Zhiqiang Wang

@inproceedings{wang_2364,

title = {Contextual classification of chatter based on unsupervised machine learning},

author = {Zhiqiang Wang and Mathieu Ritou and Benoît Furet},

url = {https://www.sciencedirect.com/science/article/pii/S2212827123001981},

issn = {2212-8271},

year  = {2023},

date = {2023-06-01},

booktitle = {Procedia CIRP 19th CIRP Conference on Modeling of Machining Operations},

volume = {117},

pages = {390-395},

address = {Karlsruhe, Germany},

abstract = {In the context of Industry 4.0, large volumes of manufacturing data are available on instrumented machine-tool. The critical point is the

exploitation of this digital content. Data contextualization is important for efficient and robust data mining, particularly for industrial production. For this purpose, a classification method of the operational manufacturing context is proposed. It relies on knowledge integration (by business rules) and unsupervised machine learning, with a Gaussian Mixture Model, applied to in-process monitoring signals. The method was evaluated on real industrial machining databases collected during one year. Manual data mining shows that this method is accurate on industrial production (98.9%). Moreover, the application of contextual classifications for the chatter detection on the same production data shows the relevancy of the proposed data mining approach.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{wang_2365,

title = {NLP based on GCVAE for intelligent Fault Analysis in Semiconductor industry},

author = {Zhiqiang Wang and Kenneth Ezukwoke and Anis Hoayek and Mireille Batton-Hubert and Xavier Boucher},

url = {https://ieeexplore.ieee.org/document/9921524},

issn = {9781665499972},

year  = {2022},

date = {2022-10-01},

booktitle = {2022 IEEE 27th International Conference on Emerging Technologies and Factory Automation (ETFA)},

address = {Stuttgart, Germany},

abstract = {In the semiconductor industry, Failure Analysis (FA) is an investigation to determine the root causes of a failure. It also involves an intermediate analysis to build the steps of the failure analysis in order to mitigate future failures and to facilitate the future FA. In the framework of the FA 4.0 project, the reporting system records three items of information using natural language: the failure analysis request description (input space) and analysis steps (paths), as well as generic categories of root cause conclusion (output space). The main objective of this article is to develop a system which is able to automatically help industries carry out fault analysis diagnoses with Artificial intelligence (AI). This article extends and validates the adapted methodology proposed by [1] to transform text data into numeric data based on Natural Language Processing (NLP). It transforms the text data from the input space and output space. Different deep learning algorithms based on a Variational AutoEncoder (VAE) are applied to the output space to reduce the dimension of the numeric data, and the performance of each VAE is evaluated with different metrics. The Generalized-Controllable VAE (GCVAE) is the one best suited to our case. A Gaussian Mixture Model (GMM) is then used to perform clustering in the latent space generated by the GCVAE. A centroid analysis is also conducted to verify the similarity of each cluster.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{wang_2623,

title = {Identification du contexte d'usinage basée sur le modèle de mélange gaussien : application au broutement.},

author = {Zhiqiang Wang and Mathieu Ritou and Catherine DA CUNHA and Benoît Furet},

url = {https://hal.science/hal-03436801/},

year  = {2021},

date = {2021-07-01},

booktitle = {Conférence Manufacturing'21 2021},

address = {Angers, France},

abstract = {Les machines-outils intelligentes gérèrent de nombreuses données et l'exploration de ces données peut faciliter la prise de décision en matière de gestion opérationnelle. Une sélection fine des données pertinentes, basée sur le contexte d'usinage, est nécessaire pour un calcul précis des indicateurs clés de performance (KPIs). Il s'agit d'une étape importante pour une exploration efficace des données. Cet article propose ainsi des règles métiers pour la classification contextuelle, basée sur un apprentissage automatique non-supervisé par le Modèle de Mélange Gaussien (GMM). Les fouilles des données manuelles montrent que la précision de la méthode est très bonne. Dés lors, différents types d'usinage et différent états de l'engagement de l'outil peuvent être identifiées automatiquement à chaque instant, par cette classification contextuelle. Ainsi, des KPIs, tel que la distribution des cas de broutement par type d'usinage ou d'engagement de l'outil, peuvent être évalués sur la base de cette classification contextuelle.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{wang_2621,

title = {Classification contextuelle pour les fouilles de données issues de machines-outils.},

author = {Zhiqiang Wang and Mathieu Ritou and Catherine DA CUNHA and Benoît Furet},

url = {https://hal.science/hal-02279417/document},

year  = {2019},

date = {2019-08-01},

booktitle = {24ème Congrès Français de Mécanique},

address = {Brest, France},

abstract = {Dans le contexte général de l'Industrie 4.0, une entreprise de fabrication moderne dispose de nombreuses données numériques qui pourraient être utilisées pour rendre les machines-outils plus

