DETAILED DESCRIPTION

[0041] An intelligent manufacturing system for tracking production information from one or more manufacturing facilities has been developed. The system is known as PIPE (Process Information Per Event). PIPE collects, stores, and reports production information such as converting machine productivity, waste, and delay information on an event basis. In this system, machine data from sensors and other control means are continually monitored for events related to productivity and/or product quality, such as product waste, machine down time, machine slow downs, product maintenance, machine failure, etc. Customized rules may be established to specify how events are classified and what types of events are to be logged (normally, all sources of delay may be logged and coupled with additional data). These events may be spaced apart in time by time steps that typically are not constant, and may be substantially randomly spaced in time, or may be characterized in that the standard deviation of the time step between successive events is large relative to the mean, such that the ratio of the standard deviation to the mean time step during a week of production is about 0.2 or greater, specifically about 0.5 or greater, and most specifically about 1.0 or greater. Time steps between successive events may range, for example, from a few seconds or minutes to hours or days, depending on the process.

[0042] An “event,” as used herein, refers to any incident that may affect the productivity of a process or machine in use to produce a product, or that may adversely affect the quality of the product being produced. Events that adversely affect the productivity of a process or machine by increasing delay are “adverse productivity events.” Productivity events that lead to waste are “waste events,” while those that cause delay are “delay events.” Events that adversely affect the quality of a product are “adverse quality events.” As used herein, “intermediate events” may refer to incidents during a first process for the production of an intermediate product to be used as a raw material (starting material) in a second process for the production of a finished product (or another intermediate product or product component), wherein the incident in the first process may affect the productivity of the second process or adversely affect the quality of the product of the second process. Thus, an intermediate event in a first process may lead to an adverse productivity event or an adverse quality event in a second process. An adverse quality event may also refer to incidents that may adversely affect the quality of an intermediate product, such that the risk of rejection of the product by a subsequent user (including an industrial user) is increased. The PIPE system may be used to track any or all types of events, including events from multiple machines and processes wherein intermediate products from early processes or machines are used as raw materials in later processes or machines, and optionally wherein the event data for the intermediate products are used by operators or process control equipment to properly execute the subsequent processes based on the events associated with the intermediate product or, in general, with the quality and property attributes of the intermediate product as recorded at least in part with a system including PIPE.

[0043] Examples of events may include a web break, a component failure in a machine, a loss of manpower (e.g., inadequate employees present during a shift), a loss of power, a fire, machine shutdown to change a grade (“changeover”) or perform routine maintenance, unacceptable quality in raw materials, market curtailment (e.g., inadequate orders or excess inventory), an experimental run, a startup, and the like.

[0044] As used herein, “production information” includes waste data, delay data, and any other data related to production. In some systems, production information is segregated from waste and delay data, even though waste and delay data are considered production information. In general, the invention is operable with any form of waste data, delay data, or other production information or event data. For example, production information includes raw material usage information. Raw material usage information includes, but is not limited to, a raw material lot number, an amount of material in a roll, a time the roll was spliced on or off, a supplier of the material, a number of products produced from the roll of material, and a date the material was produced.

[0045] PIPE event data obtained during production are stored in a database associated with descriptor information. This information may be used to generate financial reports automatically for use by an accounting department, a plant manager, financial officers, or others, or for use in an internal or public publication such as a report or web page. The PIPE information may be rolled up from multiple machines, plants, sectors, and so forth, including a corporate-wide roll-up of PIPE data, to provide roll-up productivity measures.

[0046] The data from the machine are monitored and logged by a PIPE Event Logger, which may include an event logger and a machine logger. The event logger may also serve many functions in addition to receiving and processing data, such as ensuring that the raw materials fit the specifications for the product to be made (in cooperation with a separate raw materials tracking system described hereafter), or linking operating data to the PIPE database, or ensuring that adequate explanations have been entered by operators to explain delay states that occurred on the machine. The machine logger provides an interface for operators to provide explanations about delay states or product waste, but generally does not collect data from sensors or production equipment. The event logger and the machine logger may be separate programs or be part of a single program, or functions of both may be shared or split between multiple programs and servers.

[0047] The system may be structured to support multiple converting lines in multiple plants as an enterprise information system. According to the present invention, a plant information system or enterprise information system may be adapted to allow corporate financial and purchasing systems to receive information from the PIPE system for direct use. Production reporting systems may be directly linked to multiple PIPE data streams to provide rolled-up financial information or information for a single asset. The PIPE system and its accounting module may be interfaced with or cooperatively associated with accounting software such as SAP brand software and SAP/R3, process control software such as WONDERWARE brand manufacturing and process control operator-machine interface software (Wonderware Corp., Irvine, Calif.), neural networks, expert systems, fuzzy logic systems, and many other suitable software systems. Further, the PIPE system may automatically submit work requests and purchase orders to deal with causes of delays (particularly equipment failure) as they are encountered. Further, the PIPE system may be used to mine process and quality data to identify means to improve productivity or quality.

