Snowplow Operations and Resource Management

The purpose of this research was twofold: to develop a simulation model of snowplow operations, and to develop a conceptual design for a predictive maintenance system. The first portion of the research was concerned primarily with developing a simulation model of snowplow operations for selected routes in Virginia, MN. The purpose of this model is to assist managers with decisions related to route length, assignment of plows to routes, placement of reloading points, and the collection of labor and material cost based on "what-if" scenarios. Creating this model depended on input from experienced operators, internal management reports, and archived Road and Weather Information Systems (RWIS) data. One important input variable to the model was plow speed. It was assumed that plow speed was affected by both snow accumulation rate and moisture content. The researchers relied on input from experienced operators to determine the effect of accumulation rates and moisture content on plow speed. For each event during the 2002-2003 winter, Mn/DOT reports were used to determine the length of each event, and the time to reach bare pavement. The archived RWIS data was then matched to each event to estimate snow accumulation rates and relative humidity, which was used as a pseudo-measure for moisture content of the snow. Using the estimated plow speeds, the simulation model was run for selected routes and the output of the model was then compared to the bare lane reports. Although the model structure, logic, and design were verified to be correct, the model could not be validated, as the results did not match the bare lane reports. This problem can be attributed to either incorrect input data, or incorrect reports that were used to validate the model. It was not possible to determine the source of the inaccuracy. In spite of this shortcoming, other aspects of the project were successful. A user interface was developed for the simulation model, which operates from an Excel spreadsheet. The process of developing the interface was very instructive for both the researchers and the supervisors, as details of snowplow operations were articulated and the important variables affecting operations were identified. As the interface was developed, the users identified additional constraints and variables that needed to be addressed and included in the simulation model. Some variables turned out to be relatively unimportant but others turned out to be key. Key variables were pavement temperature, accumulation rate, moisture content, and material application rate, which the user can control from the interface. The supporting Excel spreadsheet were also designed with the flexibility for the user to make additional changes in the future, such as route length, and number of snowplows assigned to a route. During the winter of 2003-2004, several snowplows in Virginia will be equipped with automatic vehicle location systems that will have the ability to determine vehicle locations. This should provide tremendous assistance with model validation by determining plow speed. In summary, the simulation model was successful on many fronts: identifying key variables, determining the operational rules and logic, developing a user interface, and designing model flexibility for future development through Excel spreadsheets. The only disappointment was in model validation, which should be solved when data is collected directly from the plows through the AVL system. It is therefore recommended that data be collected from the snowplows using systems such as AVL or GIS systems. In addition to location information, data should also be collected to determine the rate and quantity of material application. In the future, systems which identify the configuration of the snowplow (e.g. wings up or down, underbelly plow up or down) should be implemented. Gathering this data using real time systems will reduce reporting errors, thereby improving any models that are used to depict snowplowing operations and also improving the reliability of internal management reports. Similar work is being conducted elsewhere. Therefore, it is also recommended that a site visit be made to Ohio, who has recently implemented software developed by Cascade International, for managing snowplow operations. It is also recommended that the researchers for this project work closely with other organizations conducting similar work, which include two private consulting firms and the Army's Cold Weather Research Labs. The other portion of this project was to develop a conceptual model of a predictive maintenance system. The interest in predictive maintenance arose from the snowplow simulation study. It was expected that the simulation model would include snowplow breakdowns, but no data was available to characterize this type of event. Although the plows are subjected to preventive maintenance, there is no predictive maintenance program, which raised concern regarding the potential breakdowns during a snow event. This study was mostly exploratory in nature, but yielded some surprising results. The first is that the research literature on predictive maintenance for vehicles is fairly small, and tends to focus instead on industrial and manufacturing equipment and machinery. The second surprising result is how few departments of transportation have predictive maintenance programs for their vehicles. The Pennsylvania DOT stands out, as it has had a predictive maintenance program in place for several years. It has taken several years to realize the benefits of the program as historical data for each vehicle has been collected. Other DOTs, such as Michigan, conduct predictive maintenance on selected equipment. Michigan performs predictive maintenance on snowplows. Similar to Pennsylvania's experience, Michigan also realized benefits from the predictive maintenance on snowplows, reporting far fewer breakdowns and higher service levels. It is recommended that MnDOT, through a fleet management program, categorize equipment to identify which equipment should have preventive maintenance programs. Next, it is important to determine if the organization is ready to implement a predictive maintenance program and, if so, to identify the vehicles to use for a pilot program. It is also recommended that DOTs who have successfully implemented such programs be contacted to arrange a site visit, and to contact software vendors to identify potential software that will serve not only the needs of the fleet manager, but also the decentralized shops where the maintenance is performed