intelligentes et faciliter la prise de décision en matière de gestion opérationnelle. L'une des premières étapes de l'approche d'exploration de données est la sélection précise de données pertinentes. Pour ce faire, les données brutes doivent être classées dans différents groupes de contexte. Cet article présente un algorithme d'apprentissage automatique non-supervisé, par mélanges gaussiens (GMM), pour la classification contextuelle. Le nombre de clusters optimal est déterminé par le critère BIC (Bayesian Information Criterion), à partir de données venant de l'industrie aéronautique. Les vérifications par fouilles manuelles et la validation par k-fold montrent ensuite que la méthode GMM permet d'obtenir de bons résultats de classification contextuelle.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{wang_1968,

title = {Comparison of K-means and GMM methods for contextual clustering in HSM},

author = {Zhiqiang Wang and Catherine DA CUNHA and Mathieu Ritou and Benoît Furet},

url = {https://doi.org/10.1016/j.promfg.2018.12.025},

issn = {2351-9789},

year  = {2019},

date = {2019-08-01},

booktitle = {Procedia Manufacturing},

volume = {28},

pages = {154-159},

publisher = {Elsevier},

address = {Nantes, France},

abstract = {High speed machining (HSM) is widely used for the manufacturing of aircraft structures, turbine blades, etc. It greatly increases the efficiency and automation for the machining. However, in HSM, operators cannot detect incidents when they manage several machines of a production cell. Robust monitoring systems are required to protect the machine tool and the high value added parts. In the global context of the Industry 4.0, abundant digital data is available in a modern manufacturing company and could be used to turn the machines-tools smarter and to support the decision making of the operational management.



One of the first step of data mining approach is the accurate selection of relevant. To do so, the raw data need to be classified into different contextual clusters. This paper compares two different methods of the unsupervised classification of machining context: K-means and GMM (Gaussian Mixture Model). It was found that GMM method can classify correctly the machining context, whereas K-means is not suitable.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{wang_2618,

title = {Classification contextuelle pour système d'aide à la décision pour machines-outils.},

author = {Zhiqiang Wang and Mathieu Ritou and Catherine DA CUNHA and Benoît Furet},

url = {https://hal.science/hal-02100713/document},

year  = {2019},

date = {2019-04-01},

booktitle = {16ème colloque national S-mart AIP-PRIMECA},

address = {Les Karellis, France},

abstract = {Dans le contexte général de l'Industrie 4.0, une entreprise de fabrication moderne dispose de nombreuses données numériques qui pourraient être utilisées pour rendre les machines-outils plus intelligentes et faciliter la prise de décision en matière de gestion opérationnelle. L'une des premières étapes de l'approche d'exploration de données est la sélection précise de données

pertinentes. Pour ce faire, les données brutes doivent être classées

dans différents groupes de contextes. Cet article présente une étude comparative d'algorithmes d'apprentissage automatique non-

supervisé pour la classification contextuelle ; qui est utile pour un

système d'aide à la décision pour machines-outils. Les vérifications

par fouilles manuelles montrent que la méthode GMM permet d'obtenir de bons résultats de classification contextuelle, contrairement à celle des K-means.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{fabien_2614,

title = {Vertical electro-polishing at CEA saclay: commissioning of a new set-up and modeling of the process applied to different cavities.},

author = {EOZENOU Fabien and Chel Stéphane and Zhiqiang Wang and Gasser Yves and Jean Pierre Poupeau and Servouin Christophe},

url = {https://epaper.kek.jp/SRF2011/papers/tupo062.pdf},

year  = {2011},

date = {2011-08-01},

booktitle = {Proc. 15th International Conference on RF Superconductivity},

pages = {549-554},

address = {Chicago, ILLonois, USA},

abstract = {Reproducible operation at high performances of sc cavities is required for superconducting linacs, at least in their high energy section in the case of accelerators for proton beams. High beta elliptical cavities are thus of concern, and, to achieve required performances for such cavities, surface preparation including electro-polishing (EP) is recommended. Moreover, for large dimension cavities we consider the EP treatment in vertical configuration (abbreviated as VEP) to be more appropriate. For this reasons, a VEP set-up has been designed at Saclay for the electro-polishing of muticells elliptical cavities. Chosen equipment will make it possible to use a wide range of parameters (voltage, flowrate, temperature, nitrogen inerting) with R&D purpose in mind.

Optimization will be done using modeling with COMSOL software for different cavities. As examples, we present some results for the 704 MHz high-beta SPL cavity and the 1300 MHz ILC cavity and show the influence of cathode shape on both acid flow and electric field

distribution during the process. At last, importance of the size of the cavity will be commented.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}