Is it possible to use machine learning to make failure prediction?
Predictive modeling to anticipate equipment downtime is referred to as failure prediction. These models are based on data collected from past failures of a given equipment (or similar ones). Machine learning is well suited to model current equipment behavior and its potential breakdowns. Production equipment failures can be anticipated and maintenance can be scheduled before the problem happens, avoiding unnecessary costs.
Manufacturing, Maintenance and Operation Managers can benefit from predictive models. They are not data scientists and may not have the required skills in machine learning or coding experience to develop them from scratch.
They collect, in the course of their daily activities, considerable amounts of data as most machines are equipped with sensors. Data such as temperature, pressure, moisture, exposure to light, duration of use since the last downtime, are typically collected. Even though often considered as Big Data because they range in the millions of measures over the course of a year for instance, the particular case of failure prediction falls into the Small Data category as it has usually only occurred a very limited number of times over the same period. Traditional machine learning tools work well with Big Data but do not perform well for prediction of Small Data (failure prediction) within a batch of Big Data (unbalanced dataset).
MyDataModels allows domain experts such as manufacturing managers, maintenance managers, operation managers, facility managers to automatically build predictive models from their Small Data. They can use their raw data directly: no normalization, no need to handle outliers or engineer new features. Thanks to this limited data preparation, the predictive results from an historical dataset are obtained with a few clicks in less than two minutes on a standard laptop.
MyDataModels brings a self-service solution for those who have Small Data and no data scientists.
Manufacturers are constantly under pressure to stay competitive by optimizing processes, improving efficiency of aging infrastructure, reducing unplanned downtime, sudden failures and maintenance costs.
A CXP Group study found that 95% of companies describe their current maintenance processes as ‘not very efficient’. Production managers and machine operators operate normally on scheduled maintenance to prevent downtime. Unfortunately, 50% of these preventive maintenance activities are ineffective, i.e. they happen at a time when the machine does not need it.
“Predictive model reached a 96% accuracy rate”
In such a failure detection use case based on real manufacturing equipment data, the results obtained by using MyDataModels’ predictive models are more than helpful with a 96% accuracy rate. In 96% of cases, a breakdown was predicted before it happened.
By using an automated machine learning solution like TADA, companies can now proactively identify problems by running a root cause analysis and push fixes including spare-parts, software, hardware and firmware to eliminate possible points of failure or degraded performance – ultimately increasing customer satisfaction and competitive advantage.