After more than a decade of development and refinement, APS has gradually matured from both a technological and market perspective. This article analyzes some characteristics of APS intelligent scheduling systems, a key software in intelligent manufacturing , from an academic perspective, and serves as a reference for enterprise users in their selection process.
I. Basic Attributes of APS
APS (Advanced Planning and Scheduling) is a relatively new type of industrial software. In terms of its application purpose and objectives, APS differs significantly from other software. For example, financial software, inventory management software, and customer resource management software primarily leverage the high processing speed of computers and the ease of data storage, transmission, analysis, error correction, and exchange to "automate" many human tasks. These software programs lack decision-making capabilities; they only do what we instruct them to do, without directing human actions.
The most important attribute of APS (Advanced Planning and Scheduling) is its decision-making function. Its "smartness and competence" are judged by the quality of its decisions, which directly impact a company's productivity and efficiency. Scheduling, or ordering, involves informing APS of the company's resource status and a set of tasks to be addressed. APS then answers questions such as: What to do? Who (equipment, personnel)? What to do? How much to do? When to do it? And so on.
APS initially appeared in some process industries and dedicated production line environments. In fact, in these applications, APS is no smarter than a human, but it is fast, can automate processes, and does not make mistakes.
The complexity of production planning and scheduling is primarily determined by the availability of resources, the Bill of Materials (BOM), and the process. For example, if a factory divides its resources into several production lines, each dedicated to producing a different product, scheduling becomes quite simple. However, if hundreds of products are produced on mixed lines, scheduling becomes much more complex. Adding dynamic resources (such as constraints and variables related to molds, tooling, and personnel), the complexity of the material structure, and specific time constraints, even setting aside optimization issues, simply generating an executable scheduling instruction becomes exceptionally complex (how difficult it is for production schedulers! This also proves that computers are too meticulous, adhering strictly to rules; unlike humans, who are flexible enough to make mistakes and adjust on-site as needed).
In discrete manufacturing, the environment is characterized by multiple objectives, multiple constraints, and dynamic stochasticity, making enterprise production planning and scheduling an extremely complex large-scale system problem. When applied to actual production, the difficulties of scheduling go far beyond simply handling a part blank through turning, milling, planing, and grinding processes. Time, space, temperature, material form, intersections, and dispersion all present numerous constraints. The specific requirements of each process can be listed in the thousands. If relatively important constraints are not considered during scheduling, the result will be unenforceable. It is impossible to use a single mathematical model to address all problems. This is why it is virtually impossible for APS to achieve the same universality as ERP. The fact that ERP uses MRP/MRPII to address all industrial enterprise planning problems is precisely the main reason for its failures in some project implementations.
For APS (Advanced Planning and Scheduling), one of its key performance characteristics is its powerful and flexible modeling capabilities to handle the diverse production constraints faced by enterprises. This is crucial for determining whether a schedule is executable and suits the company's actual situation. Secondly, its optimization capability is paramount—its ability to find the optimal solution from an infinite number of feasible options. Generally, finding an executable schedule isn't particularly difficult, perhaps taking less than a second. However, optimizing the schedule is extremely challenging.
Optimizing scheduling results presents two main challenges. First, businesses have multiple, often conflicting, objectives. For example, maximizing order fulfillment while simultaneously minimizing production cycles is contradictory. Optimization requires human decision-making. Once the objectives are defined, the next step is finding the best job order sequence. Mathematically, sequencing involves permutations and combinations, represented by N!. If a computer can process 1,000,000 sequences per second, we can only process a maximum of 11 sorted results per minute. Given 20 sorted results, finding the optimal solution would take 77,147 years! In reality, the number of executable solutions for scheduling is far greater than tens of thousands! This necessitates the "optimization algorithms" we often discuss.
Discussions of optimization algorithms are too complex for manufacturing companies to handle, and there's no need to waste time on them. Enterprise users only care about results. One point needs clarification: implementing an optimization algorithm requires a powerful "optimization engine," and the optimization process takes time; APS (Advanced Processing Speed) cannot produce optimization results instantly. This is where cloud computing platforms will play a significant role.
Whether or not optimization calculations are performed directly results in the number of delayed orders, or whether the potential 20% capacity has been tapped. For manufacturing companies, this makes a qualitative difference, and may be precisely the original intention behind implementing APS (Advanced Planning and Scheduling).