[0048] Data from the PIPE system may also be integrated with other software systems for financial tracking, production management and planning, supply chain management, inventory control, maintenance and reliability engineering, customer relationship management (CRM), and the like. For example, PIPE data may be included in the sources of information treated by POWERFACTORE software from KPMG Consulting (McLean, Va.). With this approach, the relationship between system maintenance schedules and product quality may also be explored to optimize operations to improve financial returns.

[0049] PIPE data may also be integrated with data warehousing systems such as the SAS INTELLIGENT WAREHOUSING SOLUTION marketed by the SAS Institute, Inc. (Cary, N.C.) and the KALIDO brand computer database management programs by Kalido, Inc. (Houston, Tex.) such as the Dynamic Information Warehouse. Likewise, SAS/INTRNET brand computer software and SAS online analytical processing (OLAP) technology from the SAS Institute, Inc. may be combined with the PIPE system. Other exemplary OLAP tools include the ESSBASE DB2 OLAP software from Hyperion Solutions Corporation (Sunnyvale, Calif.) and COGNOS POWERPLAY of Cognos Incorporated (Ottawa, Canada). General principles on the combination of OLAP with data warehousing are disclosed by Surajit Chaudhuri and Umeshwar Dayal, “An Overview of Data Warehousing and OLAP Technology,” ACM Sigmod Record, March 1997.

[0050] Other data warehousing and maintenance methods may be applied. By way of example, principles of data warehousing and warehousing technology are disclosed in U.S. Pat. No. 6,418,450, “Data Warehouse Programs Architecture,” issued Jul. 9, 2002 to Daudenarde; U.S. Pat. No. 6,353,835, “Technique for Effectively Maintaining Materialized Views In A Data Warehouse,” issued Mar. 5, 2002 to Lieuwen; U.S. Pat. No. 6,178,418, “Distributed Data Warehouse Query and Resource Management System,” issued Jan. 23, 2001 to Singer; U.S. Pat. No. 6,138,121, “Network Management Event Storage and Manipulation Using Relational Database Technology in a Data Warehouse,” issued Oct. 24, 2000 to Costa et al.; and U.S. Pat. No. 5,781,911, “Integrated System and Method of Data Warehousing and Delivery,” issued Jul. 14, 1998 to Young et al. Historical, summarized, and consolidated data are typically present in data warehouses, which may be queried to guide decision making and the development of business plans, or to prepare summary financial reports.

[0051] In addition, the PIPE system may be combined with NET PROPLAN and other manufacturing execution systems (MES) software systems and modules in the NET COLLECTION by Network Systems International, Inc. (Greensboro, N.C.). For example, the PIPE data may be integrated with the NET SCHEDULER module and the NET EVENT TRACKER system. In addition, the PIPE system may be integrated or modified to communicate with the FOLDERS system and the FACTELLIGENCE brand software for assisting manufacturing operations by CIMNET (Robesonia, Pa.).

[0052] Enterprise Resource Planning (ERP) systems may be coupled with PIPE systems. Exemplary ERP systems include those marketed by suppliers such as SAP (Newtown Square, Pa.), J D Edwards (Denver, Colo.), Manugistics (Rockville, Md.), Siebel Systems (San Mateo, Calif.), ROI Systems (Minneapolis, Minn.) including the MANAGE 2000 brand pre-recorded computer programs, or custom built systems. An exemplary tool for integrating PIPE data and other data with ERP systems (SAP R/3 systems in particular) and generating financial reports is DATA INTEGRATOR of Business Objects Americas, Inc. (San Jose, Calif.).

[0053] Existing software and known methods may be used to determine the financial costs of waste and delays. A computer system for determining the financial cost of various production problems and process bottlenecks is disclosed by Van Der Vegt and Thompson in U.S. Pat. No. 6,144,893, issued Nov. 7, 2000, and in U.S. Pat. No. 6,128,540, issued Oct. 3, 2000, both of which are herein incorporated by reference to the extent they are non-contradictory herewith. Columns 1 to 12 in U.S. Pat. No. 6,144,893 disclose the computer method, and columns 12 to 19 therein disclose a method for generating a problem priority table for problems in the process. The determination of the cost of a process problem may be calculated based on whether the process is constrained by production limitations or whether the process is sales constrained (demand for the product is less than the maximum capacity of the machine).

[0054] Integrated systems, in which PIPE and other systems tie into purchasing and financial systems, may be used for many purposes. For example, information about a machine failure detected by PIPE may be used to automatically order a failed part with an asset management process utilizing SAP or other systems. Production tracked with PIPE may be combined with financial reporting tools, components, or modules as well. Neural network/fuzzy logic analysis of PIPE and related data, including raw material data that is linked to the PIPE system via a raw material tracking system, may be used to optimize profitability and improve process control, identify weaknesses in systems, parts, or vendor performance, and so forth. Results may be displayed on a web page to local or remote viewers (typically authorized viewers only); displayed via a client (e.g., through a window on a monitor for a Human-Machine Interface such as WONDERWARE brand manufacturing and process control operator-machine interface software); incorporated into weekly, monthly, and annual reports; used to guide daily operations; and so forth. Time series of productivity parameters, such as measures of waste or delay may be displayed graphically to show trends or ranges, in tabulated form, including means and for various periods of time, and so forth. Productivity results may be sorted and/or displayed according to sector, machine type, product classification, geographical location, technology or raw material types used in production (to examine the effect of a change in a production technology or raw material implemented at one or more plants), and the like. In generating reports, any suitable type of chart or graph may be used, and results may be put into any suitable software format.

