Abstract: According to aspects of the disclosure, a method is provided comprising: executing a set of threads in a storage system, the set of threads including at least a first thread; executing a plurality of performance counters of the storage system, the plurality of performance counters including at least: (i) a first performance counter that is executed when an operating state of the first thread is changed in response to the first thread accessing a synchronization object, and (ii) a second performance counter that is executed when a marker inserted in the first thread is executed; generating one or more performance data containers associated the first thread based on performance data associated with the first thread; and generating a directed graph based on the performance data containers.

Abstract: This document presents a system and method for using an intelligent belt and a remote secondary sensor to obtain instantaneous feedback during the performance of physical activities. A remote secondary sensor may be attached to a conveyance associated with, and in use by, a user. The intelligent belt uses a position sensor, such as a 3-axis accelerometer, to monitor the movements of a user and a conveyance such as an athletic apparatus, or any other apparatus upon which a user may stand, move, or with which the user may be associated. The belt is capable of determining when certain physical activities are not being performed properly in comparison to a preferred and pre-configured performance, and can provide real-time feedback using visual, aural, haptic, and other feedback mechanisms.

Abstract: The presentation of sensor network information can be provided to a user via a customized dashboard web interface. The dashboard web interface can be based on customized transformation and alert functions that can be applied to one or more sensor channels of data produced from sensors at a monitored location. The customized alerts can provide real-time monitoring of targeted conditions at the monitored location.

Abstract: A database search apparatus includes a processor that receives a search query including first keys. The processor calculates a first score based on statistical information. The first score represents search performance of searching a first database in accordance with the search query. The first database stores a data group including combinations of a key and a value of the key. The statistical information includes information regarding each key included in the data group. The processor calculates a second score based on the statistical information. The second score represents search performance of searching a second database different from the first database in accordance with the search query, the second database storing the data group. The processor selects the first database or the second database based on the first score and the second score. The processor instructs the selected database to performing a search in accordance with the search query.

Abstract: An example method comprises receiving historical sensor data from sensors of components of wind turbines, training a set of models to predict faults for each component using the historical sensor data, each model of a set having different observation time windows and lead time windows, evaluating each model of a set using standardized metrics, comparing evaluations of each model of a set to select a model with preferred lead time and accuracy, receive current sensor data from the sensors of the components, apply the selected model(s) to the current sensor data to generate a component failure prediction, compare the component failure prediction to a threshold, and generate an alert and report based on the comparison to the threshold.

Abstract: Methods, systems, apparatus, and non-transitory computer readable media are described for conducting collaborative real-time inspections of insured properties. The system may include several workstations, each operated by a different user with a respective role in processing an insurance claim. The system also may include a server system configured to receive an indication of an insured property to be inspected by users at the workstations using an unmanned aerial vehicle (UAV) configured to collect aerial imagery, provide, substantially in real time during an inspection session, aerial imagery of the insured property to each of the workstations, receive, substantially in real time during the inspection session, information related to the insured property being inspected from the workstations, and generate a database record descriptive of the inspection session. The aerial imagery may be used to verify dimensions of insured homes or estimate home damage resulting from fire, wind, hail, water, or hurricanes.

Abstract: An apparatus includes a first function module providing a master signal, a second function module providing a comparison signal, and safety logic. The safety logic includes a toggle signal generator having a comparator providing a comparison result in response to the master signal and the comparison signal, a feedback path generating a first toggle signal in response to the comparison result and providing a feedback signal to the comparator, and a first multiple input gate generating a second toggle signal in response to the comparison result. The safety logic also includes a toggle signal monitor providing a final fault search signal in response to the first toggle signal and the second toggle signal.

Abstract: A system for enabling and disabling operation of manufacturing machines provides a manufacturing machine user interface that facilitates receiving quality control information regarding the manufacturing machine from a user. For quality assurance purposes, the computer system of the manufacturing machine may decide whether to enable operation of the manufacturing machine based on the received quality control information. The computer system of the manufacturing machine may also decide to disable operation of the manufacturing machine if the quality control information provided is incomplete, out of date, or otherwise insufficient to indicate the manufacturing machine is ready for safe and effective operation.

