Model-based design (MBD) is crucial to the timely delivery of embedded control systems for many of Caterpillar’s OEM customers. It is especially critical for machines with multiple electronic control modules (ECMs) running integrated algorithms aimed at maximizing machine performance and fuel efficiency. MBD has helped Caterpillar’s OEM/Custom Products Division leverage proven Cat components to develop highly integrated machine systems for OEMs.
Locomotives with mechanical drive systems are fundamentally analogous to typical heavy-duty drive systems found in many off-highway construction vehicles. The major components are a diesel engine, an automatic mechanical transmission with an integral or separate torque converter, the axles, and final drive system.
A system architecture for a mechanically driven sugar cane locomotive was recently developed by Loadquip Engineering of Australia in conjunction with Caterpillar OEM/Custom Products. Most of the powertrain components are, or were derived from, production components currently used in other Caterpillar products.
The OEM requested the following driveline control functions be provided by the Caterpillar electronic system:
• Shift Control Logic (SCL), including Controlled Throttle Shifting (CTS)
• Traction control for both accelerating and braking conditions
• Integrated engine compression brake and retarder braking system control
• Programmable electronic Vehicle Speed Limiting (VSL)
• Interface to existing OEM operator control electronics.
The first development step for a powertrain plant model was to obtain and/or develop/modify component models. Component models need to provide fidelity suitable for the algorithms to be developed, as well as the interface I/O necessary to design the highly integrated electronic control system. A Caterpillar-developed physical components model library served as the primary model source for the cane locomotive plant model. Some components, such as the final drives and torque converter, were acquired as-is from the Caterpillar library. Other library component models were customized to provide the functionality necessary for specialized machine functions. For example, the engine component model required the addition of J1939 TSC1 control inputs and the associated internal fuel system interaction functionality that would be necessary for developing traction control and vehicle speed limiting algorithms. A model of the customer-designed oil-cooled brake pack was developed from functional descriptions and specifications provided by the customer.
Once the components were selected and/or developed, they were assembled into a plant model. The plant model was initially verified by using simulation to run the locomotive up and down through the forward and reverse gears at full engine throttle using a primitive transmission output speed (TOS) based shift control algorithm. Power consumption analysis was used to verify that the sum of the power losses (rolling resistance, losses due to efficiency, etc.) in the system was equal to the total engine power at a steady state operating condition. Test data was not available for model verification since the plant model preceded the first prototype locomotive. Model validation therefore relied on simple checks and common sense approaches such as comparing basic performance to other similar locomotive systems that had previously been developed, as well as using such tools as power consumption analysis at steady state operating conditions.
Once the plant model had been verified, work began on developing the electronic control system features that the customer requested. The heart of the control system is the shift control logic (SCL). SCL selects the appropriate transmission gear based upon operating conditions and includes all of the strategies necessary to protect the system from potentially damaging overspeed or underspeed conditions. SCL also includes algorithms that prevent hunting and ensure predictable and responsive shifts for the locomotive operator. The first pass shift points were based upon a static analysis done to optimize locomotive acceleration during full throttle conditions. Further refinements came from more detailed analysis and simulation of specific operating conditions (e.g. various grades, towed loads, driving surface friction coefficients). As the SCL algorithm matured, development of other features could proceed. Finally, the more advanced features were added around the SCL core to provide the customer specified system functionality.
Due to the limited time allowed for field tests, extensive simulation studies were relied on to thoroughly debug the control model and investigate potential system interactions. Some of the more relevant issues and unexpected benefits of this analysis include:
• Discovered need for specialized slip/slide detection in SCL to prevent potential engine damage;
• Verified proper cooler sizing for customer supplied oil-cooled brake system;
• Discovered improper assumptions in early shift point analysis;
• Determined optimal test track configuration for eventual field validation.
Control development via simulation minimized on-machine validation and significantly reduced cost by avoiding overseas travel by the control engineers. On-machine validation was completed by a technician running predetermined test cases and sending results back to the engineers for analysis. The entire control system was fully validated and approved by the customer in less than one week of on-machine test time, much to the surprise and delight of the customer who was under pressure to get the new cane locomotive into the fields and operating for the upcoming cane season.
MBD of the cane locomotive control system saved an estimated two weeks and more than $60,000. The additional time needed up front for plant model development was more than offset by a large reduction in field test time and the elimination of overseas travel. The cane locomotive control was completely developed via MBD in five weeks of modeling and algorithm development and one week of field tuning by a technician. A non-MBD design process would include building initial control algorithms directly from the customer’s specifications followed by field testing to debug, modify, and validate the final design. Using this approach for the cane locomotive would have required travel to Australia and four weeks of engineering on-site field testing and development.