II. How to Choose an APS
PK method
APS is an optimized scheduling tool. Ultimately, APS aims to improve enterprise production efficiency, not replace manual scheduling. How to verify this? For APS users, the simplest way to test and measure it is to take historical production data from a past month, specify an optimization index sequence, and have several APS software vendors schedule it. A target sequence can be proposed, such as: first, minimizing order delays; second, maximizing equipment utilization; and third, minimizing order production cycles. Enterprises only need to provide APS vendors with consistent basic data and necessary constraints. Some data (such as standard working hours) may not be perfectly accurate; approximation is sufficient. If necessary, APS vendors can conduct on-site investigations at the production site. Because the planning and scheduling experts at the production enterprise are highly experienced, they can immediately determine whether an APS is suitable for their application by looking at the scheduling results, and by comparing the results of different vendors, they can see the differences. Let several APS vendors compete, and the customer can then analyze and verify the results.
Enterprise users can also set demanding goals based on their actual needs. For example, they might provide enough orders to push the utilization rate of relatively bottleneck equipment close to 100%; or the production cycle for certain orders must not exceed a certain number of days. Simply providing simplistic data and requirements makes the competition meaningless. Even if these goals cannot be achieved, the APS vendor must provide convincing reasons.
Whether the scheduling results meet the company's needs is the most important criterion for selecting APS (Advanced Planning and Scheduling). Improvements in enterprise productivity largely depend on this, and it also reflects the technological sophistication of APS. Other requirements on the visual interface (reports, Gantt charts, etc.) are secondary.
Customer adaptability
Production planning and scheduling is not about achieving the highest possible precision. Theoretically, Advanced Planning and Scheduling (APS) can schedule tasks down to the second, down to every action of every person and every piece of equipment. This is only meaningful for automated production lines.
The management levels of Chinese enterprises vary widely. Frankly speaking, the management level of many industrial enterprises may not even reach that of Taylor's era a hundred years ago. For some industrial enterprises that primarily employ migrant workers, simply keeping track of orders and achieving synchronization of information and material flow at the team and work center levels is already quite good. Therefore, choosing an APS (Advanced Management System) is not necessarily better the more advanced it is; enterprises can choose a more economical and practical APS.
The complexity of production site management varies greatly among enterprises. Some enterprises have a limited product range, operate on dedicated production lines, have simple product structures, limited equipment resources, and simple processes. In such cases, the results of manual scheduling and APS scheduling will not differ significantly. Using APS can greatly reduce the workload of manual labor and prevent some basic errors. These enterprises do not need to choose an overly expensive APS system.
Some companies have obvious bottleneck equipment (primary resources). Their needs can be met using either APS or MRP within ERP systems.
One point needs to be emphasized. Lean management is a gradual process. Some companies' current management level can only adapt to relatively crude management, and using a simple APS (Advanced Management System) is acceptable. However, as the company's management level improves, it is essential to consider whether the APS can adapt to the new needs and environment.
An excellent APS (Advanced Management System) should embody advanced management concepts and implement these concepts throughout the system's operation. Operating under an APS system, enterprises can continuously identify and solve problems, thereby improving their management level. These requirements should be met by the manufacturing enterprise itself, rather than requiring constant assistance from the software vendor.
Is it about automating manual processes? Or utilizing artificial intelligence?
In large industrial enterprises, production planning and scheduling experts are invaluable assets. Through years of production practice, they have accumulated a wealth of experience. This experience cannot be replaced by any software. In other words, if a software system cannot absorb the enterprise's expert experience (or, in more technical terms, quantify that expert experience), then the APS (Advanced Planning and Scheduling) system may not achieve better results than those of the experts themselves.
Influenced by ERP solution models, some domestic and international APS vendors design standard templates (standardized modeling) based on common problems in industrial enterprises, and then perform some peripheral development on customized output forms according to customer needs. This modeling approach requires simplification or approximation of production site requirements and constraints, and it is difficult to incorporate expert experience. Therefore, one should not expect APS to have perfectly mature templates that can be readily applied. Compared to ERP, APS requires significantly more secondary development and more detailed enterprise research. In particular, the quantification of expert experience is reflected in the software system. Without these conditions, scheduling optimization in APS is impossible.
The management models that companies have developed over many years, including some of their details, all have their own rationale. APS (Advanced Planning and Scheduling) must adapt to the company and continuously improve upon its existing foundation; rather than the company adapting to APS by making certain changes from the outset. For example, APS should perfectly adapt to the company's habits regarding various production reports. This is very different from the implementation of ERP.