Abstract: A big data analytics system includes an event monitoring controller that detects a user targeted event from an initial digital query data flow. An event conversion controller determines at least one contextual element associated with the at least one user targeted event, and automatically converts the at least one contextual element into a search criteria. An external data query builder controller automatically executes an external search query based on the search criteria. A data fusing controller automatically fuses the at least one contextual element generated from the at least one user targeted event, together with external search results generated from the external search query and the initial query data flow to generate a fused data file.

Abstract: Performance of a storage system with data distributed substantially, evenly across a cluster of storage nodes can be dynamically managed according whether quality of service (QoS) rules based on one or more performance capacities according to performance class is satisfied, wherein each QoS rule includes a boundary condition and a threshold expressing performance guarantees to a respective performance class. Determining for each QoS rule, whether a respective boundary condition satisfies a respective threshold based on measured capacity for a performance metric corresponding to the respective boundary condition; and if so, setting a first QoS parameter associated with the set of QoS rules and throttling access to the storage system by a first client associated with a first performance class based on the first QoS parameter to enforce the performance guarantees to the first performance class.

Abstract: A computer device determines requirements of a particular device/service by analyzing designated electronic communications. The computer device then monitors performance of the device/service over a period of time to determine performance of the device/service in meeting those derived requirements. An output regarding the performance of the device/service in meeting those requirements is then generated.

Abstract: A system and method for measuring dynamic properties of a structure, and for using the measured dynamic properties to assess the dynamic performance of the structure. The system and method measures dynamic properties of the structure such as frequencies of resonance, mode shapes, and non-linear damping, and uses them in an analysis of the structure to compare the dynamic response of the structure with the anticipated properties of a structure built according to applicable building code requirements. The system and method thus quantifies a risk of failure of the structure by determining a risk ratio that compares an as-is condition of the structure with an as-designed condition of the structure.

Abstract: A determination method is for an exhaust gas treatment device mounted in a work machine and having at least one of a diesel oxidation catalyst and a catalyzed soot filter. The determination method includes acquiring heat damage information, acquiring cumulative heat damage information, and determining a usability of the exhaust gas treatment device based on the cumulative heat damage information. The heat damage information indicates an extent of heat damage of the exhaust gas treatment device based on a unique identification symbol of the exhaust gas treatment device. The cumulative heat damage information is acquired by accumulating the heat damage information.

Abstract: The present specification relates to a power grid electric transient detector, a method of detecting a power grid electric transient, and an electric provisioning system detecting the power grid electric load. The power grid electric transient is estimated by receiving an electric parameter for electricity received from a power grid at a consuming facility. A digital filter extracts transient of the electric parameter. A processor analyzes the transients of the electric parameter and generates a power grid performance index based on the analyzed transient of the electric parameter.

Abstract: A thermal energy system comprising a first thermal system having a heating demand, and a heat source connection system coupled to the first thermal system, the heat source connection system being adapted to provide selective connection to a plurality of heat sources for heating the first thermal system, the heat source connection system comprising a first heat exchanger system coupled to a first remote heat source containing a working fluid and a second heat exchanger system adapted to be coupled to ambient air as a second heat source, a fluid loop interconnecting the first thermal system, the first heat exchanger system and the second heat exchanger system, at least one mechanism for selectively altering the order of the first heat exchanger system and the second heat exchanger system in relation to a fluid flow direction around the fluid loop, and a controller for actuating the at least one mechanism.

Abstract: An in-depth behavioral analysis of time-series metric data can require time consuming and computationally expensive statistical processing. To provide a more efficient behavioral analysis of a component within an environment, an application behavioral analysis system captures time-series metric data for a component, such as an application, and generates tiles comprising metric values from sequential segments of the metric data. After generating the tiles, the system iterates through the tiles of each segment and determines whether the segment shares matching tiles with other segments of the metric data. The system generates a behavioral analysis for the component based on the segments with shared tiles. The behavioral analysis may include a matrix or graph data structure which indicates the related segments. From this behavioral analysis, the system can efficiently identify common and reoccurring behaviors for a component without performing complex statistical processing of the metric data.

Abstract: An apparatus and method for identification of air leaks in a compressed air system are presented. The method includes receiving system data including current samples of air compressors of the compressed air system from at least one self-powered power sensor; aggregating the system data; dividing the aggregated system data into a plurality of clusters; merging overlapping clusters to reduce the number of clusters to two clusters, wherein a first cluster represents a baseload current and a second cluster represents a full-load current; determining the baseload current to full-load current ratio and comparing the ratio to a predetermined standard; and determining if an air leak is present if the ratio is below the predetermined standard.

