In recent years, the rapid development of the national economy has significantly increased the demand for coal energy. Coal remains a major primary energy source, and the construction of high-yield, high-efficiency mines has placed higher requirements on the reliability of core equipment, especially shearers. Among the many components of a shearer, the drive gear is one of the most critical and vulnerable parts. It transmits large torque and traction force, meshes with the pin rack of the scraper conveyor, and resists strong cutting reaction forces. Because it is difficult to replace underground, the drive gear is often supplied as a component assembly. However, the manufacturing process of the drive gear is complex, with long production cycles, low machining efficiency, and frequent delays in delivery. These problems have long remained unsolved and have caused production stoppages and economic losses in coal mines. Therefore, I focused on the drive gear machining process and attempted to use a simulation-based method to analyze and optimize the production system. The关键词 of this study is gear machining, and I repeatedly examined how gear machining operations interact with limited resources. By building a discrete-event simulation model in Witness, I aimed to identify bottlenecks in gear machining, improve equipment utilization, and increase monthly output.
The traditional approach to solving the production bottleneck relied on experience, simple calculations, and trial-and-error adjustments. For example, production planners compared total processing hours with available machine hours and concluded that wire cutting and hobbing were the main constraints. However, such static calculations ignore random events, queueing effects, machine failures, operator availability, and the dynamic interaction between processes. A more systematic method is needed. System simulation provides a powerful alternative because it can represent the stochastic and dynamic behavior of a manufacturing system. In this study, I treated the drive gear gear machining system as a typical discrete-event dynamic system. I selected Witness, a discrete-event simulation software, to model the entire gear machining process, from raw material arrival to finished gear shipment. The model allowed me to test different configurations of machine quantities and processing times, observe the resulting utilization rates and throughput, and propose an optimized production plan.
The main objective of this research was to build a simulation model that accurately reflects the actual gear machining process for shearer drive gears, to analyze the system performance, and to optimize the production process. The specific goals included: (1) establishing a Witness-based simulation model of the drive gear gear machining line; (2) identifying bottlenecks by analyzing equipment utilization, buffer contents, and part waiting times; (3) evaluating optimization scenarios that change the number of machines, processing times, or both; and (4) providing decision support for production management, process improvement, and system analysis. I conducted field investigations, collected processing data, defined model elements, verified the model, and ran multiple independent replications. The results showed that the original system had a monthly processing capacity of about 43 pieces, while the optimized system could reach 55–60 pieces per month, increasing gear machining efficiency by approximately 35%.
System simulation is a scientific technique that constructs a model of a real-world system and experiments with it on a computer. It is especially useful when analytical solutions are difficult or impossible to obtain. The essence of simulation is to explore the behavior of a system under different parameters and constraints, thereby understanding or improving the real system. A system consists of entities, attributes, and activities. Entities are the objects of interest, attributes are their characteristics, and activities are the changes that occur over time. Simulation can be classified into continuous, discrete, and combined systems. The drive gear gear machining system is a discrete-event system because its state changes at discrete points in time when events occur, such as a machine finishing a part or a part entering a buffer.
Discrete-event simulation has several advantages. It can handle complex systems with random factors, it can model both linear and nonlinear relationships, and it can provide insights that analytical methods cannot. The main simulation strategies for discrete-event systems are event scheduling, activity scanning, and process interaction. In this study, I used the process-interaction approach, which is the natural modeling paradigm in Witness. The simulation model was built with elements such as machines, buffers, parts, and labor. Machines process parts, buffers store parts, parts flow through the system, and labor resources assist with setups and operations. The model was run for a simulated period of one month (30 days, 8 hours per day, totaling 14,400 minutes). Because the simulation is stochastic, I used multiple independent replications with different random number streams to obtain reliable statistics.
Witness is a powerful simulation software developed for discrete-event systems. It provides an interactive modeling environment, flexible execution strategies, and a rich set of elements. The main elements include machines (single, batch, assembly, production), buffers, parts, labor, and variables. Witness also offers modules for definition, display, detail, report, run, and experiment. The software allows users to define input and output rules, cycle times, setup times, breakdown patterns, and shift schedules. It can produce animated displays and detailed statistical reports. I used Witness to model the drive gear gear machining process because it is well suited for manufacturing systems and has been widely applied in automotive, aerospace, and logistics industries.
