TL;DR
Reliability in Flexible Manufacturing Systems (FMS) is often calculated by looking at individual machines in isolation, but this ignores the "ripple effect" where one failure cascades through the system. This paper introduces a hierarchical evaluation system that uses Petri nets and judgment matrices to quantify how hardware, software, and human errors interact, allowing engineers to pinpoint the true bottlenecks in system-level availability.
What Problem Does This Paper Address?
In modern manufacturing, an FMS is a complex "system of systems" involving mechanical equipment, automated transfer vehicles (UGVs), hierarchical control software, and human operators. Traditionally, reliability engineers focused on the failure rates of single components (e.g., "How often does this specific robot break?").
However, the authors argue that this approach misses coupling relations. For example, a sensor glitch in a machine tool might send a "wrong command" to the upper control system, which then halts the entire production line. Conversely, a well-designed buffer system might allow the rest of the line to keep moving even if one machine fails. The challenge is creating a mathematical framework that accounts for these interdependencies to accurately predict overall system availability during the design phase.
Methods
The authors propose a three-level hierarchical index system:
- Level 1 (System): Overall FMS availability.
- Level 2 (Subsystems): Human operators, upper control systems, processing systems, and transfer systems.
- Level 3 (Elements): Individual components like robots, machine tools, and bottom-level controllers.
To make this system work, they combine two key engineering tools:
- Petri Nets (PN): These are used to model the discrete flow of parts and information. The authors developed specific PN models for parallel processes, synchronous operations, and equipment with buffer storage. By simulating these nets, they can see how "tokens" (representing workpieces) get stuck when specific elements fail.
- Analytic Hierarchy Process (AHP): The authors use judgment matrices to assign "weights" to different components. This helps determine which elements have a "major," "general," or "minor" influence on the total system.
- Coupling Error Probability ($CEP$): A novel metric introduced to represent the probability that a failure in element $a$ causes a failure in element $b$.
Datasets
The paper does not use a standard public dataset. Instead, it validates the model using a simulated application example of a box-type component machining FMS.
- Parameters: The authors defined specific Mean Time Between Failures (MTBF) and Mean Time to Repair (MTTR) for 10 distinct elements (e.g., Loading Robot MTBF = 850h, PC-based control system MTBF = 30,000h).
- Software: Simulations were conducted using SPNP6.0, a specialized tool for analyzing stochastic Petri nets.
Key Insights and Ideas
- The Human/Software Factor: Unlike many mechanical reliability models, this framework treats the human operator (Human Error Rate) and the PC-based control system as integral "elements" with their own availability scores.
- Buffers as Reliability "Shock Absorbers": The model explicitly accounts for the fact that an FMS can often continue running during a short equipment failure if the buffer capacity is sufficient. This prevents the system availability from being a simple product of its parts.
- Weighting the "Weakest Link": The study found that not all high-failure components are equally damaging. A component's "influence level" depends on its position in the hierarchy and its coupling with other systems.
Key Results
Using their machining line example, the authors calculated:
- Overall System Availability: 0.8516.
- Critical Bottlenecks: In this specific setup, Machine Tools (weighted at 0.4) and UGVs (weighted at 0.3264) were identified as the most significant influences on system reliability.
- Actionable Intelligence: By identifying these weights, the system shows engineers exactly where to invest in higher-quality hardware or redundant systems to get the biggest boost in total uptime.
Limitations and Future Work
- Data Dependency: The model’s accuracy relies heavily on having high-quality field data for MTBF and MTTR, as well as accurate estimates for "coupling error probabilities," which can be difficult to determine for entirely new system designs.
- Complexity: As manufacturing systems grow larger (e.g., "Lights-out" factories), the Petri net models may become computationally expensive to simulate.
- Future Directions: The authors suggest that the next step is integrating this evaluation system directly into the design and layout optimization phase of FMS development to "build in" reliability from day one.
Summary updated