Notes

· We started the meeting by greeting to all participants Matt

· We discussed on the data download and extract from the drive.

o We updated that now we have Apex-Ml-Dcu-Data database and 'MLchainwearlinkdata read permission.

o Matt asked, is database is correct and complete now.

o We updated that now we are able to extract database properly but collections we have to verify it and keep him posted.

· We had long discussion about the past three days Wednesday, Thursday and Friday progress.

o Anthony and Matt emphasize that when database was not available then what team did in past three days.

o We updated that we worked on the data cleaning and query part for the data which we were having.

o We provided details later on Friday with the details what we had done in past three days.

· We discussed the primary goal, which is to predict conveyor chain issues using provided data.

o We started to give detailed review of the previous work, their structures, and relationships was conducted

o We explained our understanding the data's organization and possible anomalies.

· We updated that we recognized the importance of data preprocessing for accurate modeling.

o We discussed on to cover handling missing values, normalization, and feature engineering.

o The need for accurate timestamps was highlighted for time-dependent analysis.

o Discussion continued on the data preprocessing, emphasizing the importance of accurate timestamps.

· We highlighted the need to resolve issues related to timestamp alignment to ensure data consistency.

o Discussion on Lubrication Cycles:

o We expressed concern about lubrication cycles 3 and 4 consistently showing as false.

o We mentioned that cycles 1 and 2 have both true and false values.

o Speculation arose that the number of loop monitors may be related to this issue.

· Main Drive and its Functionality:

o The role and functionality of the main drive were discussed by Ankush.

o We questioned whether it was related to electricity due to its on/off behavior.

o It was noted that the main drive might switch to a backup drive in certain situations.

· Link Value Feature and its Significance:

o We discussed on link value feature and its relevance.

o Matt Knysz explained that the link number corresponds to specific links in a chain.

o The value represents the amount of stretching (measured in hundredths of an inch).

o We inquired if this value played a role in machine wear and maintenance.

· Matt will contact Tony for clarification on some of the questions raised during the meeting.

o We continued with a discussion about chain link pitch and measurement.

o David provided a visual explanation of a chain link and how it's measured.

o He emphasized that the value point 0.5 indicates the distance between pins as they pass by and how it reflects chain wear.

· We asked query regarding the significance of the value point 0.5.

o David and Matt Knysz clarified that it measures the extent of chain stretch.

o A value of 0 means no stretching, while 0.5 indicates stretching has occurred.

· Matt Knysz highlighted that detecting a rapid increase in chain wear over time is crucial as it could signify issues with the chain.

o David shared a real-life example of how chain wear patterns can indicate problems with lubrication and maintenance.

o We discussed the need for data table joins, as the goal is to create a single CSV file for machine learning purposes.

o They identified common factors for joining tables, including Factory ID, DCU (Data Collection Unit) ID, and Conveyor ID.

o For joining monitor-related tables, Monitor ID was recognized as a key factor.

o An important distinction was made between "air" and "take up" sensor types. Anthony Brown clarified that "air" sensors measure air pressure on the chain, while "take up" sensors,

§ which are part of the take-up mechanism, measure chain distance. It was noted that these sensor types are mutually exclusive within the data.

· We discussed on bad data flag and utilize machine learning to detect issues related to chain wear

o David explained regarding chain wear measurements.

o We discussed the significance of the "point 0.5" value in wear measurements, understanding it as an indicator of chain stretch.

o They emphasized the importance of this value in detecting issues with chain wear.

o Matt Knysz highlighted that negative values are unlikely to occur in future data due to system improvements.

· David shared a case study illustrating how the data can reveal chain wear patterns and incidents of broken links.

o It was agreed that a rapid increase in chain wear over a short period is indicative of a problem.

· Tony explained the significance of "air" and "take up" values, clarifying that "air" indicates air pressure on the chain, while "take up" refers to distance measurements for chain take-up.

· We discussed on to focus on improving data analysis to detect issues promptly.

o Matt updated that he will send schema analysis document which will help to make join and cleaning the data.

· We summarized action items and discussions and thanked Matt and Anthony for joining meeting.