MB Model Research Articles

CSIG-31. T-TYPE CALCIUM CHANNEL BLOCKERS INDUCE METABOLIC ALTERATIONS AND APOPTOSIS IN MEDULLOBLASTOMA CELLS

Abstract Medulloblastoma (MB) is the most common malignant paediatric brain tumour. In our previous studies, in a screen of a panel of calcium channel modulators in our novel high-throughput 3D assay, we identified T-type (CaV3) channel inhibitors, mibefradil and NNC-55-0396 (NNC) as selective inhibitors of MB cell growth. Mibefradil was originally approved for the treatment of hypertension and angina pectoris, and recently successfully completed a Phase-I trial for recurrent high-grade glioma. NNC is an analogue of mibefradil with multiple advantages compared to mibefradil that makes it attractive for potential future clinical trials. T-type channels have a unique low voltage-dependent activation/inactivation, and many studies suggest that they have a direct regulatory role in controlling Ca2+ signalling in non-excitable tissues, including cancers. In this study, we characterized the effect of mibefradil and NNC on MB cells and elucidated their mechanism of action. This study demonstrates that the induction of toxicity in MB cells is selective to T-type but not to L-type Ca2+ channel inhibitors. Addition of CaV3 inhibitors to vincristine sensitised MB cells to this MB chemotherapeutic agent, suggesting an additive effect. Furthermore, CaV3 inhibitors induced cell death in MB cells via apoptosis. Supported by proteomics data and cellular assays, apoptotic cell death was associated with reduced mitochondrial membrane potential and reduced ATP levels, which suggests that both compounds alter the metabolism of MB cells. We also identified that CACNA1H (encoding the CaV3.2 protein) is significantly upregulated in aggressive Groups 3 and 4 MB patient samples, and its higher expression is associated with higher metastasis and lower patient survival rates. This study offers new insights into the action of mibefradil and NNC and will pave the way to test these molecules or their analogues in pre-clinical MB models alone and in combination with vincristine to assess their suitability as a potential MB therapy.

Surface Crack Detection Method for Coal Rock Based on Improved YOLOv5

Coal mine safety may be able to be ensured via the real-time identification of cracks in rock and coal surfaces. Traditional crack identification methods have the disadvantages of slow speed and low precision. This work suggests an improved You Only Look Once version 5 (YOLOv5) detection model. In this study, we improved YOLOv5 from the perspective of three aspects: a Ghost module was introduced into the backbone network to lighten the model; a coordinate attention mechanism was added; and ECIOU_Loss is proposed as a loss function in this paper to achieve the co-optimization of crack detection speed and accuracy and to meet the deployment requirements in the embedded terminal. The results demonstrate that the improved YOLOv5 has a 92.8% mean average precision (mAP) with an 8 MB model size, and the speed of recognition was 103 frames per second. Compared to the original method, there was a 53.4% reduction in the number of parameters, a detection speed that was 1.9 times faster, and a 1.7% improvement in the mAP. The improved YOLOv5 can effectively locate cracks in real time and offers a new technique for the early warning of coal and rock dynamic hazards.

An improved contact parameter model with elastoplastic behavior for bolted joint interfaces

In this study, considering the elastoplastic behavior of contacting asperities on micro-scale and the contact pressure distribution of bolted joint interfaces on macro-scale, an improved contact stiffness model is presented. Firstly, the bolted T-stub joint is chosen as the object of study, and the distribution of normal contact load over the interface under preload conditions is studied. Then, combining the continuity, smoothness and monotonicity of the constitutive relations of normal contact load-deformation in three different regimes of plastic- elastoplastic-elastic of contacting asperities, the contact stiffness of individual asperity is analytically derived. Meanwhile, the relation between the distribution of contact stiffness and the distribution of normal contact load for bolted joint interfaces is determined based on fractal theory. The presented model overcomes the shortcomings of the contact stiffness without elastoplastic behavior obtained from the MB model and reveals the inhomogeneous distribution features of interfacial contact stiffness in bolted joints. Finally, the finite element (FE) analysis for flange-bolted joints incorporating the proposed model and the relevant experimental verification are performed. The results show that the average error of the first 4-order natural frequencies between the FE results and the experimental ones is only 0.72%, and the predicted results in frequency-domain are consistent with the measured ones, which illustrates that the proposed model can be used in interface modeling of jointed structures.