The drive gear studied here is an involute-tooth drive gear used in a shearer. Its gear machining route includes forging, normalizing and high-temperature tempering, rough turning, aging, semi-finish turning, wire cutting for rough tooth profile, hobbing for semi-finish tooth profile, deburring, carburizing, semi-finish turning, overall quenching, finish turning, grinding, wire cutting for internal spline, tooth grinding, and deburring. The process involves several machine groups: turning, wire cutting, hobbing, grinding, tooth grinding, bench work, and heat treatment. The processing times and machine quantities are summarized in Table 1. These data were collected from the actual production floor and used as inputs for the simulation model.
| Operation | Description | Machine | Time (min) | Available Machines |
|---|---|---|---|---|
| 05 | Turning: rough turning surface | CW6180 | 660 | 20 |
| 10 | Heat treatment: aging stress relief | Box furnace | 720 | 9 |
| 15 | Turning: semi-finish turning | CY-K6180 | 330 | 20 |
| 20 | Wire cutting: rough involute profile | CTWQ800Z | 4525 | 14 |
| 25 | Hobbing: semi-finish profile | YA31200K | 1860 | 2 |
| 30 | Bench work: tooth profile chamfering | Bench | 60 | 2 |
| 35 | Heat treatment: carburizing | Pit furnace | 6600 | 5 |
| 40 | Turning: semi-finish outer surface | CY-K6180 | 270 | 20 |
| 45 | Heat treatment: overall quenching | Multi-purpose furnace | 960 | 2 |
| 50 | Turning: finish turning with grinding allowance | CY-K6180 | 120 | 20 |
| 55 | Grinding: finish outer surface | MKW28125 | 300 | 3 |
| 60 | Turning: finish inner surface | CY-KX6180 | 120 | 20 |
| 65 | Wire cutting: internal spline | CTWQ800Z | 1650 | 14 |
| 70 | Tooth grinding: finish profile | P1200G | 555 | 2 |
| 75 | Bench work: deburring | Bench | 60 | 2 |
From Table 1, I aggregated the processing times by machine group. The total time for the turning group is 1,500 minutes, wire cutting is 6,175 minutes, hobbing is 1,860 minutes, grinding is 270 minutes, tooth grinding is 555 minutes, bench work is 120 minutes, and heat treatment is 8,280 minutes. These aggregated times are shown in Table 2. The total processing time per part is 11,200 minutes, which is equivalent to about 186.7 hours. This long total time indicates that the gear machining process is resource-intensive and has significant potential for bottlenecks.
| Machine Group | Turning | Wire Cutting | Hobbing | Grinding | Tooth Grinding | Bench Work | Heat Treatment |
|---|---|---|---|---|---|---|---|
| Time (min) | 1500 | 6175 | 1860 | 270 | 555 | 120 | 8280 |
Before building the simulation model, I defined the modeling objectives. The main objective was to simulate the drive gear gear machining system under normal working conditions and to calculate performance measures such as average waiting time, average queue length, average utilization, and throughput. I also aimed to identify bottlenecks and evaluate optimization scenarios. The scope of the model included the main machine groups, buffers, and the part flow. External factors such as raw material supply and demand fluctuations were considered as boundary conditions. I assumed that raw materials are always available, that machines do not fail unless specified, and that operators are available when needed. These assumptions simplified the model while preserving the essential behavior of the gear machining system.
In Witness, I defined four types of elements: machines, buffers, parts, and labor. The machine elements represented the processing equipment. The buffer elements represented the storage areas between processes. The part element represented the drive gear being machined. The labor element represented the operators who perform setups and bench work. The model included 17 machine elements, 5 buffer elements, 1 part element, and 1 labor element. The machine groups and their quantities are listed in Table 3. The buffers were placed before each machine group to hold parts waiting for processing. The part followed a fixed route: heat treatment, turning, wire cutting, hobbing, grinding, tooth grinding, and bench work. After the final operation, the part was shipped out of the system. The input and output rules were defined to pull parts from upstream buffers and push them to downstream buffers. For example, the turning group pulled parts from buffer b1 and pushed them to buffer b2 when processing was complete.
