A Data-Driven Predictive Maintenance and Process Monitoring System for Grinding Machine

For the manufacturing of products with a good surface finish such as brake disc plate for automobiles, the grinding process is used. Manufacturer’s consistency in producing good quality products put them ahead of their counterparts. Due to machine failures and their unscheduled maintenance halts production and hamper productivity. Process failures fuel quality failure occurrence. Study shows among the industrial failures, machine and process failures constitute more than 50%. In the presented work, a problem-solving approach for these two problems in manufacturing has been explained and supported by the case study of real industrial products and experiments. A data-driven model consisting of an accelerometer has been developed for the collection of vibration signals from the bearing spindles of the grinding machine. The framework for the automatic signal collection, feature extraction, health indicators derivation and failure prediction based on learning algorithms has been explained. Conventionally, these failures are detected by machine breakdown or operator’s intuition of judgment of the component sound. A detailed explanation of the collection and analysis of the vibration signals with the use of engineering tools has been given. Additionally, for the process monitoring of the grinding, the dependency is completely on the operator who characterizes the failure based on his experience or from the feedback of the produced product. A process monitoring system with the accelerometer in consideration of surface failures like chatter marks has been developed. The developed module reciprocates the real process behavior of the current product and alarms the operator if the corresponding parameter goes beyond the set threshold limit. The monitoring of the vibration behavior of the workpiece and the grinding wheel is the best way of characterizing the surface finishing of the product. Later this module is used to build a robust database of good products and bad products with corresponding process parameters and their vibration features. A rule-based process control strategy is also proposed, which suggests the operator of the machine with the best set of parameters, after searching the database, to avoid the failures. Experiments in the production line of a brake disc plate manufacturing company have been conducted to implement and validate the developed module of predictive maintenance and process monitoring. The developed system after its implementation is very helpful in terms of reducing the number of scrap parts and machine breakdown instances in the factory and subsequently saves a lot of cost as well as the resources. Presented work after its full implementation in real practice will be a boost to the field of smart manufacturing and its complete automation.