Pull-driven Scheduling for Pipe-Spool Installation: Simulation of Lean Construction Technique

Many construction processes include installation of unique materials in specific locations in the facility being built: materials and locations must match before installation can take place. Mismatches due to delay and uncertainty in supplying materials or completing prerequisite work at those locations hamper field productivity. This is illustrated here using a model of a materials-management process with a matching problem that typifies fast-track process-plant projects. The uniqueness of materials and locations combined with the unpredictability in duration and variation in execution quality of various steps in the supply chain allow for different ways to sequence material delivery and work area completion. Several alternatives are described. Their impact on process execution is illustrated by means of probabilistic process models. One model reflects total lack of coordination between delivery and work area completion prior to the start of construction; a second one describes perfect coordination. The corresponding materials staging buffers and construction progress are plotted based on output from discrete-event simulation models. A third probabilistic model then illustrates the use of the lean construction technique called pull-driven scheduling. Real-time feedback regarding the status of progress on site is provided to the fabricator off site so process steps can be re-sequenced opportunistically. This yields smaller buffers and earlier project completion and, when properly accounted for, increased productivity. 1 Associate Professor, Civil and Environmental Engineering Department, 215-A McLaughlin Hall, University of California, Berkeley, CA 94720-1712, 510/643-8678, [email protected] INTRODUCTION Construction involves installing materials according to project specifications in the facility being built. By tracking the flow of materials through their supply chain (i.e., describing when and where materials are being engineered, fabricated, transported, staged, etc.) installation work can be most effectively planned and executed. Flow data must be more or less detailed depending on whether the material of concern will be available in large quantities of identical, interchangeable units (e.g., concrete blocks, electrical conduit, nuts and bolts); in modest quantities, possibly with some degree of interchangeability (e.g., windows, structural steel, timber in precut lengths), or in small quantities of units with unique properties (e.g., engineered materials such as pipe spools or a custom-designed main entrance door). Field installation crews, responsible for the final step in the materials flow process, must find resources that match among those available to them; they must ensure that the right material gets put in the right place. For instance, they must identify the location where installation is to take place (e.g., area AR-123), then find the matching material (e.g., pipe spool SP-123) and retrieve the correct installation accessories (e.g., attachments and supports). An integral part of their work, time and again, is to solve the so-called "matching problem." In facilities that comprise thousands of materials of which many are unique, tackling the matching problem is an enormous task. Nevertheless, those performing installation have no way around it. In contrast, those responsible for engineering and design, fabrication, delivery, and site storage of materials, as well as construction managers overseeing the project often overlook the matching problem that installation crews face. Dealing with materials on an item-by-item basis means paying attention to minute details. It is a tedious task, largely irrelevant to their own. Accordingly, matching-problem details are selectively abstracted away by each party so that they can focus on problems of more direct, contractual concern to them. For example, structural designers do not worry about vendors’ ability to deliver specialty valves or nuts-and-bolts because it is outside of their scope of work. Pipe-spool fabricators optimize production schedules to suit their plant’s fabrication constraints and other projects’ needs. Shipping agents optimize travel by choosing vehicles to meet delivery schedules; they package materials to ensure that loads are stable and meet weight and dimensional constraints during transportation. Laydown yard personnel group materials by shipment, type, or finalinstallation destination to ease tracking. Project managers control progress based on percentages-of-total of materials engineered, delivered to the site, or installed. The corresponding planning systems must therefore allow for abstraction or detail as needed. Because of this abstraction, installation crews rarely have the data they need to optimally schedule and thus execute their work. They must rely on the numerous assumptions that are embedded in pre-construction schedules. How much of a problem this creates depends on the extent to which uncertainties in their supply manifest themselves during project execution. If pre-construction schedules were well thought-out and steps preceding installation had no uncertainty in duration or execution quality associated with them, then matching would be easy. In practice, unfortunately, this is not the case. Many projects are executed on a fast track, so construction starts before design has been completed or materials deliveries have been properly sequenced. Installation crews and equipment are often kept waiting because delays in materials supply and delays ASCE Journal of