The overall aim of the project is to develop an intelligent production planning for machinery and plant engineering. It aims to support companies in creating transparency in order management planning processes and making optimized planning decisions. Approaches from the field of artificial intelligence (AI) and machine learning methods are used for this purpose. Intelligent production planning provides an AI-based representative model to identify similar customer orders, a forecast-capable simulation module to create transparency concerning planning decisions, and an AI-based planning module to identify optimized decisions. The three modules are consolidated in a suitable planning process.

Motivation

This project is necessitated by the volatile, uncertain, and complex framework conditions to which manufacturing companies are currently exposed and under which they have to make ambiguous decisions. These framework conditions can be summarized under the acronym of the VUCA world and pose a challenge to existing planning approaches. On the one hand, increasingly volatile markets and external shocks are leading to frequent new planning and rescheduling. The coronavirus pandemic and the Suez Canal blockade are just two examples of recent events that demonstrate the impact on closely interlinked supply chains. On the other hand, customers expect a wide range of variants, distinct customization options, and fast delivery times. This results in complex material flows and planning decisions that have to be made under ambiguity, which increases the complexity of production planning. Within these challenges, the machinery and plant engineering (MP) sector is of great importance, as it is especially affected by these developments. The high level of customer orientation in MP leads to a diversity of variants that is more pronounced in no other industry. For efficient production planning, the material, time and space requirements of a customer order must be derived with a high degree of manual effort. The available capacity is then only scheduled in compliance with complex planning restrictions, for example due to limitations regarding the permitted sequence of different product variants. If a possible resource allocation plan has been found, the effects on operations are currently uncertain, such as the occurrence of production bottlenecks. Finally, there is a lack of decision support in order to use the generated transparency to identify optimized decisions, as planners can only handle the complex solution space to a limited extent without support. While the MP is particularly affected and challenged by the complexity of production planning due to the points outlined above, it represents an important economic sector for Germany and Bavaria and contributes significantly to strengthening the location through its export orientation.

Objective

In order to meet these challenges and secure competitive advantages in an increasingly complex and dynamic environment, flexible and efficient production planning is required. The use of AI-based planning methods as part of intelligent production planning for the MP offers considerable potential for this. As a tool, intelligent planning supports manufacturing companies in identifying resource requirements, generating transparency with regard to the effects of planning decisions and providing decision support for optimized planning.

Approach

The intelligent production planning to be developed in this project includes a representative model, a simulation module, and a planning module for decision support. The representative model uses a large number of features to describe the specifications of a customer order as input variables, such as the type of tool used in a system. Similar past customer orders are identified using advanced ML and AI methods. The material, time, and space requirements of similar customer orders then serve as predictors of the requirements of new customer orders. The simulation module uses these as input variables and makes it possible to forecast the effects of different planning alternatives on the production system. The AI-based planning module is used to identify optimized planning decisions. By using reinforcement learning (RL) as a sub-area of ML, optimized decisions can be derived from the previously simulated findings and transferred to planning.

In addition to developing these central components, the project aims to integrate the technical solutions into existing planning processes. Finally, the project results are to be implemented and validated by the application partner in addition to being designed. Specific use cases for this purpose have already been identified in workshops.

Acknowledgment

The InProdMa project is funded by the Bavarian State Ministry of Economic Affairs, Regional Development and Energy (StMWi). The project sponsor is VDI/VDE Innovation + Technik GmbH. We want to thank the partners for their excellent support.