Compressing speaker extraction model with ultra-low precision quantization and knowledge distillation

Recently, our proposed speaker extraction model, WASE (learning When to Attend for Speaker Extraction) yielded superior performance over the prior state-of-the-art methods by explicitly modeling onset clue and regarding it as important guidance in speaker extraction tasks. However, it still remains challenging when it comes to the deployments on the resource-constrained devices, where the model must be tiny and fast to perform inference with minimal budget in CPU and memory while keeping the speaker extraction performance. In this work, we utilize model compression techniques to alleviate the problem and propose a lightweight speaker extraction model, TinyWASE, which aims to run on resource-constrained devices. Specifically, we mainly investigate the grouping effects of quantization-aware training and knowledge distillation techniques in the speaker extraction task and propose Distillation-aware Quantization. Experiments on WSJ0-2mix dataset show that our proposed model can achieve comparable performance as the full-precision model while reducing the model size using ultra-low bits (e.g. 3 bits), obtaining 8.97x compression ratio and 2.15 MB model size. We further show that TinyWASE can combine with other model compression techniques, such as parameter sharing, to achieve compression ratio as high as 23.81 with limited performance degradation. Our code is available at https://github.com/aispeech-lab/TinyWASE.

MODL-29. Molecular Landscape of a comprehensive panel of pediatric brain cancer Patient-derived orthotopic xenograft (PDOX) models inform unique targets for drug responsiveness

Abstract Brain tumor is a leading cause of cancer related death in children. In addition to replicating histopathology, animal models faithfully replicating genetic/epigenetic abnormalities, molecular subtypes and broad inter-tumoral heterogeneities are needed. Through direct implantation of patient surgical or autopsied tumor tissues into matching locations in the brains of SCID mice, we developed a panel of 150 PDOX mouse models. Here, we report the analysis of 74 of the 150 PDOX models, 45 matching patient tissues and 60 non-tumorigenic samples to a well-annotated reference cohort of 2,801 methylation profiles of primary brain tumors. Our data showed that the lack of tumorigenicity was neither correlated with molecular subtypes nor predicted by low cell viabilities of the patient samples. Methylation profiling identified PDOX models representing nearly a full spectrum of molecular subtypes of pediatric brain tumors including GBM, medulloblastoma, ependymoma and ATRT. Direct comparison with the original patient tumors confirmed the replication of molecular subtypes. ONCOplot [FB1] analysis of PDOX models derived from matching pairs of primary and recurrent tumors (n=8) revealed close clustering with the patient tumors. Investigation of metastatic properties was performed in 13 MB models by harvesting and sub-transplanting matching PDOX primary tumors in the cerebella and metastatic tumors in the spinal cords. To confirm the potential and power of PDOX models in preclinical drug testing, we applied fractionated radiation (2 Gy/day x 5 days) and optimized multi-agent combinatory chemotherapies in MB models of the four major subgroups. High-throughput combination drug screening with ~ 8,000 drugs in PDOX-derived GBM cell lines and primary cultures of MB PDOX cells identified a library of ~ 3,500 drugs that were active in pediatric brain tumors. In summary, this study provides detailed information on molecular subclassification of a uniquely large cohort of PDOX models to serve as essential tools for brain tumor research.

MODL-14 Inhibition of cell cycle mechanisms to combat high-risk medulloblastoma

Abstract Medulloblastoma (MB) is the most common embryonal brain tumor in children, characterized by a high level of heterogeneity within this entity. Four major molecular groups have been defined - WNT, Sonic hedgehog (SHH), Group 3 (G3), and Group 4 (G4) - differing widely in genetic alterations and patient outcomes. Targeted treatment approaches are limited for high-risk MB including MYC-amplified G3, G4, or MYCN/GLI2-amplified SHH. Based on earlier studies describing the inhibition of cell cycle regulation as a therapeutic vulnerability in high-risk MB, we aimed to explore the efficacy and mode of action of approved CDK4/6 inhibitors in aggressive MB. First the in vitro sensitivity of pediatric G3 (n=5) MB cell models against palbociclib, ribociclib and abemaciclib was evaluated and compared to adult SHH (n=2) cells serving as MYC-negative controls. Abemaciclib most effectively reduced cell viability, exhibiting a mean IC50 value of 0.9 M in G3 compared to 2.1 M in SHH cells (p=0.007). Upon long-term exposure (7-10 days), again abemaciclib most potently decreased clonogenic survival already at 250nM in G3 MB cells. We observed a G1/S phase cell cycle arrest upon abemaciclib treatment and confirmed this by Western blot analyses, showing decreased Rb phosphorylation accompanied by reduced MAPK signaling. Interestingly, abemaciclib treatment distinctly reduced MYC protein and gene expression levels in MYC-amplified G3 cell models. In addition, exposure to abemaciclib decreased GLI2 and TERT mRNA levels, which was accompanied by induction of cellular senescence. Within the ITCC-P4 project, single mouse trials are currently being performed including abemaciclib treatment in orthotopic MB models. The detailed investigation of the precise mode of action in vitro and in vivo is ongoing. Summarizing, the CDK4/6 clinically approved inhibitor abemaciclib shows promising efficacy against high-risk MB in vitro.