Construction Engineering and Management, July/August 1998, 124 (4) 279-288 2 in completing prerequisite site work lead to mis-matches that foul up scheduled work sequences. This lowers the installation crew’s productivity and extends the project duration. In order to increase understanding of these issues, a model was created of a process that is characteristic of the processplant sector of the construction industry. Alternative strategies for sequencing materials deliveries are presented in this paper and their execution was simulated so computer data supports the comparison between them. RELATED WORK IN LEAN CONSTRUCTION Matching problems and process uncertainties pose unique requirements on construction planning systems. An analogy with manufacturing production systems is appropriate to explain what these are. Specifically, the lean production philosophy is relevant (Ohno 1988). Lean production focuses on adding value to a raw material as it proceeds through various processing steps to end up as a finished product. It advocates the avoidance, elimination, or at least reduction of waste from this so-called value stream. By considering waste not only in or produced by individual operations but in the value stream at large, lean production adopts a systems view. The late Taiichi Ohno first articulated this philosophy and implemented it in Toyota’s production system. He classified sources of waste as follows (8 added by Womack and Jones 1996): (1) Defects in products; (2) Overproduction of goods not needed; (3) Inventories of goods awaiting further processing or consumption; (4) Unnecessary processing; (5) Unnecessary movement of people; (6) Unnecessary transport of goods; (7) Waiting by employees for process equipment to finish its work or for an upstream activity to complete; and (8) Design of goods and services that fail to meet user’s needs. The lean production philosophy, since it emerged in the 1950s, has provided major competitive advantage to Japanese manufacturing companies. Its benefits gradually became known outside of Japan. In the 1980s, US manufacturing companies began to convert their operations to implement lean production techniques and, consequently, also improved their operations dramatically (Womack and Jones 1996). Some lean production techniques are: (1) Stopping the assembly line to immediately repair quality defects; (2) Pulling materials through the production system to meet specific customer demands; (3) Reducing overall process cycle time by minimizing each machine’s change-over time; (4) Synchronizing and physically aligning all steps in the production process; (5) Clearly documenting, updating, and constantly reporting the status of all process flows to all involved. Though no one will doubt that there is much waste in construction, lean production has only recently become a subject of interest in our industry. Since the publication of Koskela’s (1992) seminal report, researchers around the world have been studying its applicability to construction (e.g., Alarcon 1997). Unfortunately, translating lean concepts from manufacturing to construction is not automatic because of the unique characteristics of the architecture/engineering/construction (AEC) industry in addition to the geographic diversity among projects. Researchers in construction have begun to realize that construction management must include production control systems (e.g., Bernold and Salim 1993, Melles and Wamelink 1993) to complement the project management systems currently in use. Control systems must include not only activities being performed at the project site but also those that make up the entire supply chain (O’Brien 1995). The work described here belongs to this school of thought. Some lean concepts have already been translated to construction. Howell et al. (1993) discussed how buffers of materials can alleviate the dependencies and worker idle time otherwise incurred when process sub-cycles interact with one another. Ballard formalized the Last Planner to shield installation crews from uncertainties in work flow and demonstrated its successful implementation on actual projects (Howell and Ballard 1996, Ballard and Howell 1997). Phair et al. (1997) reported how equipment manufacturers are reducing set-up time by changing product designs (e.g., buckets and other attachments). In the same vein, this paper describes how the pull technique with feedback regarding progress on site to fabricators off site can improve construction process performance (Tommelein 1997a, 1997b). PUSH-DRIVEN VS. PULL-DRIVEN PROCESS MANAGEMENT Push-Driven Process Management Construction work traditionally is planned by articulating activities and dependencies between them, then assigning durations and resources to each activity. A schedule is developed by calculating early and late activity starts and finishes using the Critical Path Method (CPM). Resource leveling or allocation algorithms may yield some adjustments to the early-start schedule, but upon project execution, activities are expected to start at their earliest possible date in order not to delay succeeding activities or the project as a whole. Project control aims at adhering to the resulting schedule. It is assumed that all resources required to perform an activity that is about to start will indeed be available at that activity’s earlystart time. In this so-called "push-driven" approach, each activity passively waits for its ingredients (instructions, labor, materials, equipment, and space) to become available, e.g., by being released upon completion of predecessor activities. When some have become available but others