Abstract: We disclose a system that optimizes deployment of sensors in a computer system. During operation, the system generates a training data set by gathering a set of n signals from n sensors in the computer system during operation of the computer system. Next, the system uses an inferential model to replace one or more signals in the set of n signals with corresponding virtual signals, wherein the virtual signals are computed based on cross-correlations with unreplaced remaining signals in the set of n signals. Finally, the system generates a design for an optimized version of the computer system, which includes sensors for the remaining signals, but does not include sensors for the replaced signals. During operation, the optimized version of the computer system: computes the virtual signals from the remaining signals; and uses the virtual signals and the remaining signals while performing prognostic pattern-recognition operations to detect incipient anomalies that arise during execution.

Abstract: Systems, methods, and other embodiments associated with autonomous cloud-node scoping for big-data machine learning use cases are described. In some example embodiments, an automated scoping tool, method, and system are presented that, for each of multiple combinations of parameter values, (i) set a combination of parameter values describing a usage scenario, (ii) execute a machine learning application according to the combination of parameter values on a target cloud environment, and (iii) measure the computational cost for the execution of the machine learning application. A recommendation regarding configuration of central processing unit(s), graphics processing unit(s), and memory for the target cloud environment to execute the machine learning application is generated based on the measured computational costs.

Abstract: We describe a system that performs prognostic-surveillance operations based on an inferential model that dynamically adapts to evolving operational characteristics of a monitored asset. During a surveillance mode, the system receives a set of time-series signals gathered from sensors in the monitored asset. Next, the system uses an inferential model to generate estimated values for the set of time-series signals, and then performs a pairwise differencing operation between actual values and the estimated values for the set of time-series signals to produce residuals. Next, the system performs a sequential probability ratio test (SPRT) on the residuals to produce SPRT alarms. When a tripping frequency of the SPRT alarms exceeds a threshold value, which is indicative of an incipient anomaly in the monitored asset, the system triggers an alert. While the prognostic-surveillance system is operating in the surveillance mode, the system incrementally updates the inferential model based on the time-series signals.

Abstract: The disclosed embodiments relate to a system that facilitates deployment of utility repair crews to nodes in a utility network. During operation, the system determines a node criticality for each node in the utility network based on a network-reliability analysis, which considers interconnections among the nodes in the utility network. The system also determines a node failure probability for each node in the utility network based on historical weather data, historical node failure data and weather forecast information for the upcoming weather event. The system uses the determined node criticalities and the determined node failure probabilities to determine a deployment plan for deploying repair crews to nodes in the utility network in preparation for the upcoming weather event. The system then presents the deployment plan to a person who uses the deployment plan to deploy repair crews to be available to service nodes in the utility network.

Abstract: The disclosed embodiments relate to a system that camouflages electromagnetic interference (EMI) fingerprints in EMI emissions from a computing system to enhance system security. During operation, the system monitors the EMI emissions from the computer system while the computer system is operating to produce corresponding EMI signals. Next, the system performs a Fast Fourier Transform (FFT) operation on the EMI signals. The system then converts an output of the FFT operation into a frequency-domain representation of the EMI signals. Next, the system generates a camouflaging signal based on the frequency-domain representation of the EMI signals. Finally, the system outputs the camouflaging signal through a transmitter to camouflage EMI fingerprints in the EMI emissions from the computer system.

Abstract: The disclosed embodiments provide a system that detects execution of malicious cryptomining software in a target computing system. During operation, the system monitors target electromagnetic interference (EMI) signals generated during operation of the target computing system. Next, the system generates a target EMI fingerprint from the target EMI signals. The system then compares the target EMI fingerprint against a set of malicious EMI fingerprints for different pieces of malicious cryptomining software to determine whether the target computing system is executing malicious cryptomining software.

Abstract: The disclosed embodiments provide a system that generates a reference EMI fingerprint to be used in detecting unwanted electronic components in a target asset. During operation, the system gathers reference EMI signals generated by a reference asset while the reference asset is executing a periodic workload, wherein the reference asset is of the same type as the target asset and is certified not to contain unwanted electronic components. Next, the system divides the reference EMI signals into a set of profiles, which comprise EMI signals for non-overlapping time intervals of a fixed size. The system then temporally aligns and merges profiles in the set of profiles to produce a reference profile. Next, the system generates the reference EMI fingerprint from the reference profile. Finally, the system compares a target EMI fingerprint for the target asset against the reference EMI fingerprint to determine whether the target asset contains unwanted electronic components.