MEDB-76. Evaluating the B7-H3 checkpoint in Medulloblastoma

BACKGROUND: There is currently no curative therapy for recurrent/refractory MB. Novel approaches to MB include immunotherapy, such as targeting the immune checkpoint molecule B7-H3. B7-H3 is implicated in tumor metastasis and is highly expressed in MB. This study explores the effects of genetically knocking down B7-H3 in a murine model of recurrent/refractory medulloblastoma. METHODS: Murine MB cells were transduced with a CRISPR/Cas9 lentivirus to create a B7-H3 knockout. Knockout population was sorted twice via FACS by the AECOM flow cytometry CORE and confirmed by western blot and flow cytometry. Three healthy clones were used in subsequent studies, and compared to the wild type and the scramble control. IncuCyte live imaging technology was used to evaluate spheroid growth. Matrigel Boyden chambers were used to evaluate migration. Bulk RNA-seq was performed by the Yale University Core. RESULTS: B7-H3 knockout was successful in the murine MB model. Morphological differences were noted in the B7-H3 knockout cells. Spheroid formation assays show one of the clones with statistically slower growth kinetics compared to controls. Migration results are pending. RNA seq revealed similar clustering amongst knockouts, separate from controls with an enrichment in genes of morphologic development, WNT signaling and amoeboid migration. CONCLUSIONS: The morphologic changes in the B7-H3 knockouts suggest a potential growth differential. Although in vitro growth assays have shown mixed results regarding the effect of knocking out B7-H3 in spheroid formation, B7-H3 has been more directly implicated in migration and immune signaling. If migration is impaired, this will suggest that B7-H3 enhances malignant and metastatic potential in MB. Functional in vivo immune studies in syngeneic mice will investigate immune mediated effects of B7-H3 knockout in this tumor. If our studies support a role for B7-H3 in the development of MB, it may have important clinical implications, particularly for relapsed patients.

MODL-21. Multi-omics investigation of medulloblastoma resistant models reveals functional association between intracellular regulatory networks and drug susceptibility

Abstract Medulloblastoma (MB) is the most common malignant brain tumor of childhood. Despite high-dose radio/chemotherapy treatment, 15-30% of patients still display a high risk of tumor recurrence. In this context, the characterization of innovative cellular models resembling therapy-induced drug resistance represents an unprecedented opportunity for selecting relevant therapies for chemotherapy-refractory patients. In order to unveil the molecular mechanisms sustaining chemotherapy resistance in MB, we setup in vitro models of MB drug resistance by exposing Patient-Derived MB (PD-MB) cells to a combination of the commonly used chemotherapeutics for pediatric MB treatment. Integration of multi-omics data, including transcriptional, proteomic and kinase activation profiling, disclosed that drug resistant PD-MB cells are characterized by a significant deregulation of several cancer-related pathways converging to metabolism of xenobiotics, adaptation of the biochemical processes sustaining energetic metabolic demand, cell proliferation and survival, protein homeostasis, RNA processing and modification, and immune response. Moreover, this intriguing regulatory network was functionally associated to the response of drug-resistant MB cells to a large library of compounds through a semi-automated High-Throughput drug Screening (HTS) workflow, suggesting the antimetabolite class of drugs as relevant therapeutics displaying high selectivity and efficacy against resistant MB models, together with a significant synergistic action when combined with standard chemotherapeutic agents. Collectively, our results suggest that drug-resistant MB cells are subjected to a peculiar adaptation of multiple intracellular processes during adaptation to chemotherapy, which protects them from the toxic environment but, at the same time, provides targetable vulnerabilities for therapeutic purposes.

