Constraint-Guided Autoencoders to Enforce a Predefined Threshold on Anomaly Scores:An Application in Machine Condition Monitoring

Anomaly detection(AD)is an important task in a broad range of domains.A popular choice for AD are Deep Support Vector Data Description models.When learning such models,normal data is mapped close to and anomalous data is mapped far from a center,in some latent space,enabling the construction of a sphere to separate both types of data.Empirically,it was observed:(i)that the center and radius of such sphere largely depend on the training data and model initialization which leads to difficulties when selecting a threshold,and(ii)that the center and radius of this sphere strongly impact the model AD performance on unseen data.In this work,a more robust AD solution is proposed that(i)defines a sphere with a fixed radius and margin in some latent space and(ii)enforces the encoder,which maps the input to a latent space,to encode the normal data in a small sphere and the anomalous data outside a larger sphere,with the same center.Experimental results indicate that the proposed algorithm attains higher performance compared to alternatives,and that the difference in size of the two spheres has a minor impact on the performance.

Model Identification and Control of Electromagnetic Actuation in Continuous Casting Process With Improved Quality

This paper presents an original theoretical framework to model steel material properties in continuous casting line process. Specific properties arising from non-Newtonian dynamics are herein used to indicate the natural convergence of distributed parameter systems to fractional order transfer function models. Data driven identification from a real continuous casting line is used to identify model of the electromagnetic actuator device to control flow velocity of liquid steel. To ensure product specifications, a fractional order control is designed and validated on the system. A projection of the closed loop performance onto the quality assessment at end production line is also given in this paper.

Frequency‐to‐spectrum mapping GAN for semisupervised hyperspectral anomaly detection

Most unsupervised or semisupervised hyperspectral anomaly detection(HAD)methods train background reconstruction models in the original spectral domain.However,due to the noise and spatial resolution limitations,there may be a lack of discrimination between backgrounds and anomalies.This makes it easy for the autoencoder to capture the lowlevel features shared between the two,thereby increasing the difficulty of separating anomalies from the backgrounds,which runs counter to the purpose of HAD.To this end,the authors map the original spectrums to the fractional Fourier domain(FrFD)and reformulate it as a mapping task in which restoration errors are employed to distinguish background and anomaly.This study proposes a novel frequency‐to‐spectrum mapping generative adversarial network for HAD.Specifically,the depth separable features of backgrounds and anomalies are enhanced in the FrFD.Due to the semisupervised approach,FTSGAN needs to learn the embedded features of the backgrounds,thus mapping and restoring them from the FrFD to the original spectral domain.This strategy effectively prevents the model from focussing on the numerical equivalence of input and output,and restricts the ability of FTSGAN to restore anomalies.The comparison and analysis of the experiments verify that the proposed method is competitive.

Effects of substrate surface treatments on hybrid manufacturing of AlSi7Mg using die casting and selective laser melting

To balance the manufacturing cost and customizability of automotive parts,a hybrid manufacturing process combining die-casting and selective laser melting(SLM)is proposed:starting with a conventional cast substrate,SLM is utilized to add additional geometric elements on top of it.For this hybrid process,the first priority is to prepare a substrate surface suitable for the subsequent SLM addition of the top-on elements.In this study,the original cast surface of AlSi7Mg was processed by sandblasting,wire electro-discharge machining,and laser remelting,respectively.Then,additional AlSi7Mg components were built on both the original cast and treated surfaces through SLM.After hybrid builds,these surfaces and resultant interfaces were examined by optical and scanning electron microscopes.Results indicate that the defect-free metallurgical joint between the cast and additively added parts can be formed on all surfaces except for the one processed by electro-discharge machining.The observed epitaxial grain growth crossing the interface implies a strong connection between the cast and the SLMed component.Despite these benefits,also mismatches in microstructure,residual stress level and element distribution between the two parts are identified.After a comprehensive assessment,laser remelting with no additional machining is recommended as the optimal surface treatment preceding SLM fabrication,because of its user-friendly operation,low cost,and high industrial feasibility.

Transport network downsizing based on optimal sub-network

Transportation networks are sized to efficiently achieve some set of service objectives.Under particular circumstances,such as the COVID-19 pandemic,the demand for transportation can significantly change,both qualitatively and quantitatively,resulting in an over-capacitated and less efficient network.In this paper,we address this issue by proposing a framework for resizing the network to efficiently cope with the new demand.The framework includes a model to determine an optimal transportation sub-network that guarantees the following:(i)the minimal access time from any node of the urban network to the new sub-network has not excessively increased compared to that of the original transportation network;(ii)the delay induced on any itinerary by the removal of nodes from the original transportation network has not excessively increased;and(iii)the number of removed nodes from the transportation network is within a preset known factor.A solution is optimal if it induces a minimal global delay.We modelled this problem as a Mixed Integer Linear Program and applied it to the public bus transportation network of Lyon,France,in a case study.In order to respond to operational issues,the framework also includes a decision tool that helps the network planners to decide which bus lines to close and which ones to leave open according to specific trade-off preferences.The results on real data in Lyon show that the optimal sub-network from the MILP model can be used to feed the decision tool,leading to operational scenarios for network planners.

Redundancy in Biology and Robotics:Potential of Kinematic Redundancy and its Interplay with Elasticity

Redundancy facilitates some of the most remarkable capabilities of humans,and is therefore omni-present in our physiology.The relationship between redundancy in robotics and biology is investigated in detail on the Series Elastic Dual-Motor Actuator(SEDMA),an actuator inspired by the kinematic redundancy exhibited by myofibrils.The actuator consists of two motors coupled to a single spring at the output.Such a system has a redundant degree of freedom,which can be exploited to optimize aspects such as accuracy,impedance,fault-tolerance and energy efficiency.To test its potential for human-like motions,the SEDMA actuator is implemented in a hopping robot.Experiments on a physical demonstrator show that the robot's movement patterns resemble human squat jumps.We conclude that robots with bio-inspired actuator designs facilitate human-like movement,although current technical limitations may prevent them from reaching the same dynamic and energetic performance.