Page 60
www.us-tech.com
Partnering February, 2019
IBM and Optimal+ Bring Cognitive Analytics to the Factory Floor
By Michael Schuldenfrei, Technology Fellow, Optimal+ M
any companies are now real- izing the benefits of using machine learning to solve
complex problems, where traditional methods of analytics fail. Companies are rushing to recruit data scientists at an unparalleled rate. Specifically, in the semiconduc-
tor and electronics manufacturing ecosystem, there are many applica- tions that can benefit from machine learning. These include root cause analysis of yield loss to accelerate NPI; reducing expensive test steps by using fab and test data to predict failures ahead of time; and reducing field RMAs by predicting quality is- sues before shipment. Machine learning can also help
to enhance quality by detecting com- plex scratches or cracks in products, enhance coverage and reduce work for a variety of visual inspections and improve reliability of electronic sys- tems, based on the quality of their components.
Not Just a Trained Model There is lot more to machine
learning than just creating or train-
ing a model. The real challenge lies in the complex lifecycle of a machine learning project, which involves mul- tiple steps and often crosses organi- zational boundaries. Broadly speaking, the lifecycle of
a machine learning project consists of four tasks: data collection, preparation and cleaning; model development and validation; model deployment to pro- duction; and ongoing monitoring to en- sure the model stays valid. In fact, data scientists report
that they spend only a fraction of their time developing the algorithm or model. The rest of the time is spent on a variety of “menial” tasks necessary to support the complete lifecycle — collecting, connecting and unifying data. Once models are built, data scientists must work with SW and IT teams to put them into pro- duction, which is usually a complex and inefficient collaboration. Each of the four steps has its
own complexities in a real production environment and the overhead of managing them is daunting. The problem is even more complex when the model must execute in out-
sourced manufacturing operations, as integration must be performed on every manufacturing floor for the process to work seamlessly. IBM has tools to manage the full
cycle of machine learning projects. By partnering with Optimal+, customers who already run Optimal+ Innovation Platform for data collection and ana- lytics from across complex, out- sourced, mission-critical manufactur- ing environments can now benefit from the power of IBM Watson Studio (1 and 2) to build, test and deploy ma- chine learning solutions.
Generating Quality Data Optimal+ already provides data
scientists with tools to create queries that can generate large, harmonized datasets of manufacturing and test data by using the Innovation Plat- form. Not only does the platform pro- vide the necessary raw data, it can also perform a variety of data prepa- ration tasks to make the data more meaningful to a machine learning al- gorithm. This includes prefiltering data on a wide range of statistics, such as
entropy, to ensure only relevant measurements are included. It also applies advanced data enrichment algorithms, such as wafer geography, for semiconductor data (e.g. in which ring/zone each die is located). The software also derives information from quality algorithms, such as out- lier detection. From within the Optimal+ ap-
plication, the user can export this query at the click of a button as a Jupyter notebook, together with all of the required libraries, and launch it in Watson Studio. The user can then build, train
and test models in Watson Studio, leveraging all of the libraries and ca- pabilities of the Watson Studio plat- form, such as its built-in support for Python, R, Spark, and Scala. The notebook becomes a collaborative project in Watson Studio that an en- tire team can use to develop and test the model. Once the model is ready, a mon-
itoring rule can be created in Watson Studio. First, the Optimal+ platform is configured to export a daily snap- shot of production data to Watson Studio. The built-in scheduled evalu- ation features of Watson Studio are then used to evaluate the model on current production data and ensure the model is not going stale. Finally, the approved model can
then be exported into an Optimal+ “Virtual Operation Rule.” Optimal+ rule simulations can be used to test the rule on additional data sets and the final rule can be published at the click of a button to the relevant manu- facturing floors, where it is seamlessly integrated with shop-floor systems. For data scientists, the solution
is a gamechanger. They can build, test and monitor models in IBM Wat- son Studio. At the same time, they can leverage the Optimal+ infra- structure to collect and harmonize their data (both before the model is developed and during execution), managing the deployment and inte- gration across complex, outsourced manufacturing operations. At long last, data scientists can enjoy the best of both worlds. Contact: Optimal Plus, Ltd.,
2107 North First Street, Suite 310, San Jose, CA 95131 % 1-800-685-2127 E-mail:
info@optimalplus.com Web:
www.optimalplus.com r
See at IPC APEX, Booth 3513
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108 |
Page 109 |
Page 110 |
Page 111 |
Page 112 |
Page 113 |
Page 114 |
Page 115 |
Page 116 |
Page 117 |
Page 118 |
Page 119 |
Page 120 |
Page 121 |
Page 122 |
Page 123 |
Page 124 |
Page 125 |
Page 126 |
Page 127 |
Page 128 |
Page 129 |
Page 130 |
Page 131 |
Page 132 |
Page 133 |
Page 134 |
Page 135 |
Page 136