Restricting Meta-Analyses’ Source Studies to Non-consumer Reference Groups Has Significant Effects on the Link between Processed Meat and Type II Diabetes

ObjectivesTo compare estimates of the relative risk (RR) of type-2 diabetes mellitus (T2DM) from processed meat consumption between observational studies including only non-consumers in the reference exposure groups (ZB) vs studies including both no- and low-amount processed meat consumers (mixed baseline, or MB).MethodsStudies examining processed meat consumption and incidence of T2DM were selected from references of the Global Burden of Disease (GBD) project's 2020 systematic review. For studies meeting the inclusion criteria, a meta-analysis regression was fit to the RR of the lowest exposure group relative to the reference. Both fixed and random effect models and modifiers including grams/day and the type of baseline (ZB or MB) were tested for significance.ResultsSeven of the 19 cohorts that met the inclusion criteria had ZB. The MB model, using all 15 studies, was best fit by a random-effect model modified with both grams/day and the use of ZB as reference. The effect size for additional 10 g/day was 1.07 (95% CI 1.02–1.12). The coefficient for use of a ZB suggests those studies have a lower RR than MB studies when also controlling for differences in the gram/day consumed in non-reference group. The best fitting model for the ZB studies (no consumers in the baseline) was the fixed effect model without addition of any covariates. The effect for 10 g/day estimated with this meta-analysis was 1.07 (1.03–1.12).ConclusionsThe identification of zero-baseline study groups as a covariate of the meta-analysis had a significant impact on the model. In this model, both the addition of grams per day and of an identifying variable of a zero-baseline group had a significant effect, indicating that estimate of effect based on a zero-consumption reference group can be more precise than reference groups with consumers. However, the expected impact of additional precision in the exposures would be a larger effect size estimate. The overall lack of change in adjusted relative risks from the meta-analysis combined with this result suggest evidence of effect of processed meat and T2DM at low consumption levels might not be consistent, and possibly biased by residual confounding.Funding SourcesThis work was partially funded by MatPrat Norway. The funders had no input on the analysis or preparation of this abstract.

Contact characteristic parameters modeling for the assembled structure with bolted joints

Considering the non-uniformly distributed contact pressure, this study presents a new contact characteristic parameters (contact stiffness and damping) model for bolted joint interfaces. Firstly, to transform the contact pressure distribution into normal contact load distribution, the contact pressure distribution and its mathematical representation of single-bolted joint are studied. Subsequently, the relationships among contact stiffness distribution, contact damping distribution and normal contact load distribution are determined based on the improved MB model. It is found that the interfacial contact stiffness and damping exhibit significant inhomogeneous distributions. The proposed model overcomes the shortcomings of homogeneous distributions of contact stiffness and damping in bolted joints derived from previous models. In addition, the effects of bolt preload, tangential-to-normal contact load ratio and friction coefficient on the contact stiffness and damping are discussed in detail. Finally, the finite element (FE) analysis for beam lap structure with bolted joints incorporating the proposed model and the relevant experimental are performed to verify the validity of the modeling method. The results indicate that the proposed model can significantly improve the modeling accuracy of bolted joints.

Frondoside A Inhibits an MYC-Driven Medulloblastoma Model Derived from Human-Induced Pluripotent Stem Cells.

Medulloblastoma (MB) is the most common malignant pediatric brain tumor. MYC-driven MBs, commonly found in the group 3 MB, are aggressive and metastatic with the worst prognosis. Modeling MYC-driven MB is the foundation of therapeutic development. Here, we applied a synthetic mRNA-driven strategy to generate neuronal precursors from human-induced pluripotent stem cells (iPSCs). These neuronal precursors were transformed by the MYC oncogene combined with p53 loss of function to establish an MYC-driven MB model recapitulating the histologic and transcriptomic hallmarks of group 3 MB. We further show that the marine compound Frondoside A (FA) effectively inhibits this MYC-driven MB model without affecting isogenic neuronal precursors with undetectable MYC expression. Consistent results from a panel of MB models support that MYC levels are positively correlated with FA's antitumor potency. Next, we show that FA suppresses MYC expression and its downstream gene targets in MB cells, suggesting a potential mechanism underlying FA's inhibitory effects on MYC-driven cancers. In orthotopic xenografts of MYC-driven MB, intratumoral FA administration potently induces cytotoxicity in tumor xenografts, significantly extends the survival of tumor-bearing animals, and enhances the recruitment of microglia/macrophages and cytotoxic T lymphocytes to tumors. Moreover, we show that MYC levels also predict FA potency in glioblastoma and non-small cell lung cancer cells. Taken together, this study provides an efficient human iPSC-based strategy for personalizable cancer modeling, widely applicable to mechanistic studies (e.g., genetic predisposition to cancer) and drug discovery. Our preclinical results justify the clinical translation of FA in treating MYC-driven MB and other human cancers.