Abstract: Systems and methods for shadow testing are provided. A new service may be brought online that corresponds to an existing service. The new service may be designated for shadow testing. A call may be received for the existing service from a calling device. The received call may be provided to the existing service, while an asynchronous call corresponding to the received call may be provided to the new service. A result may be received from the existing service in response to the provided call, and a result may be received from the new service in response to the asynchronous call. The result received from the existing service may be used in responding to the received call. Data regarding the result received from the new service may be logged, but the result received from the new service may be excluded from use in responding to the received call. One or more metrics evaluating performance by the new service may be generated.

Abstract: The present invention relates to a control method and a wind turbine configured to determine a load signal of at least one component of the wind turbine, and to calculate a damage rate based on this load signal. The control method calculates and monitors the damage rate in real-time, wherein the damage rate is normalised by using a first function defining a first transition phase. A second function is afterwards applied to the normalised damage rate which defines a second transition phase. These transition phases allows for a smooth transition between different operating modes of the wind turbine. The control method may further change the power output of the wind turbine relative to the nominal power output when the output signal of the second function is determined to be stable over at least one time period.

Abstract: Examples herein disclose an apparatus. The apparatus includes a network interface controller (NIC) port to be dedicated to a management functionality of a server. The apparatus also includes a light emitting diode (LED), coupled to the NIC port, to provide a visible indication that the NIC port is dedicated to the management functionality of the server.

Abstract: Apparatus and methods are provided for determining, recommending, and applying a calibration parameter to collected activity data. In one embodiment, calibration parameter is estimated based on physical aspects of the user and automatically applied to collected data. In another embodiment, the calibration parameter is determined based on secondary data which is more precise than the data which is collected. The calibration factor based on the more precise data may comprise a recommended calibration factor, yet the user may be enabled to select any calibration factor he/she prefers via an interactive display. In one specific variant, the activity comprises a walk or run activity of the user, and the calibration parameter comprises the user's specific stride length. In another variant, the user selects a calibration factor by reviewing a list of previous activity against that same activity after calibration given a particular calibration factor is applied.

Abstract: In one embodiment, a heterogeneous multicore processor is described that is optimized to execute multi-stage computer vision algorithms such as cascade classifier workloads. In such embodiment the heterogeneous processor includes at least one SIMD core, such as a vector processor core, coupled with one or more scalar cores. In one embodiment the heterogeneous multiprocessor executes multi-stage compute operations, where the SIMD core computes a first set of stages and the one or more scalar cores compute the second set of stages. In one embodiment, a process for designing a heterogeneous multicore processor is disclosed which optimizes the ratio of scalar to SIMD cores based on execution time of the multi-stage compute operation in relation to processor die area consumed by a processor configuration having the ratio.

Abstract: A tool strip mark generator capable of measuring applied force includes: a bottom surface on which at least a portion of a target surface made of a material capable of generating a tool strip mark, which is a tool-scraped stripe, is formed; a rotating member configured to be rotatable in a state of being spaced apart from the bottom surface; a moving member movable on the bottom surface by laterally coupling the rotating member to a coupling portion; a force sensor attached to one side of an outer peripheral surface of the rotating member and sensing force which is applied when a tool strip mark is generated on the target surface by inserting the tool; a measurer configured to measure an output signal of the force sensor and calculate information about the applied force; and a display configured to display information about the applied force calculated by the measurer.

Abstract: Apparatus and a method for controlling a network from a layer 2 node receives multiple latency measurement histograms from the layer 1 nodes and aggregates corresponding bins of the latency measurement histograms to generate an aggregated histogram. Each corresponding bin of the multiple latency measurement histograms and of the aggregated histogram represents a respective range of latency measurement values. The apparatus and method determine from the aggregated histogram, a range of latency measurement values matching a predetermined percentile and transmit control information to at least one network device responsive to the determined range of latency measurement values being greater than a predetermined latency value.

Abstract: This data conversion system is characterized by including: a storage unit that stores a column including a plurality of data elements; a range specification module that specifies the range of each of the data elements of the column; an information amount evaluation module that calculates the information amount of the data element within the specified range of the column; and a change point detection module that detects a point at which a change in the information amount according to a change in the specified range satisfies a predetermined condition.

