Our Design Direction

• To facilitate an analytical, "augmented intelligence-based" approach (i.e. using decision-supporting tools and actionable intelligence and data), our team opted to use predictive models and artificial intelligence tools, the efficacy of which has been validated by research/literature intermittent demand Prediction and smart search engine implementation. This consists of 3 three solution segments or subsystems that centrally integrate and process information to reduce iterative communication dependencies.

1. 1. Defect Occurrence Prediction: Predict the occurrence of defects and possible spare parts requirements.
2.  
3. 2. Inventory Analysis: Determine the appropriate inventory optimization strategy based on the spare demand forecast in the defect occurrence forecast.
4.  
5. 3. Resolve Action Recommendations: Use the engineer's past experience to effectively determine the best response to system flaws similar to those encountered before.

Our Design Principles

Our system architecture (displayed on the left) illustrates how our prototype is built upon R and Python APIs, and, uses AWS as the backend and ReactJS as the frontend.

Main Features

Use Cases

Automated Defect Occurrence Forecasts

The goal of this system is to use past data and trends to predict the occurrence of system failures and prepare to replace defective system components that caused said failures.

Several forecasting methods were tried, developed, and improved through a series of programs to efficiently format maintenance log data and recover as much information as possible to aid in modelling as well as evaluation and selection criteria in terms of accuracy performance metrics. The development of the solution and the refinement or update of the forecasting methodology resulted in AutoARIMA being identified as the overall best choice forecasting algorithm to help operations managers obtain effective and accurate forecasts of demand for spare parts and appropriate management strategies.

Smart Inventory Recommendations

The second system uses the forecast demand for each spare part generated by the previous subsystem defect occurrence forecast. It applies the (S-1, S) inventory strategy to provide the main function to help department managers choose the appropriate spare parts inventory management strategy.

The two subtasks involved in developing the spare parts holding analysis method are: (1) checking whether the current spare parts holdings are sufficient to support failure recovery within a specified time, based on the results of the prediction of the occurrence of defects, and (2) recommending that if insufficient, then the number of spare parts needs to be supplemented. 

Knowledge Management System

Alongside providing spare inventory management assistance, there is a need to address the tedium and inefficiency users face of having to look through countless lines of data to find recommendations for identified defects. The third system completes the solution to tackle the needs which were identified from interviews and research.

Through exploration of methods for text processing, feature extraction and a smart search algorithm, this subsystem detracts from the dependence on individual field engineers’ experience and capitalises on records of past repairs. Such a model would involve building a smart search engine using the dataset provided to us.