Boundary graph convolutional network for temporal action detection

Temporal action proposal generation is a fundamental yet challenging to locate the temporal action in untrimmed videos. Although current proposal generation methods can generate the precise boundary of actions, few focus on considering the relation of proposals. In this paper, we propose a unified framework to generate the temporal boundary proposals with a graph convolution network based on the boundary proposals' feature named Boundary Graph Convolutional Network (BGCN). BGCN draws inspiration from boundary methods and uses edge graph convolution relay on the boundary proposals' feature. First, we use a base layer to fusion the two-stream video features to get two-branches of base features. Then the two-branches of base features enter into the same structure of Proposal Features Graph Convolutional Network (PFGCN): Action PFGCN to extract the action classification score and Boundary PFGCN to extract the ending score and staring score. In proposal features graph convolutional network, we first densely sampled the proposals' feature from the video features. We construct a proposal feature graph, where each proposal feature as a node and their relations between proposals' features as an edge with edge convolution for graph convolution. After that, map the relations into a 2D map score. Experiments on popular benchmarks THUMOS14 demonstrate the superiority of BGCN over (44.8% versus 42.8% at tIoU 0.5) the state-of-the-art proposal generator (e.g., G-TAD, TAL-Net, and BMN) at any of tIoU thresholds from 0.3 to 0.7. On ActivityNet1.3, BGCN also got better results. Moreover, BGCN has high efficiency for action detection with less than 2 MB model size and fast inference time.GCN based on boundary generation for densely produce the action proposals Efficient and novel BGCN model has a great capability to learn the proposal features Has a lower model size for temporal action proposals generation Has fast inference time for temporal action proposals generation.

EmNet: a deep integrated convolutional neural network for facial emotion recognition in the wild

In the past decade, facial emotion recognition (FER) research saw tremendous progress, which led to the development of novel convolutional neural network (CNN) architectures for automatic recognition of facial emotions in static images. These networks, though, have achieved good recognition accuracy, they incur high computational costs and memory utilization. These issues restrict their deployment in real-world applications, which demands the FER systems to run on resource-constrained embedded devices in real-time. Thus, to alleviate these issues and to develop a robust and efficient method for automatic recognition of facial emotions in the wild with real-time performance, this paper presents a novel deep integrated CNN model, named EmNet (Emotion Network). The EmNet model consists of two structurally similar DCNN models and their integrated variant, jointly-optimized using a joint-optimization technique. For a given facial image, the EmNet gives three predictions, which are fused using two fusion schemes, namely average fusion and weighted maximum fusion, to obtain the final decision. To test the efficiency of the proposed FER pipeline on a resource-constrained embedded platform, we optimized the EmNet model and the face detector using TensorRT SDK and deploy the complete FER pipeline on the Nvidia Xavier device. Our proposed EmNet model with 4.80M parameters and 19.3MB model size attains notable improvement over the current state-of-the-art in terms of accuracy with multi-fold improvement in computational efficiency.

Learning Dynamic Relationships for Facial Expression Recognition Based on Graph Convolutional Network.

Facial action units (AUs) analysis plays an important role in facial expression recognition (FER). Existing deep spectral convolutional networks (DSCNs) have made encouraging performance for FER based on a set of facial local regions and a predefined graph structure. However, these regions do not have close relationships to AUs, and DSCNs cannot model the dynamic spatial dependencies of these regions for estimating different facial expressions. To tackle these issues, we propose a novel double dynamic relationships graph convolutional network (DDRGCN) to learn the strength of the edges in the facial graph by a trainable weighted adjacency matrix. We construct facial graph data by 20 regions of interest (ROIs) guided by different facial AUs. Furthermore, we devise an efficient graph convolutional network in which the inherent dependencies of vertices in the facial graph can be learned automatically during network training. Notably, the proposed model only has 110K parameters and 0.48MB model size, which is significantly less than most existing methods. Experiments on four widely-used FER datasets demonstrate that the proposed dynamic relationships graph network achieves superior results compared to existing light-weight networks, not just in terms of accuracy but also model size and speed.