[0055] PIPE may be adapted to provide information for key performance indicators (KPIs) expressed in terms of common performance measures, wherein a standardized definition and formula based on PIPE data is applied. For example, one KPI may be percent total waste, expressed as the ratio of the total number of products discarded to the total products made. KPIs are identified by financial departments to describe profitability, efficiency, production rates, etc., for individual machines, plants, groups of plants, product categories, and so forth. Another KPI may be system rate, which is the actual machine speed divided by target speed, commonly expressed as a percentage. Actual speed may be defined as total standard units produced divided by actual hours of operation. Waste may be calculated as total units produced minus acceptable units products, or may be expressed as a percentage, (total units−acceptable units)/total units×100%. Percent yield may be expressed as 100%−percent waste. Efficiency may be expressed as percent uptime×percent yield/100%. Percent reliability may be expressed as system rate*percent uptime*percent yield/10,000%.

[0056] Another factor that may be used to characterize the productivity of a machine is the “Rate of Operation” (R/O), defined as the number of non-rejected standard unit products produced per hour (standard unit products is the number of products divided by a sector standardized unit; for example, a standard unit of diapers could be set at 1,000, to 50,000 diapers would be 50 standard units).

[0057] Machines or processes also may be evaluated in terms of opportunity costs, which generally refer to the financial cost of waste or delay (include slow machine speed). As used herein, “Waste Opportunity Cost” is the direct cost of wasted products plus the delay cost. As used herein, the “Delay Opportunity Cost” is the direct cost of machine down time plus the cost of wasted product as a result of restarting the machine plus the cost of time that is spent disposing of wasted product. Also as used herein, the “Slow Running Opportunity Cost” is the cost of the machine running at a speed less than the ideal speed (determined on a per machine level), producing less product as a result.

[0058] To achieve standardized reporting, PIPE systems may provide information about production modes. Production modes may describe the status of a machine at any given moment, such as whether a machine is operating, down for scheduled maintenance, being used for a research run, and so forth. The production mode information from the PIPE system allows down time or delays in production to be counted appropriately by financial departments. Thus, the PIPE output may include fields or records for production mode. In one embodiment, data entered into a production mode field are automatically screened for correctness (e.g., upon entry and again upon the next start up), and errors or ambiguities are flagged for correction.

[0059] Other output parameters may be modified to comply with standard definitions required by finance or other users of the data. Thus, in one embodiment, the integrated system and its method of use include a method for adapting an online production documentation system to provide financial report data for a machine, including the steps of identifying one or more key performance indicators pertaining to the machine required for a financial report, modifying the output of the production documentation system to automatically track and generate the key performance indicators suitable for use in a financial report, receiving the key performance indicator information from the production documentation system, and incorporating the key performance indicator information directly into a financial report such as an electronic report (e.g., a web page or electronic chart). In one embodiment, the generated reports are maintained by the production documentation system for a preset time interval for future retrieval. In this manner, frequently requested reports may be delivered quickly with reduced processing overhead.

[0060] PIPE reports may be generated to report on operations on any of several levels, such as at the level of section or subsection of a machine (e.g., monthly defects, waste, or delays in a film production line caused by excessive arcing in a secondary corona treatment section of an apertured film line), at the machine level (e.g., hours of delay per month for the entire film surface treatment converting machine, or percent uptime for a lotion packaging line), for a product code (e.g., a particular type of apertured films for use in sanitary napkins), at the plant level (e.g., percent waste for a film production plant, or a general plant summary), at the sector level (e.g., average percent uptime for all film plants in a sector), or at the corporate level (potential lost sales per quarter based on total waste and delay). The leading (most frequent or most costly) types of waste or delay events may be listed by machine, by section, by plant, by sector, and so forth. Detailed daily, weekly, monthly, or annual reports by machine, plant, or sector may be generated, and may be applied to specific products or product categories. Production, waste, or delay by shift or crew may also be reported.

[0061] Applications of PIPE financial information to the general ledger and various subledgers may be achieved via Charts of Accounts and other tools, as described by Dan Hughes, “Designing the Financial Data Warehouse.”

[0062] In the past, there was typically a significant delay between the acquisition of data pertaining to productivity, loss, or waste for a machine and the generation of a corresponding report for review by management or incorporation into Corporate reports. Further, such reports were generally limited in terms of what could be displayed, often being static reports rather than live, customizable reports. The scope of the present invention includes an automated reporting system adapted to provide more timely and flexible reports based on PIPE data which may be provided to management, incorporated into corporate reports or intranet pages, used to call for remedial action or other decision making processes, and the like.

