Search

Unplanned breakdown of critical equipment interrupts production throughput in Industrial IoT (IIoT), and data-driven Predictive Maintenance (PdM) becomes increasingly important for companies seeking a competitive business advantage. Manufacturers, however, are constantly faced with the onerous challenge of manually allocating suitably competent manpower resources in the event of an unexpected machine breakdown. Furthermore, human error has a negative rippling impact on both overall equipment downtime and production schedules. In this paper, we formulate the complex resource management problem as a resource optimisation problem to determine if a model-free Deep Reinforcement Learning (DRL) based PdM framework can be used to automatically learn an optimal decision-policy from a stochastic environment. Unlike the existing PdM frameworks, our approach considers PdM sensor information and the resources of both physical equipment and human as part of the optimisation problem. The proposed DRL-based framework and Proximal Policy Optimisation Long Short Term Memory (PPO-LSTM) model are evaluated alongside baselines results from human participants using a maintenance repair simulator. Empirical results indicate that our PPO-LSTM efficiently learns the optimal decision-policy for the resource management problem, outperforming comparable DRL methods and human participants by 53% and 65% respectively. Overall, the simulation results corroborate the proposed DRL-based PdM framework’s superiority in terms of convergence efficiency, simulation performance and flexibility.

The Industrial Internet of Things (IIoT) is crucial for accurately assessing the state of complex equipment in order to perform predictive maintenance (PdM) successfully. However, existing IIoT-based PdM frameworks do not consider the influence of various practical yet complex system factors, such as the real-time production states, machine health, and maintenance manpower resources. For this reason, we propose a generic PdM optimization framework to assist maintenance teams in prioritizing and resolving maintenance task conflicts under real-world manufacturing conditions. Specifically, the PdM framework aims to jointly optimize the edge-based machine network uptime and the allocation of manpower resources in a stochastic IIoT-enabled manufacturing environment using the model-free deep reinforcement learning (DRL) methods. Since DRL requires a significant amount of training data, we propose and demonstrate the use of the transfer learning (TL) method to assist DRL in learning more efficiently by incorporating expert demonstrations, termed TL with demonstrations (TLDs). TLD reduces training wall time by 58% compared to baseline methods, and we conduct numerous experiments to illustrate the performance, robustness, and scalability of TLD. Finally, we discuss the general benefits and limitations of the proposed TL method, which are not well addressed in the existing literature but could be beneficial to both researchers and industry practitioners.

Polymer formulations possessing mesostructures or phase coexistence are challenging to simulate using atomistic particle-explicit approaches due to the disparate time and length scales, while the predictive capability of field-based simulations is hampered by the need to specify interactions at a coarser scale (e.g., χ-parameters). To overcome the weaknesses of both, we introduce a bottom-up coarse-graining methodology that leverages all-atom molecular dynamics to molecularly inform coarser field-theoretic models. Specifically, we use relative-entropy coarse-graining to parametrize particle models that are directly and analytically transformable into statistical field theories. We demonstrate the predictive capability of this approach by reproducing experimental aqueous poly(ethylene oxide) (PEO) cloud-point curves with no parameters fit to experimental data. This synergistic approach to multiscale polymer simulations opens the door to de novo exploration of phase behavior across a wide variety of polymer solutions and melt formulations.

This article reports on an approach for the VeRoLog Solver Challenge 2019: the fourth solver challenge facilitated by VeRoLog, the EURO Working Group on Vehicle Routing and Logistics Optimization. The authors were awarded third place in this challenge. The routing challenge involved solving two interlinked vehicle routing problems for equipment: one for distribution (using trucks) and one for installation (using technicians). We describe our solution method, based on a matheuristics approach in which the overall problem is heuristically decomposed into components that can then be solved by formulating them as set partitioning problems. To solve these set partitioning problems we introduce a novel method we call “columnwise neighborhood search,” which allows us to explore a large neighborhood of the current solution in an exact manner. By iteratively applying mixed‐integer programming methods, we obtain good quality solutions to our subproblems. We then use a simple local search “fusion” heuristic to further improve the solution to the overall problem. Besides introducing and discussing this solution method, we highlight the problem instances for which our approach was particularly successful in order to obtain general insights about our methodology.