Comparison of the performances of GEP, ANFIS, and SVM artifical intelligence models in rainfall simulaton

In this paper, evaluation the performances of GEP (gene expression programming), ANFIS (adaptive fuzzy interference system), and SVM (support vector machine) artificial intelligence models in two scales of daily and monthly rainfall data from Urmia meteorological station (Iran) and monthly rainfall data from Diata meteorological station (India) was used in rainfall simulation. The correlation coefficient of observed and simulated values was evaluated by the R2 criterion, simulation error was evaluated by the root mean square error (RMSE), and MB criteria and model efficiency were evaluated by the Nash-Sutcliffe method. The results show that the correlation coefficients in the GEP model based on daily data from Urmia station and monthly data from Diata station are 23 and 58%, respectively, and R2 in simulation with GEP is estimated to be 55% lower than with the other two models. The R2 range in both ANFIS and SVM models varies from 91 to 93%. On average, the RMSE values in the GEP simulation are 50% and 55% higher than the ANFIS ratio for daily and monthly data at the two stations, respectively, and the RMSE values of ANFIS model are 1% and 3% higher than those of the SVM at Urmia and Diata stations, respectively. The bias values of the GEP model are 72% and 60% higher than those of ANFIS at Urmia and Diata stations, respectively. The GEP efficiency factors are 56% and 61% lower than those of ANFIS at Urmia and Diata stations, respectively. And the ANFIS efficiency ratio is 1 and 2% lower than SVM in Urmia and Diata stations, respectively. Therefore, rainfall simulation with the SVM model is associated with a lower error rate and better efficiency, the ANFIS model is close to the efficiency of SVM, and the GEP model is not suitable for rainfall simulation.

MBRS-48. IDENTIFICATION OF NOVEL THERAPEUTIC APPROACHES FOR MYC-DRIVEN MEDULLOBLASTOMA

Medulloblastoma (MB) is the most common malignant brain tumor in children and is frequently metastatic at diagnosis. Treatment with surgery, radiation and multi-agent chemotherapy may leave survivors of these brain tumors with long-term deficits as a consequence. One of the four consensus molecular subgroups of MB is the MYC-driven group 3 MB, which is the most malignant type and has a poor prognosis under current therapy. Thus, it is important to discover more effective targeted therapeutic approaches. We conducted a high-throughput drug screening to identify novel compounds showing efficiency in group 3 MB using both clinically established inhibitors (n=196) and clinically-applicable compounds (n=464). More than 20 compounds demonstrated a significantly higher anti-tumoral effect in MYChigh (n=7) compared to MYClow (n=4) MB cell models. Among these compounds, Navitoclax and Clofarabine showed the strongest effect in inducing cell cycle arrest and apoptosis in MYChigh MB models. Furthermore, we show that Navitoclax, an orally bioavailable and blood-brain barrier passing anti-cancer drug, inhibits specifically Bcl-xL proteins. In line, we found a significant correlation between BCL-xL and MYC mRNA levels in 763 primary MB patient samples (Data source: “R2 https://hgserver1.amc.nl”). In addition, Navitoclax and Clofarabine have been tested in cells obtained from MB patient-derived-xenografts, which confirmed their specific efficacy in MYChigh versus MYClow MB. In summary, our approach has identified promising new drugs that significantly reduce cell viability in MYChigh compared to MYClow MB cell models. Our findings point to novel therapeutic vulnerabilities for MB that need to be further validated in vitro and in vivo.

A multilevel finite element-multibody approach to design the suspension system for the road transportation of SSR1 cryomodule

Cryomodules (CM) represents edge-frontier assemblies in particle physics research field. The road transportation of CM is a critical phase during which the structures can be subjected to significant dynamic loads. It is therefore necessary to design a Transportation Tool (TT) equipped with an appropriate suspension system. This work describes the approach adopted for the design of the TT for the CM PIP-II SSR1 (Proton Improvement Plan-II – Single Spoke Resonators 1), which is firstly introduced in the CM research field. Initially a Finite Element (FE) model was developed, considering the main sub-assemblies of the CM. However, this model was not suitable for the design of the TT due to the high computational burden. For this reason the model was exported as a Modal Neutral File and imported into a MultiBody software (MB) where the remaining components were modeled as concentrated stiffnesses or rigid bodies. The MB model thus obtained has drastically reduced the calculation time, proving to be fundamental in the TT iterative design phase, which involves the use of Helical Isolators (HI) performing the function of mechanical filters. To validate the effectiveness of TT in reducing dynamic loads, a 3D acceleration profile measured during the transport of a similar cryomodule (Linear Coherent Light Source II, LCLS-II) was used. Furthermore, the results of the MB model were used to perform the structural verification of some critical components of the CM.

