No abstract

A NARX neural network is adapted for cylinder pressure trace reconstruction on a multicylinder engine. Following a systematic study to establish the required NARX input information (using measured pressure traces and simulated crank kinematics), two fully recurrent training algorithms are developed and applied to real engine data. These include a back-propagation-through-time algorithm (BPTT) and an extended Kalman filter (EKF). For multi-cylinder engines, two cases are examined, both assuming crank kinematics is obtained from a single shaft-encoder fitted at the forward end of the crankshaft. In one case, a NARX model is constructed to provide an inverse relationship between the kinematics at the encoder location and the pressure trace in an arbitrary cylinder. In the second case, by transforming the kinematics (to emulate a local encoder), a different NARX model is constructed to relate the kinematics at the crank location of a particular cylinder to the corresponding pressure trace. The accuracy and efficiency of both NARX models is examined for application to a three-cylinder in-line DISI engine (in which pressure traces are measured on all cylinders). The paper shows that the computational requirements of training are substantial and, although the efficiency of the EKF algorithm is better than the BPTT, the fitting accuracies are similarly good. For generalization, however (to unseen data), neither method is yet sufficiently accurate (even for steady state engine operation) unless substantially more training data are used to achieve the target accuracy of ± 4 per cent. The overall conclusion of the paper is that the NARX model has the correct architecture for multicylinder pressure reconstruction.

A two-degree-of-freedom dynamic model is constructed to simulate the instantaneous crank kinematics and total mechanical losses arising in a multicylinder gasoline engine coupled to a dynamometer. The simulation model is driven using specified cylinder gas pressures, and loaded by nominal brake torque and total friction losses. Existing semi-empirical torque loss models (based on calibrated single-cylinder diesel engine data) are used to account for the instantaneous friction losses in the piston-ring assembly, in bearings, and in auxiliaries. The model is specialized to the simulation of crank kinematics and matched brake torque for a three-cylinder in-line direct injection spark ignition (DISI) engine, without a gearbox. This allows the total friction loss to be separated from the brake torque for an engine not fitted with the very large number of sensors otherwise needed to calibrate analytical friction models. An equivalent simulation model is also constructed using GT-Crank, which excludes explicit reference to friction. In using both models to simulate steady state operation at a specified mean engine speed, the output torque is matched by iteration. The GT-Crank model necessarily compensates for internal losses by exaggerating the total output torque. Both simulation models are compared with measured crank kinematics and brake torque obtained from a dynamometer-loaded I3 DISI engine. The paper shows that by comparing the matched output torque from simulation with the measured output torque from the engine, the proposed model gives a very good high-speed prediction of the total mechanical losses. At low speed, the instantaneous model is still not accurate. It is also shown, however, that apart from the no-load condition, use of an average torque to compensate for friction (as in GT-Crank) is wholly acceptable for simulating instantaneous crank kinematics. This is the first reported instance of a simulation model (which includes the particular form of semi-empirical friction loading) being comprehensively compared and verified using multicylinder DISI engine data.

We investigate the potential of geospatiotemporal data mining of multi-year land surface phenology data (250 m Normalized Difference Vegetation Index (NDVI) values derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) in this study) for the conterminous United States as part of an early warning system to identify threats to forest ecosystems. Cluster analysis of this massive data set, using high-performance computing, provides a basis for several possible approaches to defining the bounds of "normal" phenological patterns, indicating healthy vegetation in a given geographic location. We demonstrate the applicability of such an approach, using it to identify areas in Colorado, USA, where an ongoing mountain pine beetle outbreak has caused significant tree mortality.Index Terms-Remote sensing, data mining, phenology, cluster analysis, high-performance computing THE FOREST INCIDENCE RECOGNITION AND STATE TRACKING (FIRST) SYSTEMEarly identification of forested areas under attack from insects or disease can enable timely response to protect forest ecosystems from long-term or irreversible damage. Unfortunately, given the sheer size of the United States and limited resources of agencies such as the USDA Forest Service to conduct aerial surveys and ground-based inspections, many threats go unnoticed until a great deal of damage has already been done. To improve threat detection, the USDA Forest Service, in partnership with Oak Ridge National Laboratory and the NASA Stennis Space Center, is developing The Forest Incidence Recognition and State Tracking (FIRST) early warning system. FIRST will detect and monitor threats to forests and wildlands in the conterminous United States (CONUS) as part of a two tier system: An early warning system that monitors continental-scale areas at a moderate resolution using remote sensing data to spatially direct and focus efforts of the second tier, consisting of higher resolution monitoring through airborne overflights-called Aerial Detection Survey (ADS) sketch-mapping-and ground-based inspections. Tier 2 is largely in operation today, but the strategic direction provided by the FIRST system in Tier 1 will improve the efficiency and utility of these costly and labor-intensive surveys. * forrest@climatemodeling.org † rmills@ornl.gov ‡ jkumar@climatemodeling.org § shivakar@climatemodeling.org ¶ hnw@geobabble.orgThe goals of the FIRST early warning system are to provide a single, unified system for change detection from remotely sensed vegetation properties through time over the domain of the conterminous United States at about 250 m nominal resolution-obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensors on board NASA's Terra and Aqua satellites-at frequent intervals, on the order of one week. The system must be automated, requiring unsupervised data mining methods, and provide results as close to real-time as possible. It must "learn" or improve its prognostic ability utilizing a library of previous experiences, including both true and ...

No abstract