Abstract: A data collection system of at least one wind turbine power generating apparatus includes: at least one wind turbine controller including a control part for controlling each wind turbine power generating apparatus, a representative data memory part for storing representative data of at least one parameter indicating a state or operation history of the at least one wind turbine power generating apparatus, and a representative data selection part configured to obtain data of the at least one parameter once in every control calculation cycle and determine whether to store the obtained data as the representative data for every control calculation cycle; and a data collection part configured to collect, from each wind turbine controller, the representative data of the at least one parameter related to the state of each wind turbine power generating apparatus stored in the representative data memory part of each wind turbine controller.

Abstract: A method is provided for supporting maintenance of a fleet of vehicles with intuitive display of repair analytics. The method includes receiving a user request for repair analytics for the fleet of vehicles via a GUI that includes a first frame with a date-range filter to enable the user to select a date range for the repair analytics. The method includes interpreting the user request to produce a query of dataset(s) for the fleet, and executing the query for data of the vehicles. The datasets include an in-service time and repair costs for the vehicles, and the data includes the same over the date range. The method includes determining, and generating a chart in a second frame of the GUI that graphically illustrates, a moving average repair cost per unit of in-service time for a plurality of time periods within the date range from the in-service time and repair costs.

Abstract: The human machine interface (HMI) of a vehicle, especially an automobile is responsive to outside stimuli. Signals relevant to the amount of concentration required by a driver are received by the HMI, which determines, based on the values of these signals, a driver distraction level. As the driver distraction level increases, areas of a structured display on a display screen are dynamically removed, to present the driver with a reduced distractions when greater concentration is required.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for improving the connectivity and response speed throughout a network. In one aspect, a method includes receiving network traffic data from an aggregation point on a network, the network traffic data having been sent to or received from one of a plurality of end-point devices on the network, calculating performance metrics for each of the end-point devices based the received network traffic data, for each of the end-point devices, comparing the performance metrics to respective threshold values to determine performance issues for the network, wherein the threshold values are determined based on historical network data, correlating the determined performance issues for the network to an aspect of the end-point devices, and implementing an action to correct the determined performance issues for the network based on the aspect of the at least one end-point devices.

Abstract: A wind turbine generator fault detection method is described. The method includes obtaining a first signal from a generator of a wind turbine and a second signal from a vibration sensor coupled to the wind turbine, the first signal representing an output current of the generator, and the second signal being a time-sampled signal representing vibrations of a bearing in the wind turbine. Determining a shaft rotation frequency signal from the first signal, the shaft rotation frequency signal representing a time-varying rotational speed of a shaft of the wind turbine. Resampling an envelope of the second signal based on the shaft rotation frequency signal to provide a third signal, the third signal being an angular sampled signal. Detecting, by the at least one processor, a fault in the bearing of the wind turbine by identifying a characteristic signature of a bearing fault in the third signal.

Abstract: An information measuring device includes a sensor that obtains information on a phenomenon periodically occurring on an apparatus while being synchronized with the steady repetition of an operation and a control device. The control device associates the obtained information with an operation phase in the operation so as to generate change in the phenomenon in one cycle based on the information obtained during the operation of several times. Meanwhile, an operation cycle where the operation is repeated in the apparatus, a sampling cycle, and a duration where the operation is steadily repeated in the apparatus are included in measurement parameters. The control device calculates an optimum measurement parameter from a relation between the measurement parameters on at least one of the measurement parameters and changes the measurement parameter as a target to the optimum measurement parameter.

Abstract: A state identification device includes: a feature amount extraction unit that extracts a feature amount from physical data of a machine tool unit which shifts to a plurality of operation states; a feature amount storage unit in which a region is provided as a storage destination of the extracted feature amount for each of the plurality of operation states to which the machine tool unit shifts; a similarity calculation unit that calculates a similarity between the extracted feature amount and a feature amount stored in the feature amount storage unit; a state determination unit that determines an operation state of the machine tool unit based on the calculated similarity; and a storage destination determination unit that determines the region serving as the storage destination of the extracted feature amount based on a determination result of the state determination unit.

Abstract: Methods, devices, and systems for authenticating a component are disclosed herein. One such component can be a control device. This control device can include a first button and a control device fitting. When activated, the first button can be configured to cause a first control signal to be output from the control device to a controllable component. The control device fitting can include a connection device having a connector element and a conductive authentication element. The connector element can be configured to connect the control device to the controllable component. The conductive authentication element can be configured to allow the controllable component to authenticate the control device for use with the controllable component.