Reliable object detection is essential for smart surveillance systems, and video surveillance technology has been significantly improved thanks to Deep-Learning related techniques. The deployment of resource-constraint edge device algorithms based on Deep-Learning objection detection is an attractive approach with a range of challenging practical issues to solve, such as the enforcement of Deep-Learning model inference reliability on inexpensive, general-purpose edge devices. In this paper, we transform the state-of-the-art convolutional Single Single Shot Detector Object-Detection algorithm into a task optimization problem to investigate whether the overall system reliability can be further improved by using an under-studied technique called semi-partitioned rate-monotonic scheduling algorithm. Unlike current non-scheduling based methods, our scheduling approach and system configuration is empirically shown to achieve state-of-the-art results, such as enhancing Real-Time video inference performance from 0.5 frame-per-second (fps) to 2fps (i.e. four-fold increase) with a recorded mean 68% improvement in video fps deviation. Our findings are important as fluctuations in video fps, hereby termed system reliability, is known to affect the detection accuracy of object detection algorithms. Finally, we discuss the results and highlight several observations that have not been explicitly addressed in current literature.

In order to encourage innovative medicine to address Chinese unmet medical needs, China has changed its drug regulatory landscape to speed up access to new medicines. In order to understand the fast‐changing landscape and to enable planning of more global drug development programs and study designs in China, we reviewed 15 published clinical pharmacology‐related guidances by the National Medical Products Administration (NMPA), and compared them with reference guidances from the US Food and Drug Administration (FDA), the European Medicines Agency (EMA), or the International Conference on Harmonization (ICH), to understand the similarities and differences, especially any China‐specific requirements, such as ethnic sensitivity analysis. Overall, by reviewing these clinical pharmacology‐related NMPA guidances, it is clear that NMPA guidances are very similar to FDA, EMA, and ICH guidances. There are no relevant differences in the major principles, but some differences in structure, contents, and focus were noted. The NMPA is adapting flexibility statements into newly published guidances. Ethnic sensitivity analysis needs to be implemented early in drug development plans. The NMPA encourages sponsors to conduct early clinical trials in China or include China early in multiregional clinical trials, and to obtain safety, efficacy, and pharmacokinetic data for ethnic sensitivity analysis. Depending on the stage of development, ethnic sensitivity analysis can be conducted using in vitro or literature data, other Asian clinical data, or Chinese clinical data.

Twitter has become one of the main information sharing platforms for millions of users world-wide. Numerous tweets are created daily, many with highly time sensitive content such as breaking news, new multimedia content or personal updates. Consequently, accurately recommending relevant tweets to users in a timely manner is a highly important and challenging problem. The 2020 ACM RecSys Challenge is aimed at benchmarking leading recommendation models for this task. The challenge is based on a large and recent dataset of over 200M tweet engagements released by Twitter with content in over 50 languages. In this work we present our approach where we leverage recent advances in deep language modeling and attention architectures, to combine information from extracted features, user engagement history and target tweet content. We first fine-tune leading multilingual language models M-BERT and XLM-R for Twitter data. Embeddings from these models are used to extract tweet and user history representations. We then combine all components together and jointly train them to maximize engagement prediction accuracy. Our approach achieves highly competitive performance placing 2’nd on the final private leaderboard. Full code is available here: https://github.com/layer6ai-labs/RecSys2020.

Failure of mission-critical equipment interrupts production and results in monetary loss. The risk of unplanned equipment downtime can be minimized through Predictive Maintenance of revenue generating assets to ensure optimal performance and safe operation of equipment. However, the increased sensorization of the equipment generates a data deluge, and existing machine-learning based predictive model alone becomes inadequate for timely equipment condition predictions. In this paper, a model-free Deep Reinforcement Learning algorithm is proposed for predictive equipment maintenance from an equipment-based sensor network context. Within each equipment, a sensor device aggregates raw sensor data, and the equipment health status is analyzed for anomalous events. Unlike traditional black-box regression models, the proposed algorithm self-learns an optimal maintenance policy and provides actionable recommendation for each equipment. Our experimental results demonstrate the potential for broader range of equipment maintenance applications as an automatic learning framework., 6 pages, 5 figures, accepted in IEEE WF-IoT 2020