III. Regarding APS Presentation Format
Reverse and forward arrangement:
Some APS (Advanced Planning and Scheduling) systems on the market offer the option of reverse scheduling or forward scheduling. First, this shouldn't be a choice for the user. The software can schedule however it wants; users only care about the results. Second, both theory and practice have proven that just-in-time (JIT) scheduling requires sufficient equipment resources. It's only viable when operating in a continuous flow or with very high capacity. The goal of optimizing scheduling is to maximize capacity and reduce costs. Whether to use forward scheduling, reverse scheduling, or a hybrid approach is determined by the overall solution, not by the customer.
Artificial intervention:
When the scheduling results (Gantt chart) are unsatisfactory, the Gantt chart can be dragged locally to attempt to improve the scheduling results and reflect human intervention. Ten years ago, this method was widely used in Europe and America. This was because the achievements of operations research at that time could not support the planning and scheduling calculations of complex production scenarios. Now, this method has long been abandoned in Europe and America. The ability to support global optimization is a hallmark of APS upgrades.
1. Drag and drop in a Gantt chart to immediately reschedule. This relies on the premise that the scheduling operation must be very fast; otherwise, the tedious dragging and dropping would become tedious. As mentioned earlier, finding a scheduling result among thousands of executable solutions is easy. For example, some algorithms only generate one result and then stop. The system does not automatically search for the optimal result. Therefore, manual dragging and dropping is required to find a satisfactory result.
2. In reality, manually dragging and dropping Gantt charts usually results in localized optimization, and may not necessarily benefit overall metrics. Only when production capacity is extremely abundant can a comprehensive improvement be achieved simultaneously.
3. The use of intelligent software tools means that human intervention should be reflected in "strategic deployment," not "commanding individual soldiers." Seeing a team (equipment) idle or an order delayed, you can simply drag and drop it to initiate operations and capture that "hill." However, if follow-up troops don't arrive, it's meaningless, or you might capture one hill only to lose another.
4. From the perspective of software vendors expanding their market, dragging and dropping in a Gantt chart instantly produces a new scheduling result, which certainly looks good. This is very encouraging for those new to APS (Advanced Planning and Scheduling). However, APS is an optimization scheduling tool, not a game console. From a technical implementation standpoint, dragging and dropping in a Gantt chart is not difficult. The key issue is that when scheduling considers global optimization, this approach is meaningless. Furthermore, human vision has significant limitations. When hundreds of orders are presented in a Gantt chart, if capacity is already tight, dragging and dropping will inevitably result in neglecting some aspects.
Human intervention should be reflected in the overall layout and objectives. For example, establishing a sequence of scheduling optimization targets for this phase; prioritizing certain orders; allocating resources to a critical customer with high quality requirements; arranging regular maintenance plans; reducing and adjusting non-bottleneck equipment and manpower resources and allocating them to bottleneck processes; improving the supply chain, etc. Tell your needs and strategy to APS, and let it make the specific arrangements to meet your requirements—that's the kind of human intervention needed.
IV. APS is merely a software tool.
Enterprise production management is a systems engineering problem; the planning and scheduling problems seen on the production floor are merely superficial. Solving enterprise production management challenges requires analysis and verification from a holistic system perspective. Do not assume that implementing an APS (Advanced Planning and Scheduling) system will solve an enterprise's production planning and scheduling problems. APS is just a tool, and it is only effective when a correct overall solution is in place.
Many factors contribute to confusion in enterprise production planning and scheduling, with varying degrees of impact. These include the enterprise's management system and mechanisms, sales and marketing management models, supply chain issues (especially supplier management), inventory management, layout and environmental factors, equipment management systems, process technology management, human resource management mechanisms, and workshop production management models and processes. Improving production site management for an enterprise begins with a comprehensive and scientific diagnosis of its production management, identifying the core problems and the relationships between them. Only then can an APS (Advanced Planning and Scheduling) solution be designed based on the enterprise's specific circumstances. This is where the true value of APS lies. Unlike ERP and other software, APS must adapt to the enterprise's current situation, not the other way around. Influenced by traditional software tool implementation models, enterprises often overlook this crucial issue. Enterprise diagnosis and consulting are another closely related scientific field.
V. Other technical specifications: System integration (interfaces)
An enterprise's IT architecture should be centered on its core business, meaning it should start with the IT infrastructure for production planning and scheduling, and then expand outwards from there. Unfortunately, many enterprises have ERP, MES, PDM, CAPP, and other software to varying degrees, leaving a gap in the core functionality. Regardless of the current situation, APS (Advanced Planning and Scheduling) must carefully consider its interface with other software.
1. Open Standards
2. Module Design
3. Well-designed Application Programming Interface (API)
4. Distributed Protocols
5. Integration Mechanism