Abstract: The present disclosure describes systems, methods, and computer readable media that provide a hybrid approach that uses machine learning techniques and phenomenological reactor models for optimization of steam cracker units. While the phenomenological model allows capturing the physics of a steam cracker using molecular kinetics, the machine learning methods fill the gap between the phenomenological models and more detailed radical kinetics based steam cracker models. Also, machine learning based models can capture actual plant information and provide insight into the variation between the models and plant running conditions. The proposed methodology shows better interpolation and extrapolation capabilities as compared to stand-alone machine learning methods. Also, compared to detailed radical kinetics based models, the approach utilized in embodiments requires much less computational time in order to carry out whole plant-wide optimization or can be used for planning/scheduling purposes.

Abstract: The systems and methods disclosed herein include an assessment system and process for assessing an industrial machine and its various sections, sub-sections, and parts. In embodiments, the assessment system includes an assessment device that includes an assessment overviewer and a part assessor. The assessment overviewer provides an assessor with a selectable industrial machine schematic that illustrates the assessment status of each section of the industrial machine, such as by changing the appearance of each section based its assessment status. The part assessor provides a part assessment interface with engineering instructions and a part grading user interface that provides for a comparison of digital images to complete the part grading.

Abstract: Disclosed herein are system, method, and computer program product embodiments for automating component management in enterprise applications. An embodiment operates by receiving metadata associated with the enterprise application implementation and storing an inventory including at least a portion of the metadata. The system then determines one or more component dependencies of the enterprise application implementation based on the inventory and providing one or more recommendations for component installation or deletion based on the inventory and one or more component dependencies. The system also generates one or more testcases based on the inventory and the one or more component dependencies.

Abstract: A system and method is for production control, and comprises an electronic device, and one or more network connectors associated with the electronic device configured to connect to a server computer. The server computer contains a database of one or more work instructions that are downloadable to the electronic device through the network connector. The one or more work instructions are configured to provide operation instructions for one or more tools and the one or more network connectors are further configured to provide work instructions to the one or more tools.

Abstract: Certain aspects and features include a system and method for validating fundamental-only power measurements. A meter testing application calculates an expected registration for a meter under test by comparing a fundamental demand indicating an amount of power expected to be delivered by the fundamental component to a known load with a wideband demand indicating an amount of power expected to be delivered by the wideband voltage to the known load. The application further measures, from the meter under test, a measured registration by applying the wideband voltage component and the known load to the meter under test, which is configured to measure power consumption from the fundamental component only. The application calculates based on the expected registration and the measured registration, an error for the meter. When the error is outside a tolerance, the application identifies the meter under test as non-compliant.

Abstract: A computing system may involve a time-series server device and computing devices. The time-series server device may be configured to: receive and store pre-defined trigger configurations; receive and store time-series data, wherein the pre-defined trigger configurations define states and/or state transitions for the received time-series data; apply, by way of a trigger engine, the pre-defined trigger configurations to the received time-series data to determine observed states and/or state transitions in the time-series data; and store, in transition storage, representations of the observed states and/or state transitions. One or more applications operating on computing devices may be configured to: transmit the pre-defined trigger configurations to the time-series server; transmit a stream of the time-series data to the time-series server; and repeatedly poll and receive, by way of a plurality of worker threads, the representations of the observed states and/or state transitions from the transition storage.

Abstract: According to the present disclosure, a sensor controller configured to collect a variety of machine related operations data may receive operation data from at least one measurement sensor connected to a machine and configured to detect the operation data of the machine and transfer the operation data to a communication module to be transmitted to and collected by a server. Even if a different measurement sensor is connected to a sensor controller, the operation data of the connected measurement sensor can be still recognized.

Abstract: The present invention is directed to a system and process for monitoring jet fuel quality control procedural compliance. One system of the current invention includes a portable computer in communication with a server over network, an equipment database, and a report module. In exemplary process, jet fuel equipment is input into the system and stored in the equipment database, along with the process for its inspection. Inspector profiles are input into the system. The system facilitates notification of required inspections for a facility. The system presents an interface guiding an inspector through inspection of jet fuel and jet fuel equipment inspections. The input is stored by the system, whereby the report module generate reports based on inspection reports, equipment, and facilities.