needed at the same time have not, those available will wait in a queue or buffer for the combination of resources—the set of "matching parts"—in its entirety to be ready. While it may be possible to start work with an incomplete set of resources, chances are this will negatively affect productivity (e.g., Thomas et al. 1989, Howell et al. 1993). Because of uncertainty in duration as well as variation in execution quality and dependency logic of activities, schedules are bound to change as construction progresses. It may be possible to model this uncertainty during the planning stage, as is done by using probabilistic distributions to characterize activity durations in the Program Evaluation and Review Technique (PERT). However, the actual manifestation of uncertainty is known only upon plan execution and must thus be ASCE Journal of Construction Engineering and Management, July/August 1998, 124 (4) 279-288 3 dealt with in real time. At that point, rigorously adhering to the initial schedule may not be the best approach for successful project completion as network characteristics and resource availability will deviate from those assumed when that schedule was generated. Moreover, traditional CPM schedules do not necessarily show individual resources and their allocation to activities. Certainly, procurement schedules highlight milestone delivery dates of major items, but most materials will arrive in multi-unit shipments. If a schedule reflects only groupings, then it is too coarse to guide work that involves unique parts. When missing parts are identified during the on-site allocation process, it is much too late to prevent delays. In addition, current expediting practice is to regularly touch base, e.g., with the engineering design firm or fabricator of whom goods or services are expected. Contact is made prior to the deadline of completion of their work, in order to make sure the target delivery date, e.g., of key materials or pieces of equipment, will be met. Yet, most expediters fail to (e.g., are not authorized to) reschedule activities when it can be anticipated that deadlines will not be met. Accordingly, the traditional, push-driven approach to scheduling prior to the start of construction with no corrective re-scheduling as work progresses leads to process inefficiencies and less-than-optimal project performance. Pull-Driven Process Management The main objective of a "pull-driven" approach is to produce finished products as optimally as possible in terms of quality, time, and cost, so as to satisfy customer demand. Achieving high process throughput while minimizing operating expenses including in-process inventories is key. Keeping busy by processing just any one of the resources in the input queue of an activity requiring a combination of resources is insufficient. To pull means that resources must be selectively drawn from queues—so the activity that processes them will be busy just the same—but chosen so that the activity's output is a product needed further downstream in the process, and needed more so than its output using other resources in the queue would have been. Resources' wait time in queues should be minimized. To implement a pull-driven approach, selective control is needed over which resources to draw for any given activity. This selection is driven by information not solely about resources in the queues immediately preceding the activity under consideration, but also about work-in-progress and resources downstream (successor queues and activities) in the process. Resources will get priority over others in the same queue if they are known to match up with resources forecast to be or already available in queues further downstream in the process. This way, those downstream resources will not unduly await their match and be in process for any time longer than needed, though their planned processing sequence may be violated. EXAMPLE PROCESS SCENARIO: PIPE-SPOOL INSTALLATION Constructing an industrial process facility, such as an oil refinery, involves installing many hundreds or thousands of unique pipe spools. This process is simplified here as comprising two chains of activities: pipe spools are designed and fabricated off site while work areas are prepared on site. After spools have been shipped to the site, the chains merge with the installation of spools in their designated areas. Pipe spools are fabricated off site according to the availability of engineering design information, the fabricator's plant production capacity, etc. Individual tags denote that each spool has unique properties and each has a designated destination in the facility under construction as shown in the project specifications. Spools are subject to inspection before leaving the fabricator's plant. The outcome of the inspection activity is that a spool will be found fit-for-installation with an x% likelihood, and, thus, that there will be a problem with (100 x)% of them. In the latter case, the fabricator must rework the spool to rectify the problem, prior to shipping. Concurrently with this off-site process, construction is under way on site. Roads are built, temporary facilities are brought in, foundation systems are put in place, structural steel is being erected, etc. Crews of various trades must complete their work in each area where spools are to be hung, prior to spool installation. When a specific set of ready-for-installation spools is available on site, and all prerequisite work in the matching area has been completed, spools can be installed. Completion of an area's installation work then signals to other trades that subsequent work can start.