| Element Name | Type | Quantity | Description |
|---|---|---|---|
| Duanjian | Part | 1 | Raw drive gear |
| Rechuli | Machine | 1 | Heat treatment group |
| Chechuangzu | Machine | 4 | Turning group |
| Xianqiege | Machine | 4 | Wire cutting group |
| Gunchizu | Machine | 2 | Hobbing group |
| Mochuangzu | Machine | 3 | Grinding group |
| Mochizu | Machine | 2 | Tooth grinding group |
| Qiangongzu | Machine | 1 | Bench work group |
| b1 | Buffer | 1 | Input buffer for turning |
| b2 | Buffer | 1 | Input buffer for wire cutting |
| b3 | Buffer | 1 | Input buffer for hobbing |
| b4 | Buffer | 1 | Input buffer for grinding |
| b5 | Buffer | 1 | Input buffer for bench work |
The cycle times for the machine groups were set according to the aggregated processing times. For example, the turning group had a cycle time of 12 minutes per part when considering four parallel machines, the wire cutting group had a cycle time of 25 minutes, the hobbing group had 31 minutes, the grinding group had 4.5 minutes, the tooth grinding group had 9.25 minutes, the bench work group had 2 minutes, and the heat treatment group had 12 hours (720 minutes) per batch. The input and output rules were defined using Witness logic. For instance, the turning group pulled from b1 only if the number of parts in b2 was less than or equal to 4, otherwise it waited. This rule prevented excessive blocking. The wire cutting group pulled from b2 only if the number of parts in b3 was less than or equal to 2. These rules were designed to mimic the actual pull-based production control observed in the factory.
After building the model, I verified it by checking a specific event. The first part should complete the turning operation at simulation time 37 hours (2,220 minutes). I set the simulation clock to 37 hours and ran the model. The results showed that the turning group completed its first part exactly at that time, and the part was pushed into buffer b2. The variable V1, which counted finished parts, became 1. The buffer statistics showed that b2 had one part, and b3 had zero parts. The machine statistics showed that the turning machines had an average utilization of about 50%, the wire cutting machines about 70%, the hobbing machines about 25%, and the other machines were mostly idle. These results were consistent with the expected behavior, confirming that the model logic was correct. Table 4 presents the verification results for buffers and machines.
| Name | Total In | Total Out | Now In | Max |
|---|---|---|---|---|
| b2 | 2 | 1 | 1 | 1 |
| b3 | 3 | 0 | 0 | 0 |
| Machine Group | % Idle | % Busy | No. of Operations |
|---|---|---|---|
| Heat treatment | 4.98 | 95.02 | 15 |
| Turning (1) | 50.25 | 49.75 | 4 |
| Turning (2) | 51.24 | 48.76 | 3 |
| Turning (3) | 57.21 | 42.79 | 3 |
| Turning (4) | 62.69 | 37.31 | 3 |
| Wire cutting (1) | 23.38 | 76.62 | 1 |
| Wire cutting (2) | 29.35 | 70.65 | 1 |
| Wire cutting (3) | 35.32 | 64.68 | 1 |
| Wire cutting (4) | 41.29 | 58.71 | 1 |
| Hobbing (1) | 74.13 | 25.87 | 1 |
| Hobbing (2) | 80.10 | 19.90 | 1 |
| Grinding (1) | 95.52 | 4.48 | 2 |
| Grinding (2) | 100.00 | 0.00 | 0 |
| Grinding (3) | 100.00 | 0.00 | 0 |
| Tooth grinding (1) | 95.40 | 4.60 | 1 |
| Tooth grinding (2) | 97.76 | 2.24 | 0 |
| Bench work | 99.00 | 1.00 | 1 |
After verification, I ran the model for a full month of production. The simulation time was set to 14,400 minutes (30 days × 8 hours × 60 minutes). Because the system is stochastic, I used the method of independent replications. I ran the model three times with different random number streams. The results for the part element are shown in Table 5. The average work-in-process (WIP) was 20.33 parts, and the average time in system was 157.52 hours. The standard deviation of the average time was 0.179 hours, and the variance was 0.032. These statistics indicate that the part spends a significant amount of time waiting in buffers, especially before the wire cutting operation.
