راه حل های سرمایه گذاری دیجیتالی برای برنامه ریزی تولید و کنترل یکپارچه

Digital enterprise technologies combined with sophisticated optimization algorithms can significantly contribute to the efficiency of production. The paper introduces a novel approach for integrated production planning and control, with the description of the mathematical models and solution algorithms. The deterministic optimization algorithms are complemented by a discrete-event simulation system to assess solution robustness in case of disturbances. The methods are illustrated by describing two prototype systems and by some experimental results obtained in an industry-initiated project.

The concept of digital enterprise – the mapping of the key processes of an enterprise to digital structures by means of information and communication technologies – gives a unique opportunity for planning and controlling the operation of enterprises [22]. Digitalized solutions are capable to connect customer order management with production planning, scheduling and control (PPC in short). On the other hand, the importance of optimization in decision making at various levels of an enterprise has greatly increased over the past decades, as companies invest in complex advanced planning and scheduling systems to replace their out-dated material requirement planning software. Digitalized data and sophisticated algorithms together can provide additional competitive advantage which cannot be achieved by applying solely the latest production technology. The scope of our work is set to complex engineer-to-order and make-to-order production, where products (like turbines, assembly lines, etc.) are traditionally associated with high quality, advanced, cutting-edge technology. These industries usually require skilled and expensive human workforce. Recently, business focus has shifted from selling products to supplying a combination of products and services (like engineering design, installation, on-site customization, maintenance). Such “extended products” are highly customized and their value is sensitive to the time of the delivery. Human resource intensive repair and recycling activities have to be planned together with normal production. Due to complex production processes and long lead times, production of components often starts before the overall design has been completed, or customization is executed in parallel with some production activities. In our target sector, a project-oriented approach is taken in general for planning and controlling operations. Relying on the conventional wisdom that it can never be wrong to get work done early, existing project planners typically try to sequence activities as early as possible, subject to technological constraints and resource availability profiles. Such methods are only capable of finding a particular solution, without the ability to explore and evaluate alternatives. Hence, they cannot be used for optimization. Further on, manual intervention is typically required to deal with overloaded resources and violated deadlines. According to our experience, medium-term production plans are re-adjusted manually time and again, and less than 50% of the original plan is executed finally. Our first objective was to develop intuitive and flexible models and fast, reliable solution techniques that scale-up well also to large production planning and control problems. Hence, we tackled both the medium-term production planning and the short-term detailed scheduling problems as resource-constrained project scheduling problems (RCPSPs) [7]. The solution methods had to respect all the main temporal, capacity and material availability constraints and find an optimal trade-off between various costs and due date performance criteria. Our second objective was to find novel, aggregate formulations of the production planning problem which ensure the integrity of results that are generated on two different hierarchical levels, on various horizons, by using distinct models and solution algorithms. Following the usual planning hierarchy, production planning determines what to do on the medium-term so as to achieve high-level business objectives. On the other hand, scheduling is responsible for refining a segment of the production plan into a detailed and executable schedule. However, since the two levels use different models, it is open whether production plans can really be unfolded into feasible, executable detailed schedules. An essential practical concern – especially in make-to-order production – is also that the representation of the planning problem should be generated automatically, from data readily available in de facto standard databases of production information systems. Both planning and scheduling problems are burdened by various uncertainties like estimated resource needs, uncertain capacity availability, unspecified activities due to evolving problem definition, uncertain orders, hypothetical projects, uncertain processing times, as well as unreliable delivery dates of necessary components and materials. Any method that neglects these issues is prone to generate fragile solutions. Disruptions hardly stop at the boundaries of the particular shop-floor of an enterprise; they spread upwards in the decision hierarchy and even to other members of a production network. Due to reasons of complexity, the direct inclusion of any main uncertainty factor into our RCPSP models was out of question. Hence, our third objective was to assess the sensitivity of deterministic solutions and improve the robustness of production schedules by using discrete-event simulation techniques. In the sequel, first we present our integrated approach to production planning and scheduling (Section 2). While we capture problems at both levels of the planning hierarchy as RCPSPs, the details of the models and the solutions techniques are fairly different. Section 3 gives an account of how we built simulation models automatically from common master data and applied simulation techniques for assessing the sensitivity of production schedules. This work had a strong industrial motivation: in Section 4 we present two prototype systems developed for two enterprises operating in the engineer-to-order and make-to-order sectors, respectively. Finally, we conclude the paper in Section 5 and give an outlook for related research activities.

The main goal of the paper was to demonstrate how digital enterprise technologies combined with sophisticated optimization algorithms can contribute to planning and scheduling the operation of complex production systems in an efficient and consistent way. In an industry-initiated and involved project, novel models and algorithms were developed for integrated production planning and scheduling. As the results of computational experiments with two systems have shown, our original objectives could be attained. The main contributions are as follows: • We have developed a novel project-based model for production planning in make-to-order and engineer-to-order industries. Planning with variable-intensity activities and feeding precedence constraints enables to combine decisions about the timing of production activities and the utilization of resources. The model that can be amended with various objective functions results in more accurate production plans. • We put special emphasis on developing efficient solution algorithms. On the level of planning, a new custom-tailored branch-and-cut algorithm solves the planning problem with the automatic generation of cutting planes that cut off the fractional solutions at the nodes of the search tree. The constraint-based scheduler was augmented with two methods that reveal – and exploit – typical but hidden structural properties of our industrial problem instances. Thanks to the efficient solvers, large-scale, real-life problem instances can also be solved, and decision makers are able to explore and evaluate a number of alternative future scenarios, too. • In case of hierarchical planning and scheduling, the integrity of solutions can be ensured by an appropriate aggregation method that builds up the high-level planning model from detailed, de facto standard master data. The integration of planning with detailed scheduling and execution leads to a better due date observance and to more efficient use of resources. As a result, planners are capable of accepting more customer orders and reducing production costs. •Our component-based simulation technology proved to be appropriate for assessing the sensitivity of deterministic solutions in face of typical uncertainties. Hence, results of the integrated production planning and scheduling system can be tested and further improved by a discrete-event simulation module which is able to analyze the effects of various types of changes and uncertainties related to operator skills and availability, machine breakdowns, and processing time variations. We emphasize that all the three models are built up automatically, from a common data store. Finally, two possible connections of the integrated production planning and control approach introduced in the paper are to be outlined here. One is the execution control level [23] and the other is planning on the level of production networks. However, in any case, local powerful PPC methods that produce cost-efficient, executable and robust plans and schedules both on the medium and the short-term are prerequisites of these extensions.