[0063] In one embodiment, a method for automatically generating an alert comprising a financial report based on event data comprises:

[0064] a) setting alert criteria for automatic report generation of an alert, such a setting including a cost threshold for a predetermined unit of time (e.g., a shift, day, or week, or moving time frames such as the past hour, 24 hours, 3 days, week, and so forth), such as the total cost of waste and delay during the unit of time, the total cost of waste and delay from a specified subcategory of event types during the unit of time (e.g., web breaks or equipment failure), or, rather than considering costs over a unit of time, also or alternatively setting a threshold for the cost of any single event or any event of a predetermined type (e.g., generate an alert if any waste event has a cost of $2,000 or greater, or results in a lost of at least 500 units of production);

[0065] b) repeatedly calculating costs for events during manufacturing based on event information being recorded in a PIPE database associated with the manufacturing process, the costs being calculated for the events and periods of time specified in the alert criteria, and comparing the costs to the alert criteria to determine if the alert criteria have been met;

[0066] c) in response to the alert criteria being met, automatically generating an alert comprising (or directing attention to) an electronic financial report conveying information pertaining to the costs that have met the alert criteria, and issuing the alert electronically to a supervisor.

[0067] For example, the alert may comprise a message indicating that cumulative waste or delay events during a predetermined period of time have exceeded a specified threshold, and provide a chart showing the top ten categories of waste and delay events in terms of cost, or a table of events showing the nature and cost of the most expensive events or all events that contributed to the alert. The financial report may comprise interactive electronic information such as a bar graph with electronic controls (drop-down box, radio buttons, etc.) to allow the viewer to control the format and content of the displayed information (e.g., selecting the top N waster or delay events as a function of user-selectable periods of time, product categories, machine sections, shifts, and the like). The alert may be sent by e-mail, or another electronic notification means may direct the viewer to use a link to the financial report information that is provided separately from the notice. The user, who may be a supervisor or executive, may then call for remedial action to deal with possible causes of the production problems that led to issuance of the alert. In one embodiment, the method further comprises automatically indicating one or more possible remedial actions that may be taken to reduce the production problem. The indicated remedial actions may be suggested by an expert system or other means, and information on the costs associated with the remedial action may be automatically included to enable better or more rapid decision-making.

[0068] The time delay between the occurrence of an event that contributes to a cost threshold being exceeded and subsequent issuance of an alert coupled with access to electronic financial reports based on event data may be arbitrarily short. The time delay between events and reports according to the present invention may less than a day, less than eight hours, less than an hour, less than ten minutes, less than three minutes, or less than a minute. Indeed, live reports may include financial information about events that have occurred only a few seconds before generation of the live report.

[0069] The time frame for computation of cumulative costs may be a moving time frame, whose endpoint continually advances in time (e.g., a span of one week ending with the current time), or a fixed time frame, with fixed starting and end points, such as the days, weeks or months of the calendar.

[0070] Alerts comprising electronic reports may also be issued to appropriate personnel in response to other information extracted by analysis of event data in the PIPE database. An increased rate of occurrence of one type of event may, for example, be indicative of excessive wear of a machine component. Not only may an alert be sent to maintenance staff that a component is in need of replacement, but the report system may be configured to automatically compile historical event data associated with that particular component of the machine to calculate historical and recent or projected maintenance costs for that component to allow a supervisor to assess the need for improvements in machine or component design to reduce costs associated with maintenance of the component. In one embodiment, a report comprising historical cost information associated with the performance and/or maintenance of a machine component (including the entire machine itself) is generated when problems with the component's performance or maintenance appear to be causing waste and delay at a rate or level beyond a predetermined threshold. In that case, management may be alerted that unusual or unanticipated costs are being accrued and that remedial action may be needed. Again, an expert system may recommend remedial action and include information on the costs associated with the remedial action to enable better or more rapid decision-making.

[0071] A subset of the PIPE system, herein referred to as STORM (System for Tracking Online Raw Materials), may be used to provide database information about raw materials accepted by a plant for use during the production of a product. The STORM system may provide access to raw material properties, vendor information, and so forth. Productivity data obtained by the PIPE system for a product may be combined with raw material information from STORM to provide archived information about the ingredients of a product, to permit analysis of the effect of various raw material attributes on the productivity of the process or the quality of the resulting products, and so forth. Possible functions of the STORM system in the context of the present invention may include:

[0072] Tracking and reporting consumed raw material.

[0073] Linking raw material data to finished or intermediate products.

[0074] Validating raw material (e.g., shutting down the machine if an incorrect raw material is loaded).

[0075] Collecting raw material waste data.

[0076] Rejecting and tracking reject material.

[0077] Tracking partially consumed raw materials (e.g., partially used roll goods or bales).

[0078] Linking specific lots of raw material to machine waste and delay results.

[0079] Linking specific raw material events (e.g., splicing) to machine waste and delay results.

