全球制造商在数据主导的产品测试现代化后提高了生产线生产率

The refrigerator testingprocess — before the project Our client designed their refrigeratortesting method 20 years ago. It wastime for an upgrade. The refrigerator units are tested in custom-designed stands, so only a limited numbercan be tested at one time. Testing each pieceof equipment can take up to four hours. The business gained a competitive edgeby deploying advanced data solutionsto improve production methods. The testing relied on static standardparameters and metrics — not actual datacaptured from the machines. In addition: Our customer is a leading manufacturerof equipment and value-addedservices integral to the manufacturingpractices of semiconductors andmicroelectronic devices. The product in thiscase study is a closed-cycle heliumrefrigeration system, typically used inlaboratories for the direct cooling ofexperiments. •The testing process acceptance criteria,metrics, and limitations were based onpartial data and educated guesses. •The process lacked data logic for many ofthe rules and decisions in the testing steps. •Only a few engineers fully understood thetesting process. Technical manuals existed,but the testing process itself wasn'tdocumented. When our client began looking for ways tostreamline manufacturing practices, theyfocused on the refrigeration testing process.It was a solid business decision as theirrefrigeration units are used in a variety ofcommercial and research applications, soincreasing product output would give them acompetitive advantage. They wanted to fix the testing methods andcreate advanced analytics to predict potentialrefrigerator equipment failures. They had limited success with partners in thepast and wanted to find a company withexperience in manufacturing. They choseSoftServe because of our reputation forimplementing successful data analysis andprediction solutions for manufacturingcompanies. The plan was to improve the refrigerationtesting process using advanced datatechnology solutions. Intelligent data captureand data analysis of the testing process plusmachine learning models shortened therefiguration testing time and sped up productproduction. A project is only as successful as the quality of its data Data was collected throughout themanufacturing process. However,the refrigerator data was scatteredamong many databases, files, andprintouts. We knew of at least tendata-storing systems, and each needed tobe integrated. SoftServe started the project by conductingan extensive data review to reveal datamerging and reliability pain points in ourclient's ecosystem. Althoughtime-consuming, our team cleaned,contextualized, and merged the datarequired. The success of any production optimizationproject relies on meticulous planning anddiligent data organization. Here's ahigh-level glimpse of the project steps; theoffsite work was often a trial-and-errorprocess until we found the correct analysisand modeling criteria. Finding refrigerator equipment historydata, an essential part of building predictiveanalytics, took a lot of effort. Their datamanagement problems had to be improvedbefore any analysis could begin. Onsite visit: Knowledge transferwith experts Clarify and scopeproject goals Gain main data flowdetails and dataexplanations Prioritize thehypothesis of timereduction Offsite work: Identifying datainsights and datachallenges Preparing reports,models, and dataflow suggestions Data modeling The refrigerator testing process now — a resounding success Our client’s executives were pleased with the impressive and extensive projectresults. As we wanted you to fully understand the scope of improvements that occurwhen a data project is done right, we broke down the results and successes intothree categories:data analysis,project outcomes, andgraphical interfaces. Data analysis In our analysis, we explored reducing therefrigerator testing time using pass/failpredictions during the test and in thefield. Valuable analysis during this processprovided exceptional results, like: •Building several machine learning modelswith a 67% prediction accuracy rate. •Predicting correctly in 73% of caseswhether the test will pass or fail in futurestages. •Discovering refrigeration unit retesting andreloading can increase testing productivityby 10%-15%. •Realizing 10%-15% cost savings as a resultof optimizing the testing flow. •Eliminating unnecessary retesting ofrefrigerators saves at least 6% of the totaltime spent on testing. •Designing new refrigerator stands basedon data analysis saves even more testingtime. Project outcomes The goal of this project was to improve the refrigeration testing process. Our clientconsidered this the first step to becoming a data-driven company that uses applieddata science techniques. The business intelligence and machine learning modelscreated during this project put them on the right trajectory. •The client noticed the difference betweenthe behavior of refrigerators that reachedtheir warranty and those that came back