2020/01/06: Data Description Update

Data Description—Overview: The number of Vp3AaLm is changed into 10.
Competition page has been updated with several supplement and changes in content.

Background

Bosch is a German industrial enterprise, specialized in a number of industries, including automobile, smart traffic, manufacturing, consumer goods, energy and construction technologies. Bosch is the initial champion of “Industry 4.0” platform. At present, it has more than 100 4.0 pilot projects in over 250 factories worldwide, including 13 in 9 production bases in China.
 
Conventional manufacturing factories have a complex product inspection workflow. If failure is found and rework is needed, there is still uncertainty regarding where the product can pass after rework. Machine Learning approach can play a role in predicting failure rate and improve process and efficiency. Such a problem prevail traditional manufactures. Therefore, it would be generally applicable to establish a failure prediction model to assist them to optimize production.
 
In the product series used in the competition. Bosch factory usually implements certain steps of inspection the number and phases vary from product to product. Each step contains certain number of phases. In each phase, the inspection feedbacks a result (1 for pass and 2 for failure) and parameters (value or string). If in all inspection phases, the inspection result is pass, then the product item is rendered as pass. If the inspection result is failure in certain phase, the inspection process will stop and the product item is regarded as failure. The competition expects machine learning algorithms to predict failed products in advance, so as to save inspection time.
 

Task Description

The competition expects participants to predict product quality based on INSPEC Dataset. It contains production records of a certain family of Bosch product. Participants need to establish and improve models from training set and predict final inspection result (pass or failure) according to the finished phases of target product items.
 

INSPEC (Industrial Specification Inspection Data)

Beijing Academy of Artificial Intelligence (BAAI), together with Bosch, releases inspection data of a Bosch product family in recent years. It contains production records of around 30,000 production items, including thorough description of each procedure and relevant parameters. Compared with counterparts, it has distinctive features.
First, INSPEC Dataset extracts real desensitized production records of top manufacturing factory in the world. It re-presents on-site procedures and steps of product inspection. Second, INSPEC offers a detailed illustration of inspection results and parameters, including product ID, item, phases, results, specification, etc. High data dimensions increase competition difficulty as well as the possibility for building robust models.
 

Discussion

For any problem, please send email to support@biendata.com, or join our WeChat group.
 
To join the WeChat group, participants can send message “name+org+bosch” to our competition assist (ID: shujujingsai), if you have already befriended the account.
 
For new participants, please add “biendata02” on WeChat (scanning the following QR code) with the info “bosch+org+name”. wrong info won’t be accepted.