[0080] STORM may employ a separate database of raw material information that may be linked to a PIPE database and software. Raw material or pointers to such data may be integrated as a component of a PIPE database, if desired. A related system for electronically tracking material properties of raw materials and generating certificates of analysis for their use is disclosed in commonly owned U.S. patent application Ser. No. 10/253,200, “Supplier Data Management System,” filed Sep. 23, 2002 by Amy H. Boyd et al., herein incorporated by reference. In this system, raw material data and certificates of correction, as well as information about product specifications, delivery and use dates and locations, etc., may all be included in the PIPE database or linked to data in the PIPE database.

[0081] In general, electronic means of receiving and processing raw material data in order to create electronic certificates of analysis may be integrated with PIPE such that the PIPE database provides access to a certificate of analysis or a link (pointer) to the certificate and its associated data (vendor, manufacture date, raw material properties, test methods used, batch number, date of receipt, etc.), such that the raw material data may be considered in subsequent analysis of delay or waste based on the PIPE database.

[0082] In one embodiment, a raw materials database (e.g., a certificate of analysis database) is used to store and merge raw material data. The data may be provided by a vendor or collected by the manufacturer or both. For converting operations with roll goods, for example, the data may be collected in three steps of the converting process: material load, material start (or splice on), and material expire (or splice off). Prior to loading a raw material onto the converting machine, material label information is transferred to the raw materials database (such as from label bar codes using bar code scanners). This includes material label information such as part number, lot number, and quantity. The converting line keeps a product counter that resets at a fixed preset. This product counter is a reference number that is transferred to the database on certain machine events. At the time a material starts to be consumed and at the time a material expires the converting machine transfers reference information such as a timestamp and product count to the database. This information gets merged with the label information.

[0083] Another source of data that may be combined with a PIPE system is a database of consumer complaints or other post-manufacturing quality indicators. Many producers of consumer products and other goods maintain one or more databases of information obtained from users of products, either from users or consumers contacting the manufacturer to register a complaint (e.g., data logged by customer service representatives, including type of complaint and lot number of the product, if available, or date and place of purchase to help identify the time period of manufacture), or from surveys of users, focus groups, test markets, responses to targeted promotions, and so forth. Such data, when associated with lot numbers or other information regarding the manufacture of the product, may be linked to the corresponding PIPE data. Establishing a link between post-manufacturing quality measures and the PIPE database may permit data analysis to be performed to identify possible relationships between operating conditions and consumer complaints or other measures of quality.

[0084] The PIPE system and other related systems disclosed herein, as well as methods of using such systems for improved productivity, financial reporting, raw materials handling, system optimization, and the like, may be applied to any manufacturing system, including continuous, batch, and semi-continuous manufacturing operations. The present invention may be adapted for a single unit operation, a single machine, a series of unit operations or machines, a group of related or unrelated machines at a single production facility (plant or mill), groups of production facilities (for all production operations or a subset thereof, such as operations of a single type or for a single product), or for corporate-wide operations for all products or a subset of products and processes. Exemplary products include cosmetics and toiletries, health care products, absorbent articles such as diapers or feminine care products, foods such as baby food or canned goods, paper and tissue products, pharmaceutical products, automobiles, electronic goods, petrochemicals, agricultural products, wood products, textiles, plastics, and the like. In one embodiment, the PIPE system, including any of the STORM system, the PipeMap utility, and the PIPE Data logger, may be adapted for products produced under regulatory guidelines such as FDA regulations, and includes audit tools needed for Good Manufacturing Practices (GMP). For example, pertinent data from the PIPE system and other sources may be archived and verified with electronic signatures. Raw materials sources and their certificates of analysis may be recorded electronically and associated with the materials produced. Information regarding the recipes, materials, process conditions, crewmembers, and other issues may be electronically recorded and associated with the archived data for future audits or reviews.

Tracking Delay and Waste

[0085] The PIPE system may be used to track delay and waste, or other productivity problems, as well as the apparent causes of those problems.

[0086] As used herein, “delay time” for a machine is any time when products are not being made during a time that was scheduled for production of a product. Even if the delay is due to circumstances outside the control of the company or plant, such as a shipment of raw materials from a vendor that has been delayed due to bad weather or that was shipped to the wrong plant due to a clerical error on the part of the vendor, the result is still delay of production.

[0087] Whenever a converting line stops, a delay record is created in the database. The record may include fields such as delay code, delay duration, timestamp, product count, and other information from the machine controller (see FIG. 10 A and FIG. 10B ).

[0088] In addition to delay data, waste data may also be obtained in much the same way. Whenever defective product is culled (e.g., culled in response to machine vision sensors in a converting line), a waste record is created in the database using a waste code, number of defects, timestamp, product count, and other information from the machine controller (see FIG. 10 A and FIG. 10B ).

[0089] In general, waste and delay information, as well as other productivity parameters, may be automatically captured on an event basis and stored in the PIPE database.

[0090] Productivity and performance of a machine, plant, or business unit may be reported using any suitable set of measures. For example, the total available hours for a reporting period (typically taken as 24 hours per day multiplied by the number of days in the reporting period) may be reported in terms of several categories, such as:

[0091] Development outages, which may include down time for special research runs;

[0092] Market-driven curtailment, when a machine is taken down deliberately because of inadequate sales or due to inventory factors;

[0093] Planned asset outages, as approved by manufacturing leadership;

[0094] Planned holiday shutdowns;

[0095] Force majeure, when an uncontrollable event precludes operation, including floods, hurricanes, disruption of energy supply, etc. Catastrophic equipment failure for reasons other than acts of nature would not normally be included under force majeure.

