For past 20+ years client has been manufacturing auto parts for world’s leading auto manufacturers worldwide. Client takes pride in producing quality parts in the frequency they produce to meet the customer’s demand.
1 - Different combination of raw material from different vendors causing machines to
loss of productivity.
2 - Line workers did not have visibility to the quantity, quality, and acceptance criteria of production.
Cominfo was engaged to developed machine learning and predictive models to ensure management is notified prior to breakdown occurs so preventive measures are taken, and vendors are notified about their material in real time, so they can take corrective actions as well. Secondly, develop visual analytical solution, which could be displayed on big screen in manufacturing floor to show what is been produce, what’s coming up next, who is testing, and what are the test results.
Cominfo developed multiple predictive models evaluating over 500 data points in real time collected from 8 source systems such as relational databases, IoT, flat files, and web logs. Our Predictive analytics developed using R and Python enabled client to make better use of downtime due to breakdowns. Automating the analysis of data from over 8 source systems helped client cut down downtime by over 60%. Essentially, client was able to predict when the machines may need to be brought online or shut off to prevent an issue. We also developed open APIs to talk to vendor systems, so they are notified about the breakdowns cause by material or human errors. We developed analytical reporting and data presentation layer using Microsoft Power BI platform. Which were refreshed every 2 mins using Microsoft Flow. Reports were broadcasted to over 20 manufacturing plant located in 7 different countries. Row Level Security (RLS) was also implemented so data could be segregated by plants. We took in the consideration that line workers wear gloves during production, so we used big square buttons and eliminated any scroll bars resulting in all the data presented in a single screen. Client was able to produce tier 3 parts every 8 minutes compare to every 10 minutes net result improvement of 20%