Abstract: During operation, the system receives time-series signals gathered from sensors in a utility system asset. Next, the system uses an inferential model to generate estimated values for the time-series signals, and performs a pairwise differencing operation between actual values and the estimated values for the time-series signals to produce residuals. The system then performs a sequential probability ratio test (SPRT) on the residuals to produce SPRT alarms. Next, the system applies an irrelevance filter to the SPRT alarms to produce filtered SPRT alarms, wherein the irrelevance filter removes SPRT alarms for signals that are uncorrelated with previous failures of similar utility system assets. The system then uses a logistic-regression model to compute an RUL-based risk index for the utility system asset based on the filtered SPRT alarms. When the risk index exceeds a threshold, the system generates a notification indicating that the utility system asset needs to be replaced.

Abstract: The disclosed embodiments relate to a system that performs low-temperature desalination. During operation, the system feeds cold saline water through a liquid-cooling system in a computer data center, wherein the cold saline water is used as a coolant, thereby causing the cold saline water to become heated saline water. Next, the system feeds the heated saline water into a vacuum evaporator comprising a water column having a headspace, which is under a negative pressure due to gravity pulling on the heated saline water in the water column. This negative pressure facilitates evaporation of the heated saline water to form water vapor. Finally, the system directs the water vapor through a condenser, which condenses the water vapor to produce desalinated water.

Abstract: During operation, the system receives a set of input signals containing electrical usage data from a set of smart meters, which gather electrical usage data from customers of the utility system. The system uses the set of input signals and a projection technique to produce projected loadshapes, which are associated with electricity usage in the utility system. Next, the system identifies a closest time period in a database containing recent empirically obtained load-related parameters for the utility system, wherein the load-related parameters in the closest time period are closest to a present set of load-related parameters for the utility system. The system then iteratively adjusts the projected loadshapes based on changes indicated by the load-related parameters in the closest time period until a magnitude of adjustments falls below a threshold. Finally, the system predicts electricity demand for the utility system based on the projected loadshapes.

Abstract: During operation, the system obtains time-series sensor signals gathered from sensors in an asset during operation of the asset in an outdoor environment, wherein the time-series sensor signals include temperature signals. Next, the system produces thermally-compensated time-series sensor signals by performing a thermal-compensation operation on the temperature signals to compensate for variations in the temperature signals caused by dynamic variations in an ambient temperature of the outdoor environment. The system then trains a prognostic inferential model for a prognostic pattern-recognition system based on the thermally-compensated time-series sensor signals. During a surveillance mode for the prognostic pattern-recognition system, the system receives recently-generated time-series sensor signals from the asset, and performs a thermal-compensation operation on temperature signals in the recently-generated time-series sensor signals.

Abstract: The disclosed embodiments provide a system that detects unwanted electronic components in a target asset. During operation, the system generates a sinusoidal load for the target asset. Next, the system obtains target electromagnetic interference (EMI) signals by monitoring EMI signals generated by the target asset while the target asset is executing the sinusoidal load. The system then generates a target EMI fingerprint from the target EMI signals. Finally, the system compares the target EMI fingerprint against a reference EMI fingerprint for the target asset to determine whether the target asset contains unwanted electronic components.

Abstract: After sensors are placed at three or more non-collinear locations on a surface of the component, the system receives time-series signals from the sensors while the component operates on a representative workload. The system then defines one or more triangles on the surface of the component, wherein each triangle is defined by three vertices, which coincide with different sensor locations on the surface of the component. For each triangle, the system applies a barycentric coordinate technique (BCT) to time-series signals received from sensors located at the vertices of the triangle to determine a candidate location within the triangle to place an additional sensor. The system then compares the candidate locations for each of the one or more triangles to determine a globally optimal location for the additional sensor, and a new sensor is placed at this location. This process is repeated until a desired number of sensors are placed.