Abstract: A measurement result management apparatus includes one or more memories, and one or more processors coupled to the one or more memories and the one or more processors configured to acquire, from a plurality of devices worn by a plurality of users, location information indicating locations of the plurality of devices, perform, in response to acquisition of a first measurement result including a first measurement time from a first measurement apparatus, a selection of a first device from the plurality of devices in accordance with a location of the first measurement apparatus, the first measurement time, and the location information, and store the first measurement result in association with the selected first device in the one or more memories.

Abstract: Disclosed are systems, methods, and computer-readable media for assuring tenant forwarding in a network environment. Network assurance can be determined in layer 1, layer 2 and layer 3 of the networked environment including, internal-internal (e.g., inter-fabric) forwarding and internal-external (e.g., outside the fabric) forwarding in the networked environment.

Abstract: A system according to the present disclosure includes an identification module, a data recording module, and a data upload module. The identification module is configured to identify at least one of a vehicle and a user of the vehicle. The data recording module is configured to record a location of the vehicle, an acceleration of the vehicle, and data received from a controller area network (CAN) bus of the vehicle during a driving session. The data upload module is configured to upload the vehicle location, the vehicle acceleration, the CAN bus data, and at least one of the vehicle identification and the user identification to a remote server.

Abstract: A performance management system includes: an information collection function unit configured to collect database utilization information indicating a utilization status of the database and component utilization information indicating a utilization status of components, which are constituent elements of the information system; a related component calculation function unit configured to acquire a component utilization, which is a proportion of an actual usage of the component to a maximum usage of the component, on the basis of the component utilization information and specify a related component related to the performance of the information system among the components on the basis of the component utilization and the database utilization information; and a prediction function unit configured to predict a future performance of the information system on the basis of utilization information of the related component.

Abstract: An athlete tracking system and method for tracking an athlete during training sessions. The system includes a mobile tracking device wearable around a portion of the athlete's head, and a training log server for maintaining an athlete profile that includes a training record updatable by the athlete and a training plan supplied by a coach.

Abstract: A control device includes a processor that is configured to execute computer-executable instructions so as to control driving of a robot capable of performing work including a screw-tightening process for tightening a screw, wherein the processor is configured to: receive an input of at least one of characteristics of an object including the screw used in the screw-tightening process; calculate, on the basis of the characteristics received, a value concerning screw-tightening torque at a time of the tightening of the screw by the robot; and cause a display to display the value concerning the screw-tightening torque.

Abstract: A system and method is presented for efficiently enforcing control limits in the calculation of powerflow studies of electrical power systems, using the Holomorphic Embedding Load-flow Method. The method applies to any automatic control in which the controlling resource is limited by a maximum or minimum value, such as Mvar limits in generators, the tap ratio in load-changing transformers, or real power output in AGC-participating generators. One key element of this invention consists in devising an equality constraint, holomorphically embedded, able to encode the switching behavior between the different control modes when limit thresholds are reached. The other key element is a novel analytic continuation scheme, here referred to as “Padé-Weierstrass”, which achieves the continuation to s=1 by means of several intermediate steps of power series re-expansions on the path from s=0 to s=1 in embedding parameter space.

Abstract: A method analyzing a machine having a rotating shaft includes generating an electric measurement signal dependent on mechanical vibrations from the shaft rotation; sampling the measurement signal to generate a digital measurement data signal; performing a decimation of the digital measurement data signal to achieve a digital signal having a reduced sampling frequency, where the decimation includes controlling the reduced sampling frequency such that the number of sample values per revolution of the shaft is kept at a substantially constant value, and receiving the digital signal at an enhancer input performing a correlation in the enhancer so as to produce an output signal sequence where repetitive signals amplitude components are amplified in relation to stochastic signal components, and performing a condition analysis for analyzing the condition of the machine dependent on the digital signal having a reduced sampling frequency.

Abstract: One variation of a method for predicting manufacturing defects includes: accessing a set of inspection images of a set of assembly units recorded by an optical inspection station; for each inspection image in the set of inspection images, detecting a set of features in the inspection image and generating a vector representing the set of features in a multi-dimensional feature space; grouping neighboring vectors in the multi-dimensional feature space into a set of vector groups; and, in response to receipt of a first inspection result indicting a defect in a first assembly unit, in the set of assembly units, associated with a first vector in a first vector group, in the set of vector groups, labeling the first vector group with the defect and flagging a second assembly unit associated with a second vector, in the first vector group, as exhibiting characteristics of the defect.