[0096] The total available hours, minus the sum of any hours falling into the five categories immediately above (development outages, etc.) may be taken as the scheduled hours. The delay hours are the total number of hours that the unit is precluded from operating for any reason during scheduled hours. The scheduled hours minus the delay hours is the actual hours operated.

[0097] In one embodiment, “preliminary waste data” may also be obtained and stored by the PIPE system. “Preliminary waste data,” as used herein, refers to data regarding defects encountered or observed during production of an intermediate product in a first process, wherein the intermediate product is intended for use as a raw material in a second process, and wherein the defects did not cause waste in the first process but are likely to cause waste in the second process. Thus, for example, production of a roll of tissue for use as a barrier material in a diaper may lead to a PIPE table in the PIPE database describing events encountered during the production of the tissue, including machine vision or other sensor input pointing to a serious defect in product quality, such as a hole or tear in the web at a particular distance into the web from the exposed outer end of the roll (e.g., 113 yards from the end of a 200-yard-long rolled web). The problem may not have created a need for discarding the defective portion of the product, which continued to be wound until the defective region was deep within a large roll of tissue web ready for use in a diaper line. Thus, no waste or delay was incurred, but a known problem has been detected in a product quality event that was recorded in the PIPE database for the product. When the product is subsequently used, the PIPE database may again be accessed to alert the second process and its control system of a position in the roll having a defect that will need to be eliminated by culling the affected product or culling the respective portion of the web before it is incorporated into a final product. In other words, PIPE data during a first process is used for feed-forward control of a second process. The “preliminary waste” of the intermediate product thus became actual waste in the final product, but with improved control over the second process.

[0098] The PIPE data may be correlated with process information and other parameters, such as the nature of the shift or crew, material properties of raw materials, season of the year, etc., to better predict causes of waste and to better align machine operation and the “recipe” for the product being made to ensure the less waste is encountered in future production efforts.

[0099] In one embodiment, the PIPE system provides a software application to allow users to view and add comments to the event records. The tools for adding comments may provide a customizable, multi-level menu structure (e.g., machine section, sub-section, problem, root cause, action, and comment) for user entry of machine delay reasons. In one embodiment, a neural network system continually processes event records to mine the database for information that may allow reduced waste. In another embodiment, a fuzzy logic expert system scans operator input to check for discrepancies, as well as to suggest improvements in operation to reduce waste and delay.

[0100] The PIPE system may also assist in identifying the various apparent causes of delay. For example, when delay is due to force majeure that persists for a prolonged period of time, one may recognize that the problem will persist and alter the schedule of time. In this case, one may wish to only count as delay the first time unit in which the force majeure occurred, the time unit being chosen as desired from units such as a shift of eight hours, a day, or other period of time.

[0101] The PIPE system may also account for down time due to market curtailment, wherein the machine has excess capacity due to inadequate customer orders, or because inventory of a product is sufficient to supply customer demands for a period of time without producing more product.

PIPE Fields and Tables

[0102] The PIPE database includes output tables and support tables with fields that specify the machine and numerous aspects of the performance of a machine or process. The output tables include information obtained from a production event, such as a delay or waste event. Fields that may be of use in an output table for delay, by way of example, may include:

[0103] Machine Reference—a field identifying the machine

[0104] Timestamp—a field giving the date and time of the event

[0105] Delay Code—a field indicating the nature of a delay, which may be related to the alarm in a programmable logic controller (PLC) that caused the machine to stop. The delay code may be linked to a particular PLC and machine section, and to a particular cause of delay, or the field may be more general and be coupled with additional fields for section and details of the delay.

[0106] Delay Trigger—a field indicating whether a delay code was caused by a manual/operator stop.

[0107] Grade Shift—a field indicating when a shift in the grade of production began (or other shifts in production parameters, such as selection of raw materials, if desired), to serve as a reference to a Grade Shift table, as illustrated below.

[0108] Operator comments—a field containing text entered by an operator. Alternatively, this may be stored in a separate table to which a link may be established based on the timestamp or other information.

[0109] Duration—the length of a delay

[0110] The output tables may be supported by support tables (lookup tables or maps) that are used to interpret information in the output tables and provide links to other information in other databases or tables. Support tables provide relatively static information to be used in conjunction with the active output tables. Support tables may be developed for delay events, waste events, quality problems, and so forth. Exemplary fields in support tables for a delay event may include:

[0111] Delay Code—a field indicating the nature of a delay, which may be related to the alarm in the PLC that caused the machine to stop.

[0112] Description—a field that contains a brief description of the Delay Code.

[0113] PLC Address—the PLC Address that relates to the Delay Code.