A lightweight deep learning algorithm for inspection of laser welding defects on safety vent of power battery

With the wide applications of power battery in the automobile industries, the safety of power battery is becoming an increasingly prominent problem. At present, a safety vent welded on the battery plate could prevent unpredictable explosions. To perform quality control, the inspection of laser welding defects on safety vent is a critical issue. In this paper, based on the theory of convolutional neural network (CNN) and the technique of transfer learning, a pre-trained SqueezeNet model with small model size and low computation complexity was proposed. We totally collected 34537 images from the production line, and built a 2-classifications dataset and a 7-classifications dataset respectively. It proves that our proposed model achieved better accuracy than the other six contrastive CNN models in these two classification tasks. Specifically, it obtained an accuracy of 99.57 % in the 2-classifications task and an accuracy of 95.58 % in the 7-classifications task. Besides, the model features lightweight and high-speed with only 1.2 MB model size and 4.9 ms average test time; so, it is more suitable for welding quality inspection of safety vent in an industrial production line. Additionally, we ported our pre-trained SqueezeNet and the other four contrastive models to Raspberry Pi embedded system to evaluate their response time. Our proposed model has achieved well response time of less than 330 ms. Our experiments indicate that the proper CNN model can help to conduct the laser welding quality inspection tasks that are usually performed by humans.

Driven by Data or Derived Through Physics? A Review of Hybrid Physics Guided Machine Learning Techniques With Cyber-Physical System (CPS) Focus

A multitude of cyber-physical system (CPS) applications, including design, control, diagnosis, prognostics, and a host of other problems, are predicated on the assumption of model availability. There are mainly two approaches to modeling: Physics/Equation based modeling (Model-Based, MB) and Machine Learning (ML). Recently, there is a growing consensus that ML methodologies relying on data need to be coupled with prior scientific knowledge (or physics, MB) for modeling CPS. We refer to the paradigm that combines MB approaches with ML as hybrid learning methods. Hybrid modeling (HB) methods is a growing field within both the ML and scientific communities, and are recognized as an important emerging but nascent area of research. Recently, several works have attempted to merge MB and ML models for the complete exploitation of their combined potential. However, the research literature is scattered and unorganized. So, we make a meticulous and systematic attempt at organizing and standardizing the methods of combining ML and MB models. In addition to that, we outline five metrics for the comprehensive evaluation of hybrid models. Finally, we conclude by shedding some light on the challenges of hybrid models, which we, as a research community, should focus on for harnessing the full potential of hybrid models. An additional feature of this survey is that the hybrid modeling work has been discussed with a focus on modeling cyber-physical systems.

Shallow SqueezeNext: Real Time Deployment on Bluebox2.0 with 272KB Model Size

The significant challenges for deploying CNNs/DNNs on ADAS are limited computation and memory resources with very limited efficiency. Design space exploration of CNNs or DNNS, training and testing DNN from scratch, hyper parameter tuning, implementation with different optimizers contributed towards the efficiency and performance improvement of the Shallow SqueezeNext architecture. It is also computationally efficient, inexpensive and requires minimum memory resources. It achieves better model size and speed in comparison to other counterparts such as AlexNet, VGGnet, SqueezeNet, and SqueezeNext, trained and tested from scratch on datasets such as CIFAR-10 and CIFAR-100. It can achieve the least model size of 272KB with a model accuracy of 82%, a model speed of 9 seconds per epoch, and tested on the CIFAR-10 dataset. It achieved the best accuracy of 91.41%, best model size of 0.272 MB, and best model speed of 4 seconds per epoch. Memory resources are of high importance when it comes down to real time system or platforms because usually the memory is quite limited. To verify that the Shallow SqueezeNext can be successfully deployed on a real time platform, bluebox2.0 by NXP was used. Bluebox2.0 deployment of Shallow SqueezeNext architecture achieved a model accuracy of 90.50%, 8.72MB model size and 22 seconds per epoch model speed. There is another version of the Shallow SqueezeNext which performed better that attained a model size of 0.5MB with model accuracy of 87.30% and 11 seconds per epoch model speed trained and tested from scratch on CIFAR-10 dataset.