Manufacturing companies are confronted with a high demand for product individualization and are also operating in a dynamic market environment. In order for a company to strengthen its market position in the long term, it is not only the price and quality of the products that are important – short delivery time to the customer and therefore the performance of the company’s production process is also an essential lever. Effective and efficient production monitoring and control is required to ensure high performance. In the event of poor performance, the causes must be identified and appropriate measures derived in order to improve logistics performance, for example.
However, complex cause-effect relationships in the company’s production process often obscure the actual causes. The increasing availability of data enables systematic, data-driven root cause analysis with the help of logistics models. However, model-based root cause analysis is limited to the level of detail of generally valid cause-effect relationships. The use of data analytics, on the other hand, enables context-specific analyses with a high degree of individuality.
Data analytics should therefore be systematically integrated into model-based root cause analysis as part of a hybrid, automatable production monitoring and control methodology. By transferring the hybrid production monitoring and control methodology into easy-to-use software, small and medium-sized companies are to be supported in analyzing the actual primary causes of poor logistics performance and deriving targeted improvement measures without the need for in-depth specific methodological knowledge.
Data-driven production monitoring and control and its challenges
Production monitoring and control is an essential management function for ensuring efficient production operations. Its tasks include the procurement, analysis and preparation of data in order to promote the target-oriented planning, implementation and management of improvement measures. In this way, production monitoring and control enables companies to position themselves between the conflicting priorities of high logistics performance and low logistics costs. A high logistics performance is expressed by target figures such as short throughput times and high schedule reliability. Low logistics costs are achieved through low inventories and high resource utilization, which in turn can reduce logistics performance.
If one of the target variables mentioned shows deviations, model-based root cause analysis can be used to identify the primary cause with the help of logistics models based on production data. In so-called cause-effect relationshiptrees, generally valid cause-effect relationships of the company’s production process are used to analyze potential primary causes of poor target achievement with an increasing level of detail based on a target value. For example, low schedule reliability in order-specific procurement can be caused by internal delays in incoming goods or by supplier delays. Potential causes of supplier delays are, for example, low supplier adherence to delivery dates or late order release. The introduction of safety times or the automation of order triggering can then be derived as potential improvement measures.
However, company-specific cause-effect relationships are not taken into account in the model-based cause analysis. In the example of order-specific procurement, questions such as: “How is the target value influenced by the company-specific structures in procurement? What influence do individual suppliers or the range of articles have?” remain unanswered. In addition, a high level of experience is required in order to prioritize potential improvement measures in the correct order from a logistical and economic perspective.
Data analytics in production monitoring and control
Data analytics is an interdisciplinary science that aims to generate knowledge from data – for example, non-trivial patterns, cause-effect relationships or trends. Among other things, production data can be used for this purpose. Data analytics combines methods of machine learning and statistics. From a process perspective, data analytics fundamentally comprises the collection and selection of data, its pre-processing in order to subsequently enable data analysis, as well as the interpretation of results and visualization of the new knowledge generated in this way.
In initial case studies, for example, cluster analyses and regression models were used in production monitoring and control to identify key context-specific influencing factors on the schedule reliability of manufacturing companies and to quantify the strength of the influence. Examples of these context-specific influencing factors include the complexity of the production order, the planning horizon, individual process steps, the product type, the intensity of rework and also quality characteristics.
Hybrid production monitoring and control
In order to exploit the potential of data analytics, the Institute of Production Systems and Logistics (IFA) is developing a hybrid production monitoring and control methodology in the deep.control research project that combines the advantages of model-based root cause analysis and the use of data analytics. Components of the methodology are a hierarchical target system of production logistics target variables, a process model for hybrid root cause analysis and derivation of measures, a generic catalog of measures and a data architecture that lays the foundation for implementing the methodology in easy-to-use software.
The IFA is currently working on the detailed design of the systematic integration of data analytics methods into model-based root cause analysis, while deepIng business solutions GmbH is designing the software implementation.
The following application scenario for the production monitoring and control methodology and therefore the software was defined by the researchers for a generic company: The company must initially define the performance targets for the production area under consideration and create an interface to production databases – for example, to the enterprise resource planning system (ERP system) – in order to automatically monitor target achievement in the software and visualize it on a dashboard. If a deviation from the target is detected, the existing production data is automatically pre-processed and merged in a central database in the software to ensure uniform data access. Using drag and drop, the company is able to semantically link the data input with standardized data requirements for root cause analysis.
As part of its research activities, the IFA is detailing automatic data processing and analysis in the data architecture. For the development of the hybrid root cause analysis, the IFA uses both logistical modeling along the cause-efffect relationship trees and data analytics methods to identify the primary causes based on generally valid and context-specific cause-effect relationships. The model-based approach is characterized above all by a systematic delimitation of the production area to be considered, in which the potential causes of the target deviations lie.
The case studies mentioned above show the potential of using data analytics to additionally quantify the strength of the influence of potential causes and thus to be able to draw conclusions about primary causes. deepIng business solutions GmbH creates a visually appealing user interface in the software and enables the company to enter feedback so that the automatic root cause analysis is enriched with empirical knowledge.
Recognize improvement measures and implement them sensibly
Once the company has identified the primary causes with the help of the automated production monitoring and control methodology, the software, which deepIng business solutions GmbH will develop as part of the research project, will recommend which measures should be implemented to improve target achievement. To this end, the software will draw on a generic catalog of measures to be developed by the IFA. The IFA also provides extensive knowledge about the cause-effect relationships, making it possible to assess both the positive and negative effects of the implementation of measures on individual target variables in the area of conflict between logistics performance and logistics costs.
This ultimately enables the company to prioritize logistically and economically advantageous improvement measures and implement them in a sensible sequence.