[0114] Section—the section of the machine in which the alarm was generated

[0115] Machine Type—a field indicating the type of machine

[0116] Alarm Source—a field indicating where the alarm originated (which PLC/processor)

[0117] The choice of how tables are constructed and linked, and which fields are used, may be the subject of many alternatives known to those skilled in the art. The specific examples shown for exemplary purposes here are not intended to limit the scope of the invention.

[0118] By way of example only, a line of an output table may include the data shown in Table 1, which indicates what machine is being used, what code describes the delay, and when the delay occurred. The table also indicates when the current grade began being produced (the Grade Shift Start time). Many other fields (not shown) may be present as well, including fields indicating what machine components were involved, which crew and shift was involved, what product was being made, which recipe file was being used, what corrective actions were taken, who made the corrective actions, what the machine speed was prior to failure or averaged during the time since the last event, whether any planned maintenance occurred, whether the down time was a scheduled down, etc. Other records may include data or provide links to data from selected sensors for parameters such as air temperature, humidity, process water pH, etc., which may be known to be relevant to runnability or quality. A field for general operator comments may also be included (not shown). 1

[0119] Further, a line of a support table serving as a delay map may include the data shown in Table 2, wherein the meaning of code 257 in the Delay Code field is given. 2

[0120] The description field may be more generic, such as “Web break at transfer to first imprinting fabric” so that it may be applicable to more than one machine. It may also be simply “Web break” if the Delay Code were accompanied with additional information in the Delay Table to specify where the break occurred.

[0121] Additional output and support tables may be used to link the information in the Delay Table to other production data. A waste table may show waste-related information, similar to that of Table 1 but using parameters pertaining to waste. A material usage table may indicate what raw materials were used and in what quantities for each period of production. A Grade Shift table is an output table that may provide basic production information for the grade being produced. For example, an exemplary Grade Shift Table associated with the Delay Table of Table 1 is shown in Table 3. The data shown indicate when the grade shift began, what the machine was, which crew of employees ran the machine, how many rolls of product were produced for shipping, how many rolls were discarded as waste, and what the product was (e.g., white facial tissue according to recipe C2). A link between Table 1 and Table 3 is possible by means of the Grade Shift Start field, which may serve as pointer in the Delay Table to additional information in the Grade Shift Table. The Grade Shift Start value may be used combination with the machine reference field to serve as a pointer if multiple machines may be considered. Thus, by looking up the entry in the Grade Shift Table having the same Grade Shift Start field as that recorded for an event in a delay table, details of production information associated with the delay may be obtained. 3

[0122] PIPE data may also provide a continuous time series of machine state information to show machine status and history before an alarm event occurs, or to allow tracking of the long-term effects of a process modification on machine efficiency and modes of operation. The machine state may be described by a Machine State record in a database, which may include labels such as startup, shutdown, thread, acceleration, full speed, etc. The state of operation may further be described by information from a statistical process control program, which may generate associated data to indicate at any point in time if the machine was operating within specifications or whether it was out of control on one or more variables. Analysis of machine state data and other measures of productivity in combination with statistical process control data may be a rich source of information about the interaction between statistical quality control practices and machine productivity.

[0123] Similar fields may be used to describe other events such as waste, slow downs, process control excursions, quality problems, remedial actions to correct a delay, or other events. For example, an incident of waste on a machine (e.g., the culling of one or more products) may be described by a waste code may be associated with a description field, a PLC address, a section code, a delay trigger field, a machine type field, and an alarm source.

[0124] If desired, PIPE data may be stored on a server at each plant and be periodically “rolled up” to the corporate PIPE database. Support tables or maps may be kept identical across a sector or across the corporation or other unit. Thus, the support tables or a subset of the support tables may be maintained at the corporate level and provided to individual plants to ensure uniformity. The plants may then ensure that the output of their PIPE systems is adapted to comply with the applicable support tables.

Combining Human and Machine Input

[0125] To fully document the reasons for delay and optionally the corrective actions taken, the PIPE system allows human input to supplement machine-generated data for any event. Human-machine interfaces are often used for the entry of such data. Human input may be required to explain what the event was or to identify planned corrective action. Human input may also be required to validate a possible error state detected by a control system, or to select one of several automatic responses to a delay problem.

[0126] Human input is subject to many forms of error. For example, a shutdown may be caused by a web break that normally takes 5 minutes to correct, after which time an operator may elect to initiate routine machine maintenance for several hours. The source of the several-hour delay, when queried by the PIPE system, may be entered simply as “web break,” leading to a greatly inflated apparent cost of web breaks in the financial reporting for the shift, day, or other period of time in which the web break occurred. More accurate financial reporting may be done by ensuring that the machine delay is properly identified, such as ascribing five minutes of the delay to a web break and the remaining delay time to routine maintenance.

