off data to shells. The outer shell is exposed to the shell is roughly 62 ft (≈19 m) in diame- be obtained weekly instead of over sev- ambient environment while the inner ter, so the evacuated region is about 4 ft eral months, is ongoing, and will then shell holds the liquid hydrogen. The re- (≈1 m) wide. provide a relatively rapid indication of gion between these two shells is evacu- A small leak’s primary effect is to in- the presence of a leak. ated and typically filled with a powder- crease the boil-off of the tank. It was This work was done by Robert Youngquist, like insulation to minimize radiative preferable to install a more accurate fill Stanley Starr, and Mark Nurge of Kennedy coupling between the two shells. A tech- level sensor than to implement a boil-off Space Center. Further information is con- nique was developed for detecting the meter. The fill level sensor would be tained in a TSP (see page 1). KSC-13211 presence of an air leak from the outside composed of an accurate pair of pres- environment into this evacuated region. sure transducers that would essentially
Antenna Calibration and Measurement Equipment
NASA’s Jet Propulsion Laboratory, Pasadena, California A document describes the Antenna pointing model generation require- factor of two. It improved the pointing Calibration & Measurement Equipment ments. Other data includes antenna performances of the DSN antennas and (ACME) system that will provide the subreflector focus, system noise temper- productivity of its personnel and calibra- Deep Space Network (DSN) with instru- ature and tipping curves, antenna effi- tion engineers. mentation enabling a trained RF engi- ciency, reports system linearity, and in- This work was done by David J. Rochblatt neer at each complex to perform an- strument calibration. and Manuel Vazquez Cortes of Caltech for tenna calibration measurements and to The ACME system design is based on NASA’s Jet Propulsion Laboratory. Further generate antenna calibration data. This the on-the-fly (OTF) mapping tech- information is contained in a TSP (see page data includes continuous-scan auto- nique and architecture. ACME has con- 1). NPO-47599 bore-based data acquisition with all-sky tributed to the improved RF perform- data gathering in support of 4th order ance of the DSN by approximately a