Monitoring machines to make a smarter shop
Problem statement
There is a nano precision CNC machine owned by the Manufacturing Innovation Network Laboratory, whose main task is ultra-precision machining. To guarantee this CNC machine to work correctly, the surrounding conditions (e.g. temperature and vibrations) must be monitored and controlled to stay within a certain range of values. Otherwise, the precision of the manufactured parts could be affected, which may lead to scrapping parts and may cause damage to the machine. Currently, it is hard to identify the abnormal operation of the CNC machine without a manual check by professionals. Lack of real-time environmental and utilization data of the CNC machine has already led to a huge loss of time and unforeseen additional expenses, affecting the quality of work. In commercial applications, lacking real-time utilization data is impactful: a manufacturing manager returning to work may not know if he or she will walk in on a finished job or a destroyed tool. A quality supervisor will not know whether a defect caused by increased room temperature impacted the last 5 parts produced or the last 500. A maintainer won’t know when machine part replacements need to be made or services performed. All of these unknowns compound to individual and organizational anxiety, which not only causes undue stress on the worker but also wastes worker brainpower on mundane risk mitigation where it could be used on process improvement and product innovation. It could be argued that the resources invested in mitigating this problem could be kept on hand as a reactive measure to downtime issues. However, according to the International Society of Automation (ISA), the average factory loses 5% of its productivity due to machine downtime, with the most extreme cases approaching 20%. At large plants, this can amount to tens of millions of dollars. Smaller plants are even more adversely affected since they likely have fewer units of redundant machines. According to the ISA, the average cost of any given downtime incident is $2 million. Overall, the relative cost of implementing a predictive solution to this problem is far lower than the relative cost of implementing a reactive solution.
Team members
Rakip Alimi – leader
Mark Vandenberg – communicator
Amy Conard – accountant
Antonio Rubio – admin
Client
Sangkee Min, UW-Madison