Discovery Summit China
12 April 2018
Projections and Encapsulations
John Sall, Co-Founder and Executive Vice President, SAS
As our tables get more numerous, our data gets bigger and our models have more features to encapsulate, we are looking for improvements to organize, contain and model. With the release of JMP 14, there are new ways to do all this.
FIT in JMP: Monitoring giant panda populations
Sky Alibhai, Director and Co-Founder, WildTrack
Zoe Jewell, President and Co-Founder, WildTrack
Binbin Li, Assistant Professor of Environmental Sciences of the Environmental Research Center, Duke Kunshan University
Arguably the world’s most iconic species, the giant panda needs help. And to help this endangered species, conservationists need reliable data on numbers and distribution. That's where WildTrack and Duke Kunshan University scientists come into play. Together with researchers from the Chinese government, these scientists employ a cost-effective, non-invasive and community-friendly footprint identification technique to collect footprint data from giant pandas. Then WildTrack builds highly accurate algorithms to classify individuals and determine their sex, which is the most successful giant panda data collection and analysis method to date. Yet the conservation methods continue to evolve, deploying high-resolution drones to capture footprints in previously inaccessible areas, and exploring the application of deep learning to help filter increasing volumes of data as they come in. This work helps provide a promising future for wild panda populations and the protection of the other species that share this habitat that hosts some of the richest biodiversity on earth.
Applications of JMP in R&D, Manufacturing and Quality Management in the Pharmaceutical Industry
Jiali Luo, Vice President, Boehringer Ingelheim China
Drug regulatory science is a scientific discipline that is concerned with the development of new tools, standards and methods to evaluate drug safety, efficacy, quality control and utility, and is based on data quantification. Drug safety and efficacy is based on clinical trial or observation data, and the study design and analysis methods that are relevant to this type of data fall within the realm of biostatistics. On the other hand, statistical applications related to the research, manufacturing and quality management of drug formulation and processes are classified under quality statistics. As the proportion of generic drugs used by patients increases year by year, so will the statistical requirements for pharmaceutical R&D, manufacturing and quality management domains in various national regulatory bodies. Safety problems of generic drugs are mainly attributed to quality issues during drug R&D and manufacturing, and the application of statistical methods has been proven by traditional manufacturing industries to be an effective solution. In my experience teaching and managing research, I have observed that local enterprises and drug regulatory agencies use statistical tools widely. However, they are not familiar with the prerequisites for using these tools, when to use what tools to collect what data, and the areas to focus on when using these tools. Therefore, it becomes difficult for them to understand the practical significance of key variables and parameters within mathematical formulas, and thereby realize the application of statistical methods for drug development, manufacturing and quality management. JMP is a statistical and data analysis tool that is intuitive, effective, and widely used. Its application can greatly promote data statistics and analysis for drug R&D, manufacturing, and quality management by pharmaceutical companies in accordance with drug laws and regulations. Data analysis and usage can improve a company’s understanding of its products and processes and help achieve consensus with the regulatory agency. This presentation will focus on introducing the application of JMP in quality statistics, with practical case studies in drug R&D, manufacturing and quality management.
Big Data Applications and Analysis in the Pharmaceutical Industry
Peng Xiaodan, Deputy General Management, Operations Management Department, Fosun Pharma
Data acquisition is carried out throughout the entire product life cycle in the pharmaceutical industry, and supports GxP management.The pharmaceutical industry does not lack for data, but how can we effectively analyze and utilize such data? From application in formulation and control of QbT-assisted internal control standards to QbP control of statistical processes; from analysis and application of essential technical process fluctuations to QbD applications in product development, process R&D and analysis methods - how can we systematically analyze these processes, implement continuous iterative upgrading, and utilize JMP to examine models and apply in practice?