| Run | No. Entered | No. Shipped | No. Rejected | W.I.P. | Avg W.I.P. | Avg Time (h) | Sigma |
|---|---|---|---|---|---|---|---|
| 1 | 31 | 0 | 210 | 31 | 20.35 | 157.58 | 0.00 |
| 2 | 31 | 0 | 210 | 31 | 20.30 | 157.45 | 0.00 |
| 3 | 31 | 0 | 210 | 31 | 20.33 | 157.52 | 0.00 |
| Statistic | No. Entered | No. Shipped | W.I.P. | Avg W.I.P. | Avg Time (h) |
|---|---|---|---|---|---|
| Mean | 31 | 210 | 31 | 20.33 | 157.52 |
| Std. Dev. | 0 | 0 | 0 | 0.046 | 0.179 |
| Variance | 0 | 0 | 0 | 0.002 | 0.032 |
The output analysis revealed several bottlenecks. The highest machine utilization was in the wire cutting group, reaching 99.06%, followed by the heat treatment group at 98.54%. The hobbing group had a utilization of about 71.26%, while the grinding and tooth grinding groups had very low utilization, below 10%. The turning group had a blockage rate of about 86.3%, which was caused by the large number of turning machines, short processing times, and the pull-based input rule. Buffer b2 and b3 contained many parts. On average, b2 held 12 parts, and the waiting time in b2 was about 102 hours. Almost all of the part’s total time in system was spent waiting in b2. The main reason was that the wire cutting group required setup and adjustment time, and the wait for an operator to perform the setup accounted for 24.84% of the total time. These findings clearly identified wire cutting and hobbing as the primary bottlenecks in the gear machining system.
The first optimization scenario focused on changing the number of machines. I increased the turning group from 4 to 6 machines and the wire cutting group from 4 to 6 machines. The hobbing group remained at 2 machines. I ran the model seven times for 14,400 minutes each. The results are shown in Table 6. The average utilization of the turning group was 66.16%, wire cutting 66.33%, hobbing 96.86%, grinding 76.72%, and tooth grinding 9.86%. The hobbing group was still highly loaded, above 90%, and the output increased to 51 pieces per month. This showed that increasing the number of turning and wire cutting machines improved output, but the hobbing group became the new bottleneck.
| Machine Group | 1 | 2 | 3 | 4 | 5 | 6 | 7 | Average |
|---|---|---|---|---|---|---|---|---|
| Turning | 66.21 | 66.09 | 66.22 | 66.14 | 66.13 | 66.18 | 66.14 | 66.16 |
| Wire cutting | 66.35 | 66.26 | 66.43 | 66.29 | 66.32 | 66.40 | 66.23 | 66.33 |
| Hobbing | 96.88 | 96.81 | 96.94 | 96.83 | 96.85 | 96.90 | 96.79 | 96.86 |
| Grinding | 76.71 | 76.71 | 76.74 | 76.75 | 76.70 | 76.65 | 76.76 | 76.72 |
| Tooth grinding | 9.85 | 9.85 | 9.84 | 9.87 | 9.85 | 9.86 | 9.85 | 9.86 |
In the second optimization scenario, I further increased the hobbing group from 2 to 3 machines while keeping the turning and wire cutting groups at 6 machines each. The results are shown in Table 7. The average utilization of the turning group was 74.64%, wire cutting 75.75%, hobbing 55.01%, grinding 55.39%, and tooth grinding 84.70%. The output increased to 58 pieces per month. The bottleneck shifted again, and the tooth grinding group became highly utilized. This indicated that simply changing machine quantities could improve output but also created new bottlenecks.