We perform a general thermodynamic analysis for the salt partitioning behavior in the coexisting phases for symmetric mixtures of polycation and polyanion solutions. We find that salt partitioning is determined by the competition between two factors involving the ratio of the polyelectrolyte concentration in the coacervate phase to that in the supernatant phase and the difference in the exchange excess chemical potential Δμ_(ex) -- the excess chemical potential difference between PE segments and small ions -- between the coexisting phases. The enrichment of salt ions in the coacervate phase predicted by the Voorn−Overbeek theory is shown to arise from its neglect of chain connectivity in the excess free energy which results in Δμ_(ex) = 0 under all conditions. We argue that chain connectivity in general leads to a finite value of Δμex, which decreases with increasing PE concentration. Explicit calculations using theories that include the chain connectivity correlations -- a simple liquid-state theory and a renormalized Gaussian fluctuation theory -- show nonmonotonic behavior of the salt-partitioning coefficient (the ratio of salt ion concentration in the coacervate phase to that in the supernatant phase): it is larger than 1 at very low salt concentrations, reaches a minimum at some intermediate salt concentration, and approaches 1 at the critical point. This behavior is consistent with recent computer simulation and experimental results.

Using a recently developed renormalized Gaussian fluctuation (RGF) field theory that self-consistently accounts for the concentration-dependent coupling between chain structure and electrostatic correlations, we study the phase behavior of polyelectrolyte solutions, with a focus on the effects of added salts and chain structure. For solutions of a single polyelectrolyte species plus salt, the RGF theory predicts the existence of a loop in the phase boundary at Bjerrum lengths (inverse temperature) below (above) the critical value of the salt-free system. This loop behavior can occur at electrostatic interaction strengths lb/b at which the loop no longer exists for the TPT-1 theory, and at fixed lb/b the loop can persist for infinitely long chains, in contrast to theories using a fixed-Gaussian structure (fg-RPA). For systems of oppositely charged (but otherwise symmetric) chains, we again find that the fg-RPA greatly overpredicts the driving force for phase separation, especially at higher charge fraction...

Receptors to glutamate of the AMPA type (AMPARs) serve as the major gates of excitation in the human brain, where they participate in fundamental processes underlying perception, cognition and movement. Due to their central role in brain function, dysregulation of these receptors has been implicated in neuropathological states associated with a large variety of diseases that manifest with abnormal behaviors. The participation of functional abnormalities of AMPARs in brain disorders is strongly supported by genomic, transcriptomic and proteomic studies. Most of these studies have focused on the expression and function of the subunits that make up the channel and define AMPARs (GRIA1-GRIA4), as well of some accessory proteins. However, it is increasingly evident that native AMPARs are composed of a complex array of accessory proteins that regulate their trafficking, localization, kinetics and pharmacology, and a better understanding of the diversity and regional expression of these accessory proteins is largely needed. In this review we will provide an update on the state of current knowledge of AMPA receptors subunits in the context of their accessory proteins at the transcriptome level. We also summarize the regional expression in the human brain and its correlation with the channel forming subunits. Finally, we discuss some of the current limitations of transcriptomic analysis and propose potential ways to overcome them.

Electro-reduction of spent nuclear oxide fuels in molten salt was the key step of pyroprocessing for oxide fuel treatment. In the present study, the E - p O 2− diagram for rare earth elements in molten LiCl-KCl at 450 °C were developed based available experimental data. E - p O 2− diagrams could show the stability of each chemical compound in the salt, and therefore, the diagrams could be applied to predict experimental conditions for electro-reduction of spent nuclear fuel efficiently. Compared with the available E - p O 2− diagrams, the present study concerned the activity coefficient of the element studied in the molten salt in all reactions, which made the diagram be more reliable and accurate.

Radio Frequency powered Cognitive Radio Networks (RF-CRN) are likely to be the eyes and ears of upcoming modern networks such as Internet of Things (IoT), requiring increased decentralization and autonomous operation. To be considered autonomous, the RF-powered network entities need to make decisions locally to maximize the network throughput under the uncertainty of any network environment. However, in complex and large-scale networks, the state and action spaces are usually large, and existing Tabular Reinforcement Learning technique is unable to find the optimal state-action policy quickly. In this paper, deep reinforcement learning is proposed to overcome the mentioned shortcomings and allow a wireless gateway to derive an optimal policy to maximize network throughput. When benchmarked against advanced DQN techniques, our proposed DQN configuration offers performance speedup of up to 1.8x with good overall performance., 5 pages, 2 figures, accepted in IEEE VTC-Fall 2019