Abstract: During operation, the system receives time-series signals from sensors in the asset while the asset is operating. Next, the system obtains real-time environmental parameters for an environment in which the asset is operating. The system then selects an environment-specific inferential model for the asset based on the real-time environmental parameters, wherein the environment-specific inferential model was trained on a golden system while the golden system was operating in an environment that matches the real-time environmental parameters. Next, the system uses the environment-specific inferential model to generate estimated values for the received time-series signals based on correlations among the received time-series signals, and performs a pairwise-differencing operation between actual values and the estimated values for the received time-series signals to produce residuals. Finally, the system determines from the residuals whether the asset is operating correctly.

Abstract: The system receives a set of load signals from an archive that contains historic load information gathered at various locations throughout an electrical grid, which distributes electrical power for the utility system. Next, the system applies a first difference function to the set of load signals to produce a set of difference signals. The system then performs a spike-detection operation on the set of difference signals to identify pairs of positive-negative and negative-positive spikes, which identify gaps in the set of load signals associated with periods of network disruption. Next, the system modifies the set of load signals by filling in each identified gap with projected load values determined by performing a localized loadshape forecasting operation based on the continuous load values immediately preceding the identified gap. Finally, the system forecasts electricity demand for the utility system based on the modified set of load signals.

Abstract: The disclosed embodiments provide a system that performs seasonality-compensated prognostic-surveillance operations for an asset. During operation, the system obtains time-series sensor signals gathered from sensors in the asset during operation of the asset. Next, the system identifies seasonality modes in the time-series sensor signals. The system then determines frequencies and phase angles for the identified seasonality modes. Next, the system uses the determined frequencies and phase angles to filter out the seasonality modes from the time-series sensor signals to produce seasonality-compensated time-series sensor signals. The system then applies an inferential model to the seasonality-compensated time-series sensor signals to detect incipient anomalies that arise during operation of the asset. Finally, when an incipient anomaly is detected, the system generates a notification regarding the anomaly.

Abstract: The disclosed embodiments relate to a technique for training a prognostic pattern-recognition system to detect incipient anomalies that arise during execution of a computer system. During operation, the system gathers and stores telemetry data obtained from n sensors in the computer system during operation of the computer system. Next, the system uses the telemetry data gathered from the n sensors to train a baseline model for the prognostic pattern-recognition system. The prognostic pattern-recognition system then uses the baseline model in a surveillance mode to detect incipient anomalies that arise during execution of the computer system. The system also uses the stored telemetry data to train a set of additional models, wherein each additional model is trained to operate with one or more missing sensors. Finally, the system stores the additional models to be used in place of the baseline model when one or more sensors fail in the computer system.

Abstract: During operation, the system receives time-series signals from sensors in the asset while the asset is operating. Next, the system obtains real-time environmental parameters for an environment in which the asset is operating. The system then selects an environment-specific inferential model for the asset based on the real-time environmental parameters, wherein the environment-specific inferential model was trained on a golden system while the golden system was operating in an environment that matches the real-time environmental parameters. Next, the system uses the environment-specific inferential model to generate estimated values for the received time-series signals based on correlations among the received time-series signals, and performs a pairwise-differencing operation between actual values and the estimated values for the received time-series signals to produce residuals. Finally, the system determines from the residuals whether the asset is operating correctly.

Abstract: The system receives a set of present time-series signals gathered from sensors in the asset. Next, the system uses an inferential model to generate estimated values for the set of present time-series signals, and performs a pairwise differencing operation between actual values and the estimated values for the set of present time-series signals to produce residuals. The system then performs a sequential probability ratio test (SPRT) on the residuals to produce SPRT alarms with associated tripping frequency (TF). While the TF exceeds a TF threshold, the system iteratively adjusts sensitivity parameters for the SPRT to reduce the TF, and performs the SPRT again on the residuals. The system then uses a logistic regression model to compute a risk index for the asset based on the TF. If the risk index exceeds a threshold, the system generates a notification indicating that the asset needs to be replaced.

Abstract: The disclosed embodiments relate to a system that preprocesses sensor data to facilitate prognostic-surveillance operations. During operation, the system obtains training data from sensors in a monitored system during operation of the monitored system, wherein the training data comprises time-series data sampled from signals produced by the sensors. The system also obtains functional requirements for the prognostic-surveillance operations. Next, the system performs the prognostic-surveillance operations on the training data and determines whether the prognostic-surveillance operations meet the functional requirements when tested on non-training data. If the prognostic-surveillance operations do not meet the functional requirements, the system iteratively applies one or more preprocessing operations to the training data in order of increasing computational cost until the functional requirements are met.