Big Data Mining in IC Manufacturing: Defect Prediction Model
Sheng Kang, Senior Manager of the Quality and Reliability Center, SMIC
In 2016, AlphaGo defeated the top Go players one by one. And in 2017, AlphaGo Zero beat all its predecessors through self-learning without human experience. Since then, humans entered the first year of AI. With the continuous development of information technology, the application of data, especially the application of big data in industry, plays an increasingly important role in the manufacturing industry. Integrated circuit (IC) manufacturing is a high-tech-intensive industry. The products’ very small dimensions and high-precision requirements are doomed to have a very close relationship with data analysis. This talk mainly introduces how to use the real-time equipment monitoring data in IC production to find the root cause of defects with data mining and machine learning methods to achieve quality improvement, cost reduction and productivity enhancement. Let’s see the powers and prospects of AI in the industry. We’ll share some actual cases to introduce how to use the model predictive function of JMP in defect prediction and virtual metrology of WAT based on IC manufacturing data. We hope that these cases will inspire.
Commonality Analysis in Manufacturing Applications of Big Data and Predictive Quality Assessments
Cindy You, Quality Manager of Western Digital China (Shanghai) Co., Ltd.
In the manufacturing industry, we are always concerned about process fluctuations. So what are process fluctuations? What types of fluctuations will result in quality problems? How much volatility will result in product failures? The semiconductor manufacturing process is very complex. After these fluctuations are superimposed on the manufacturing process, what kind of results can we expect? This text will introduce the process of implementing data connectivity from wafer to package to finished products within an industrial big data framework, as well as evaluating business applications on this basis. Being able to identify and control key parameters among dozens, hundreds, or even thousands of input variables while ensuring quality assurance and low costs at the same time is a difficult – and pivotal – part of the complex industrial process flow. Commonality analysis can be very helpful in this area, for the purpose of both diagnosis and prediction. In order to enable commonality analysis, we need to focus on all areas of this process, including data collection, data connectivity, algorithm selection and results analysis. The cycle then repeats and iterates itself, making improvements with each step of experimentation and validation. We have chosen to use JMP analysis software to help us achieve these goals.
Data Analysis in NAND Flash Development and Testing
Thomas Chen, Director of Product Engineering and Test Engineering, Yangtze Memory Technologies
When we get into to the age of big data, we look at two important indictors to measure how deep to get involved in this revolution – memory size and network bandwidth. About 20 years ago, was a time when most people in the high-tech industry were still talking about the “computing capability” and were excited about the releases of Pentium serials. Today, only memory size, 128GB, 256GB …, differentiates an iPhone price. People are now more interested in how data is stored and transferred. Memory, especially NAND flash, is widely utilized in an individual’s daily life. From computers to cell phones, from data centers to enterprise grade servers, NAND flash has helped different applications solve data storage problems with higher efficiency, higher security and lower power. At the same time, big data applications also drive the memory industry, making it the first segment in semiconductor, occupying a third of the total semiconductor market. YMTC is the first and only IDM Company in China to develop and manufacture NAND flash. Given the capacity target of YMTC, which is over 300K wafers per month, over 1020 transistors will be tested every month in YMTC’s factory. Accurate and efficient data analysis is one of the most important ways to help quickly develop new products, stabilize mass production line and improve yield, quality and reliability. A couple of examples will be introduced during the presentation on how JMP is utilized during YMTC’s product development stage, and how it can help in future mass production.
Design of Experiments: The Efficient Method to Predict the System Signal Integrity Performance in Volume Production
Feng Wu, Technical Lead, Signal Integrity, Intel
Along with technology and market trends, the data centers, networks and end devices demand more data bandwidth than ever. This has led to the application of the high-speed transceiver to the next-generation silicon chip. Meanwhile, the signal integrity (SI) performance of the high-speed transceiver and its link is a big challenge. SI performance can be simulated, yet it is very slow, due to the complexity of the customer system and its manufacture tolerance. This presentation will show how to use design of experiments to simplify the simulation plan, predict the SI performance and explore the solution space with limited simulation cases.
Efficient and Effective Biopharmaceutical Process Development: A DOE Approach
Xun Liu, Vice President, Shanghai Hengrui Pharmaceuticals; and Chief Scientific Officer of Suncadia Biopharmaceuticals, a subsidiary of Hengrui Medicine
Biologic medicines, also known as biologics, are experiencing a rapid growth in China as one of the most promising frontiers in medicine; the speed to bring a product from DNA to BLA is the key element of success for every organization in the pharmaceutical industry. To reduce the development time and establish all the critical and key process parameters, statistical design of experiments (DOE) is a valuable tool to help achieve those objectives efficiently and effectively. We have used DOE in various stages of process and method development for screening and optimization. For example, cell line engineering applied the principle for cell line screen and selection, cell culture development used DOE for culture media screening and optimization, purification process development used DOE to test characterization range and to establish optimal operating range, analytical method development established assay method with the help of DOE, and formulation development was able to screen and obtain the most suitable final product formulation buffer with a fractional factorial design in JMP. The presentation will provide an overview of these applications with case reports.