It was experimentally demonstrated by Migler and his collaborators [Phys. Rev. Lett., 2001, 86, 1023; Langmuir, 2003, 19, 8667] that a strongly confined drop monolayer sheared between two parallel plates can spontaneously develop a flow-oriented drop-chain morphology. Here we show that the formation of the chain-like microstructure is driven by far-field Hele-Shaw quadrupolar interactions between drops, and that drop spacing within chains is controlled by the effective drop repulsion associated with the existence of confinement-induced reversing streamlines, i.e., the swapping trajectory effect. Using direct numerical simulations and an accurate quasi-2D model that incorporates quadrupolar and swapping-trajectory contributions, we analyze microstructural evolution in a monodisperse drop monolayer. Consistent with experimental observations, we find that drop spacing within individual chains is usually uniform. Further analysis shows that at low area fractions all chains have the same spacing, but at higher area fractions there is a large spacing variation from chain to chain. These findings are explained in terms of uncompressed and compressed chains. At low area fractions most chains are uncompressed (spacing equals l(st), which is the stable separation of an isolated pair). At higher area fractions compressed chains (with tighter spacing) are formed in a process of chain zipping along y-shaped structural defects. We also discuss the relevance of our findings to other shear-driven systems, such as suspensions of spheres in non-Newtonian fluids.

Altered expression of GABA receptors (GABAARs) has been implicated in neurological and psychiatric disorders, but limited information about region-specific GABAAR subunit expression in healthy human brains, heteromeric assembly of major isoforms, and their collective organization across healthy individuals, are major roadblocks to understanding their role in non-physiological states. Here, by using microarray and RNA-Seq datasets—from single cell nuclei to global brain expression—from the Allen Institute, we find that transcriptional expression of GABAAR subunits is anatomically organized according to their neurodevelopmental origin. The data show a combination of complementary and mutually-exclusive expression patterns that delineate major isoforms, and which is highly stereotypical across brains from control donors. We summarize the region-specific signature of GABAR subunits per subject and its variability in a control population sample that can be used as a reference for remodeling changes during homeostatic rearrangements of GABAAR subunits after physiological, pharmacological or pathological challenges.

Achieving thermodynamic faithfulness and transferability across state points is an outstanding challenge in the bottom-up coarse graining of molecular models, with many efforts focusing on augmenting the form of coarse-grained interaction potentials to improve transferability. Here, we revisit the critical role of the simulation ensemble and the possibility that even simple models can be made more predictive through a smarter choice of ensemble. We highlight the efficacy of coarse graining from ensembles where variables conjugate to the thermodynamic quantities of interest are forced to respond to applied perturbations. For example, to learn activity coefficients, it is natural to coarse grain from ensembles with spatially varying external potentials applied to one species to force local composition variations and fluctuations. We apply this strategy to coarse grain both an atomistic model of water and methanol and a binary mixture of spheres interacting via Gaussian repulsions and demonstrate near-quantitative capture of activity coefficients across the whole composition range. Furthermore, the approach is able to do so without explicitly measuring and targeting activity coefficients during the coarse graining process; activity coefficients are only computed after-the-fact to assess accuracy. We hypothesize that ensembles with applied thermodynamic potentials are more "thermodynamically informative." We quantify this notion of informativeness using the Fisher information metric, which enables the systematic design of optimal bias potentials that promote the learning of thermodynamically faithful models. The Fisher information is related to variances of structural variables, highlighting the physical basis underlying the Fisher information's utility in improving coarse-grained models.

Altered expression of GABA receptors (GABAARs) has been implicated in neurological and psychiatric disorders, but limited information about region-specific GABAAR subunit expression in healthy human brains, heteromeric assembly of major isoforms, and their collective organization across healthy individuals, are major roadblocks to understanding their role in non-physiological states. Here, by using microarray and RNA-Seq datasets—from single cell nuclei to global brain expression—from the Allen Institute, we find that transcriptional expression of GABAAR subunits is anatomically organized according to their neurodevelopmental origin. The data show a combination of complementary and mutually-exclusive expression patterns that delineate major isoforms, and which is highly stereotypical across brains from control donors. We summarize the region-specific signature of GABAR subunits per subject and its variability in a control population sample that can be used as a reference for remodeling changes during homeostatic rearrangements of GABAAR subunits after physiological, pharmacological or pathological challenges., Adolfo Sequeira et al. report an analysis of GABA receptor subunit expression from single-cell and global transcriptome data generated by the Allen Institute. They find that GABA receptor subunit expression is spatially organized according to neurodevelopmental origin and generate a reference from the control population for future studies.