| Machine Group | 1 | 2 | 3 | 4 | 5 | 6 | 7 | Average |
|---|---|---|---|---|---|---|---|---|
| Turning | 74.61 | 74.65 | 74.62 | 74.69 | 74.66 | 74.68 | 74.59 | 74.64 |
| Wire cutting | 75.76 | 75.72 | 75.77 | 75.74 | 75.78 | 75.77 | 75.69 | 75.75 |
| Hobbing | 55.08 | 54.98 | 55.02 | 54.98 | 55.03 | 55.09 | 54.92 | 55.01 |
| Grinding | 55.37 | 55.38 | 55.46 | 55.38 | 55.35 | 55.42 | 55.34 | 55.39 |
| Tooth grinding | 84.72 | 84.68 | 84.70 | 84.72 | 84.76 | 84.72 | 84.58 | 84.70 |
The third optimization scenario focused on changing processing times. By introducing advanced equipment and improving the gear machining methods, I reduced the processing times as shown in Table 8. The turning time was reduced from 1,500 to 1,200 minutes, wire cutting from 6,175 to 5,600 minutes, hobbing from 1,860 to 1,660 minutes, grinding remained 270 minutes, tooth grinding remained 555 minutes, bench work remained 120 minutes, and heat treatment was reduced from 8,280 to 6,800 minutes. The total time dropped from 11,200 to 10,085 minutes. I then ran the model with these reduced times while keeping machine quantities unchanged. The results are shown in Table 9. The average utilization of the turning group was 98.28%, wire cutting 80.31%, hobbing 63.93%, grinding 8.20%, and tooth grinding 98.28%. The output increased slightly, but the turning group became a serious bottleneck with nearly 100% utilization.
| Machine Group | Turning | Wire Cutting | Hobbing | Grinding | Tooth Grinding | Bench Work | Heat Treatment | Total |
|---|---|---|---|---|---|---|---|---|
| Time (min) | 1200 | 5600 | 1660 | 270 | 555 | 120 | 680 | 10085 |
| Machine Group | 1 | 2 | 3 | 4 | 5 | 6 | 7 | Average |
|---|---|---|---|---|---|---|---|---|
| Turning | 98.28 | 98.24 | 98.31 | 98.25 | 98.27 | 98.29 | 98.32 | 98.28 |
| Wire cutting | 80.14 | 80.54 | 80.54 | 79.98 | 80.00 | 80.27 | 80.67 | 80.31 |
| Hobbing | 63.66 | 64.40 | 64.00 | 64.01 | 63.57 | 63.83 | 64.07 | 63.93 |
| Grinding | 8.18 | 8.21 | 8.19 | 8.21 | 8.23 | 8.19 | 8.20 | 8.20 |
| Tooth grinding | 98.28 | 98.24 | 98.31 | 98.25 | 98.27 | 98.29 | 98.32 | 98.28 |
In the fourth optimization scenario, I further reduced the processing times for all operations except heat treatment. The results are shown in Table 10. The average utilization of the turning group was 98.53%, wire cutting 57.40%, hobbing 64.33%, grinding 8.17%, and tooth grinding 98.53%. The turning and tooth grinding groups were still highly utilized, and the output did not improve significantly. This indicated that changing time alone was insufficient to resolve the bottlenecks.
| Machine Group | 1 | 2 | 3 | 4 | 5 | 6 | 7 | Average |
|---|---|---|---|---|---|---|---|---|
| Turning | 98.50 | 98.47 | 98.53 | 98.48 | 98.56 | 98.61 | 98.59 | 98.53 |
| Wire cutting | 57.82 | 56.93 | 57.18 | 57.23 | 57.90 | 57.51 | 57.20 | 57.40 |
| Hobbing | 64.57 | 64.18 | 63.90 | 64.07 | 64.62 | 64.70 | 64.24 | 64.33 |
| Grinding | 8.17 | 8.22 | 8.14 | 8.16 | 8.18 | 8.16 | 8.13 | 8.17 |
| Tooth grinding | 98.50 | 98.47 | 98.53 | 98.48 | 98.56 | 98.61 | 98.59 | 98.53 |
Finally, I combined both quantity and time changes. I used the reduced processing times and increased the number of machines for the bottleneck groups. The turning group was set to 6 machines, wire cutting to 6 machines, hobbing to 3 machines, grinding to 3 machines, and tooth grinding to 2 machines. The heat treatment time remained unchanged. The results are shown in Table 11. The average utilization of the turning group was 67.54%, wire cutting 66.47%, hobbing 39.37%, grinding 39.35%, and tooth grinding 84.70%. The output reached 62 pieces per month. This was the best result among all scenarios, with balanced utilization and significantly higher throughput.