Multivariate Analysis of Sensory and Consumer Data With JMP®
Jianfeng Ding, JMP Senior Research Statistician Developer, SAS
Data sets resulting from sensory and consumer studies can be quite large, with many different columns and data types. A variety of multivariate data analysis methods can be useful in the exploration and analysis of sensory data. Over the past few years, many of these methods have been added to JMP. In this paper, we present how to apply methods such as analysis of variance (ANOVA), K means cluster, principal components analysis, partial least squares, multiple factor analysis, multiple correspondence analysis and Text Explorer to sensory and consumer data, emphasizing how each of these procedures operates, how each is interpreted, and how they relate to one another. By illustrating the best methods to address sensory and consumer preference problems, the goal of this paper is to familiarize analysts with sensory evaluation, consumers’ preferences and appropriate multivariate methods so that each analyst can effectively use these methods in JMP for their own sensory studies.
Statistical Applications Used in the Validation of Drug Testing Methods
Dejiang Tan, Deputy Director of Pharmacology Laboratory, Institute of Chemical Drug Control, National Institutes for Food and Drug Control
- Methodological assessments (including validation, testing, transferral, and replacement) and current problems with their use;
- What methods can be used to achieve standards for intended use? How should we analyze them?
- Which performance parameters should be chosen in order to evaluate whether standards have been met?
- How can we reliably and scientifically determine performance parameters?
- Current outlook and recent regulatory improvements.
Statistical Modeling to Empower Fundamental Study in the Chemical Industry
Juliana Zhou, Statistician and Black Belt Coach, Core R&D Dow Chemical
Undersigned and designed data generated from research experiments with proper statistical modeling can help researchers and businesses better understand the fundamental on chemistry and physics. The further studies and learning can empower innovation solutions and services in daily research work. However, there are a lot of tactics to fit models to better represent and accurately reflect the relationship between performances and independent variables, in the areas of 1) data verification 2) data transformation 3) model selection. We will present two statistical modeling cases, one of which shows how we used JMP tools as measurement system analysis (MSA), design of experiment (DOE) and responses transformation to solve the challenges on 1) & 2) . And another case shares how to build and select proper models by multiple modeling methods as Principle Component Analysis (PCA), Standard Least Square Regression. Based on those statistics tools, we proposed structure-property, and property-performance relationships, identify the key contributors and finally predict performance which help to lead success solutions.
The Standardization of Big Data Analysis and JMP® Experimental Flow
Voca Lai, R&D Manager, Taiwan's Semiconductor Assembly & Testing Giant
The recent upsurge of big data is still hot, but most of its applications in the market is limited to summaries or regression and does not include analysis. Without analysis, we cannot do further exploration of root causes and then make decisions or make predictions. The general application of big data is incomplete, and people should not fall to its magical charms. This lecture, based on the author’s practical experience in the semiconductor field for many years, will show you the data analysis steps with a standardized flow from sorting to analysis for decision making or making predictions and then solving the problem. The unique tools in JMP provide a template for a perfect demonstration.
Warranty/Performance Text Exploration for Modern Reliability
Scott Wise, JMP Beacon Account Technical Manager, SAS
With the increasing amount of warranty/performance data available, we have a golden opportunity to learn much more about our designs and products and come up with better reliability/maintainability models. However, this additional information often comes in the form of unstructured text from technicians, researchers and customers. Inability to find groupings and trends in this data can hide the real reliability impacting issues that our traditional data analysis on censored test data, warranty failure time stamps, etc., won't be able to uncover. This session will show how modern analytic software can help provide a way to quickly and easily explore this unstructured text data to gain better insights into our true reliability. We will cover the basics of how to visualize, group, analyze, model and find trends using warranty/performance unstructured text data. Industry examples will show real warranty/performance data where text exploration was helpful in uncovering real reliability trends.