Düsseldorf, July 28, 2021

How to increase profitability through predictive production planning

A 5-step guide to getting the most out of advanced analytics

Modern production planning needs to incorporate holistically all aspects of manufacturing execution. In principle, a manufacturing plant transforms raw materials into semi-finished or finished products by utilizing the manpower and equipment and implementing internal and external logistic processes. Different product types are often produced in campaigns, and the profitability of a new order might depend on how well the products of that order fit into the given production scenario. As campaign planning might consider time horizons in the order of weeks, it is of little use to plan for the moment only, but one has to foresee future demand, too. Modern, connected production planning therefore takes all these aspects into account and automatically adapts to the current production state.

The overall goal of manufacturing execution scheduling is to optimize the profitability of production. When it comes to optimizing the total profit, the time horizon of the optimization is critical. With very short time scales, the advantages can be optimized at the cost of a deterioration in the health of the asset or reduced profitability in later stages. On the other hand, planning horizons that are too long naturally reduce the reliability of the schedules and might include suboptimal short term decisions, as future orders are not taken into consideration. This is due to limited knowledge of market developments and production-related incidents. However, since potential benefits on longer time scales do not necessarily occur, this worsens the overall economic efficiency.

The production state is a complex description comprising:

  • Order backlog and forecast
  • Raw material stocks and scheduled supply
  • Asset health state and forecast
  • Work-in-progress
  • Available manpower for all tasks
  • Available spare parts & auxiliary materials
  • Internal yard states
  • External transport situation
  • Cost calculators and estimators (e.g., future raw material costs)

In production planning, numerous effects of individual decisions and measures must be taken into account and precisely evaluated in order to identify the best course of action. For instance, the average product quality can be improved by investing more effort in maintenance and replacing parts on shorter time scales. On the other hand, a higher number of spare parts obviously cost more money, and more frequent maintenance downtimes reduce equipment availability. This example shows that the optimization of individual goals such as product quality or maintenance costs on their own do not necessarily lead to improved overall profitability. A sophisticated production planning system thus incorporates all these effects for optimum performance.

1. Setting up the raw material supply

Depending on the order backlog and the forecast for potential future orders, a first draft of the long-term production order is generated. Based on this order, it is checked to see if sufficient raw material will be available by the time it is needed for scheduled production, including already scheduled supplies. If not, a check is carried out to see if other material required can be purchased and supplied on time. If this is the case, purchase orders can be generated automatically. The question as to whether they are placed automatically or manually only is just a technical detail, though of major practical importance, of course. In order to automatize this process, a standardized specification for raw material properties and purchase conditions is essential. This standard is governed by a purchasing platform. If the required raw materials cannot be supplied on time, the production schedule needs to be adapted accordingly. A valid production schedule and an appropriate raw material supply plan are generated by iterative means.

2. Determining the personnel planning

Once a production schedule is defined, the necessary manpower can be determined. This includes the operating personnel, maintenance crews, and possibly external contractors if work is outsourced. If there are different types of maintenance tasks, a specific mix of experts including mechanical, hydraulic, electrical etc., may be required for a certain period of time. A proper personnel operating schedule can thus be derived from this.

3. Scheduling the asset health

Part of the production scheduling process is considering the asset health as a property that varies over time. Every production item produced reduces the remaining lifetime or the estimated time before failure for any kind of equipment. Yet it is not just failures that can result from part wear, it can also cause a deterioration in the product quality. Examples of this could include work roll wear during rolling, which may lead to poor flatness, profile or surface quality. As such, the achievable quality must also be considered for any future state within the production schedule. If a production schedule requires a higher achievable quality, then the appropriate maintenance must be performed according to the asset condition forecast. In such cases, the availability of requisite skilled personnel is checked and an appropriate maintenance task is scheduled accordingly. At the same time, spare parts that may potentially be required are checked for availability and, if they are not available, ordered in a similar way to the raw materials. This guarantees that, at least according to the current state of information, the production schedule is capable of producing products with the required quality at their scheduled production time in the process.

4. Defining intelligent sequencing

The production scheduling itself must fulfill several tasks. First, the customer order requirements need to be converted into a set of technical production orders. The number of product items, the process route, and the main technological process characteristics need to be defined. Then, the scheduling system needs to split the technical order backlog into subgroups to be produced in a campaign or sequence, as this is required in many process steps like continuous casting or annealing. However, not all products can be successively produced without losses. With casting, changes to the chemical concentrations can only be made based on different heats, and any changes typically have to be small. In the case of annealing, similar temperatures and geometry are common requirements for keeping transition losses low. A key goal of the sequence generator, therefore, is to group the backlog in such way that the transition costs within each sequence are kept low. It is clear that logistical and contractual information must also be taken into consideration, in addition to the technical limitations. In this sense, contractual due dates must be respected and so must the available yard space between the process steps. Here again, the time horizon for optimization plays a crucial role. It can be easy to define just one perfect sequence at the cost of leaving rather demanding products for later or having a rather fragmented backlog that could be hard to produce later on. On the other hand, it might not make sense to schedule for the entire backlog, as additional orders would be active at the time when the last products are produced, so there may be other options for improved sequences. Profitability would be hit if in the short-term sequences had to include products that could be produced in a sequence of far more similar products at a later stage. Here, a forecast of potential future orders is extremely important.

The Demand Planner app shows entire product segment slots and open weights that can be used as a guideline for sales.

Once the composition of a sequence has been determined, the order of the individual products within the campaign can be determined as a further objective. The nature of the process and the characteristics of the production line largely determine the transition costs from one product piece to the next. In simple cases, the sequence can be determined manually. In more complex scenarios, optimization algorithms must be used instead, as it is almost impossible for humans to create an optimal plan.

5. Holistic interlinking of manufacturing and sales

A thorough and system-based analysis of the current order backlog further on allows the manufacturing and sales to be linked more tightly. Before a quotation is sent to a potential customer, the system checks how well the requested product fits into the current order backlog. If there are currently no other orders compatible with the requested product, the quotation should consider the cost of additional downtimes and material losses due to the transition losses or setup times that would then be required. However, the system does not only work reactively. It also proactively makes suggestions to the sales team regarding the material type and quantity to be sold that are a good fit with the current order backlog in a particular way. This is achieved by holistically analyzing the value added on a single product level, taking losses and reduced utilization due to product changeovers into account. The plant itself thus actively optimizes its profitability proactively by supporting the sales process.