| Machine Group | 1 | 2 | 3 | 4 | 5 | 6 | 7 | Average |
|---|---|---|---|---|---|---|---|---|
| Turning | 67.57 | 67.61 | 67.46 | 67.53 | 67.54 | 67.50 | 67.58 | 67.54 |
| Wire cutting | 66.48 | 66.54 | 66.44 | 66.41 | 66.52 | 66.43 | 66.47 | 66.47 |
| Hobbing | 39.39 | 39.43 | 39.36 | 39.34 | 39.37 | 39.35 | 39.33 | 39.37 |
| Grinding | 39.35 | 39.33 | 39.35 | 39.37 | 39.31 | 39.34 | 39.38 | 39.35 |
| Tooth grinding | 84.71 | 84.76 | 84.71 | 84.69 | 84.67 | 84.65 | 84.72 | 84.70 |
To compare the optimization scenarios, I calculated the overall equipment utilization and output for each. The utilization formula used was:
$$ U = \frac{T_{busy}}{T_{total}} \times 100\% $$
where \(T_{busy}\) is the total busy time of all machines in a group and \(T_{total}\) is the total available time. The output was measured as the number of shipped parts per month. The comparison is shown in Table 12. The original system had an output of 43 pieces per month. Changing quantity factors increased output to 51 and then 58 pieces. Changing time factors alone increased output only slightly and created new bottlenecks. The combined optimization achieved the highest output of 62 pieces per month and improved the balance of utilization across machine groups.
| Scenario | Turning (%) | Wire Cutting (%) | Hobbing (%) | Grinding (%) | Tooth Grinding (%) | Output (pieces/month) |
|---|---|---|---|---|---|---|
| Original | 50.25 | 99.06 | 71.26 | 4.48 | 4.60 | 43 |
| Quantity 1 | 66.16 | 66.33 | 96.86 | 76.72 | 9.86 | 51 |
| Quantity 2 | 74.64 | 75.75 | 55.01 | 55.39 | 84.70 | 58 |
| Time 1 | 98.28 | 80.31 | 63.93 | 8.20 | 98.28 | 48 |
| Time 2 | 98.53 | 57.40 | 64.33 | 8.17 | 98.53 | 50 |
| Combined | 67.54 | 66.47 | 39.37 | 39.35 | 84.70 | 62 |
The results of this study demonstrate that discrete-event simulation is an effective tool for analyzing and optimizing gear machining systems. The Witness model captured the dynamic behavior of the drive gear production line, including queueing, blocking, and resource constraints. The model was verified and validated against actual production data. The bottleneck analysis showed that wire cutting and hobbing were the primary constraints in the original system. By increasing the number of machines in these groups and reducing processing times through advanced gear machining technology, the monthly output could be increased from 43 to 55–60 pieces, an improvement of about 35%. The optimized configuration also achieved a more balanced utilization of equipment, reducing the risk of overloading any single machine group.
The findings have several practical implications. First, the simulation model can be used as a decision support tool for production planning and capacity expansion. Managers can test different scenarios without disrupting actual production. Second, the model can help identify the most cost-effective improvement measures. For example, adding machines to the bottleneck groups is more effective than simply reducing processing times everywhere. Third, the model can be updated as new equipment is installed or as product demand changes. In the future, I plan to incorporate more detailed factors such as machine failures, operator shifts, and quality inspection into the model. I also intend to use optimization algorithms to search for the best configuration automatically. Overall, this research contributes to the application of simulation in gear machining and provides a foundation for continuous improvement in shearer drive gear manufacturing.
In conclusion, I built a Witness-based simulation model of the shearer drive gear gear machining process. The model included machines, buffers, parts, and labor, and it was verified against actual production data. The output analysis identified wire cutting and hobbing as bottlenecks. Through multiple optimization scenarios, I found that a combined change in machine quantities and processing times yielded the best results, increasing monthly output from 43 to 62 pieces and improving overall equipment utilization. The simulation approach proved to be more systematic and informative than traditional experience-based methods. The model can support production management, process optimization, and system analysis for drive gear gear machining. Future work will focus on refining the model with additional stochastic factors and applying metaheuristic optimization to further improve the system.