Research On Dynamic Prognosis And Predictive Maintenance Scheduling For Health Management Of Manufacturing Systems

With the increasing international competition and advanced manufacturing patterns,modern manufacturing enterprises equipped with multi-unit manufacturing systems are facingthe challenges of growing production scale, diversified equipment health, complex systemstructure, random customer orders and dynamic decision-making. As the core ofhealth management theory, predictive maintenannce (PdM) strategies make maintenancearrengments according to the machine health trends. These maintenance scheduling havethe integrated remarkable reliability, efficiency and economic advantages in protectingmanufacturing systems, reducing downtime costs, increasing equipment availability. Thepredictive maintenannce strategies have important scientific significances for supporting andguidancing manufacturing enterprises to implement effective machine health management.This paper studies PdM strategies for series-parallel manufacturing systems consist ofindividual machines with different types. A health forecasting method has been improved forsupporting PdM scheduling. The maintennace decision-making process can be divided intothe machine-level optimization and the system-level scheduling. Based on machine healthdegradations, different system structures and various production plans, dynamic PdMstrategies are developed to reduce total system maintenance cost, decrease maintenancescheduling complexity and provide real-time PdM scheduling. In the reaserach on dynamicprognosis and predictive maintenance scheduling for health management of manufacturingsystems, this thesis focuses on the following parts: (1) For the machine health prognosis, this paper proposes a W-variable Forecasted-staterolling grey model (WFRGM) to provide efficient and accurate machine health prediction,while effects of influencing factors such as operating load are analyzed. In this greyforecasting model, generating coefficient W values corresponding to variable operating loadsare dynamically generated to overcome the shortage of a static W value. This improvedrolling grey forecasting model offers a potential to predict the health trend for supportingpredictive maintenance schedule.(2) For the machine-level optimization, a model-iteration algorithm using amulti-attribute model (MAM) is developed to optimize the maintenance schedule for eachsingle machine. The multiple attribute value theory is used to determine the optimal PMintervals in different maintenance cycles. Furthermore, a hybrid hazard rate recursionevolution is developed, where both maintenance effects and environmental condition areintegrated into maintenance scheduling. Effective and sensitive analysis shows that theproposed MAM results in a noticeable increase in the availability-benefits and thecost-benefits.(3) For the system-level scheduling of flow production, a maintenance-driven opportunity(MDO) decision-making strategy is proposed by considering both machine degradation andsystem structure. According to real-time single-machine schedule, the maintenance timewindow (MTW) programming is applied for series-parallel manufacturing systems. Adowntime caused by a machine could be used to perform PM on non-failed machines, whileunnecessary breakdown of the whole system should be avoided. The aim is to systematicallydetermine the system maintenance schedules that optimize the cost effect and decrease thedecision-making complexity.(4) For the system-level scheduling of batch production, a production-driven opportunity(PDO) approach is developed for multi-unit manufacturing systems. Advance-postponebalancing (APB) programming is presented to make real-time maintenance schedules that not only satisfy the requirement of no-disruption to batch production, but also reduce the overallmaintenance cost. This joint scheduling of dynamic maintenance and batch production is aviable and effective policy that can eliminate unplanned downtime during batch production,lead to a significant cost reduction and overcome the complexity issue of maintenancescheduling.The dynamic prognosis and predictive maintenance scheme achieved by using theproposed methodologies are demonstrated through a case study in a hydraulic steering factory.This research work can provide maintenance decision-making support for modern industrialenterprises. The developed methods and strategies can promote equipment management,health prognosis, machine-level and system-level PdM scheduling for multi-unitmanufacturing systems. The research results can help assist a plant manager in making adynamic maintenance plan to improve equipment reliability, reduce maintenance cost,increase system availability and enhance enterprise competitiveness. Therefore, this researchcan provide great support to develop machine health management and promote the industrialproductivities.