[0127] For example, if a tissue machine stops due to a web break, which normally causes a delay of several minutes, an operator may choose to prolong the down time for other scheduled maintenance that might require an hour. If the act of maintenance is not properly recorded, the hour of delay may be falsely credited to a web break. After a predetermined period of time or after start-up, the PIPE system may then alert the operator that the down time may have been incorrectly attributed to a web break, and query if it was prolonged for other reasons, which may then be recorded. Seemingly erroneous entries may also result in e-mail or other alerts directed to a supervisor for review and corrective action or further verification, if needed. Additional operator or management input optionally may be required before the machine is allowed to start again to ensure that proper documentation is provided

[0128] To validate human input in this manner, an expert system may be used. The expert system may include a simple table of rules and responses to deal with common problems, or may be a more sophisticated fuzzy logic expert system optionally coupled with a neural network that learns over time how the process should perform and what conditions are anomalies.

[0129] Though fuzzy logic and neural networks may be powerful tools in data mining a PIPE database, it is to be understood that any known statistical or mathematical technique may be applied to determine correlations, find optimum process conditions, predict instabilities or runnability problems, and the like. Such methods include statistical analysis such as regression or time-series analysis, signal processing techniques such as autocorrelation analysis, etc.

[0130] The expert system may be an intelligent agent to automatically check data integrity as it is recorded in the database, adapted to tag the record for human intervention if the data was suspect. If a data record violated a set of particular rules or was determined to be a statistical anomaly, the agent may flag the record and send e-mail or other communications to appropriate people for intervention. If the record was found to be in error, it may be manually corrected; if the record was correct, a tag may be marked in the database to signal to the agent that it had been checked and verified for accuracy.

[0131] The agent may be intelligent in two aspects. First, human experts may impart their learning to the agent through a fuzzy-rule-based inference system. There are many types of errors in a machine process log that humans may quickly and easily detect upon inspection. For example, a machine that made product during a particular day may report an average machine speed of zero due to a recording error. A person reviewing this record may easily spot this inconsistency. A list of known errors and inconsistencies would be compiled into fuzzy if-then rules, and the agent may automatically navigate a large amount of data and check the data using the expert-based rules. Second, the agent may use a neural network to learn patterns in the data. Deviations from learned patterns may be flagged as anomalies. The neural network may be trained with historical data and may be re-trained after a given time period to be updated with the most current process information.

Integration of Data for Multiple Systems

[0132] PIPE data from multiple machines or plants may be integrated and summarized in a common display or report. Database results from multiple sources may also be sorted or searched in any way desired, such as sorting waste data by geographical region and machine type, or searching for plants of a certain kind having waste delays in the upper quartile.

[0133] When PIPE is applied to multiple machines or plants for financial reporting of a plant, sector, or other business unit, there may be a need to obtain useful information from a variety of control systems or hardware and software systems. One useful method for establishing communication and common standards between multiple machines and control systems is the use of maps that identify the relationship between parameters required by PIPE and the data structures employed by the diverse systems communicating with PIPE. Map definitions (e.g., support tables) may be established by a central supervisor and then downloaded to the respective business units or plants, to ensure that the data transmitted from the machines at the plants is in the proper fields and format.

Feed-Forward Control Systems

[0134] When a raw material for use in a process was produced under a PIPE system of the present invention, electronic data in time series form about production defects may be available that may be of value for a process control system. For example, during production of a roll of cover material used in feminine care products, a PIPE system may record that defects were observed at two positions in the roll (e.g., at 210 meters and 318 meters within a roll of material having a total length of 500 meters). The defects may have been associated with a web break and represent the location of splices, or they may have been holes or color defects that did not result in machine delay but were detected as quality problems that may or should result in waste during subsequent manufacturing. The information about the nature of the defects and their location in the roll is sent to the machine that subsequently processes the roll as a raw material. A feed forward control system then allows the machine to anticipate the problem areas in the roll as they are about to be fed into the machine or any of its unit operations. The machine may, in response to the problem supplied by the PIPE system for the raw material, slow down or invoke a cull to remove potentially defective product or initiate other compensating action. Thus, waste may be predicted at an early stage and the cause may be properly identified when the material is culled. Through anticipating the problem, the impact of the defects in the raw material on runnability and productivity of the machine may be reduced (e.g., web breaks may be avoided, or other machine problems may be averted), while quality of the final product is improved.

[0135] Basic information identifying the raw materials used in production may be supplemented with detailed information from an electronic certificate of analysis or other information accessible, for example, via a license plate system, described hereafter.

[0136] In one embodiment, the feed-forward system employs information obtained from a subset of PIPE specifically engineered to track raw materials (e.g., STORM—System for Tracking Online Raw Materials). STORM enables detailed data about the production history of a material to be generated during production and stored.

[0137] In another example, in producing a roll of tissue, the STORM system may provide the quality attributes of the tissue, including a record of measured basis weight from a beta-radiation-based scanner or other means as a function of position in the roll, and perhaps a record of optically detected web defects in the roll, again as a function of position in the roll (distance from the end of the roll). The tissue may then be slitted and converted into multiple smaller rolls for use in a diaper mill, for example. Each slitted roll may have an electronic file associated with it indicating the basis weight and presence of defects as a function of roll position. When the roll is received at the mill, this information may be accessed by scanning a bar code to obtain an identifier that links to the data file. The raw material data file is accessed by process control systems for the machine. The system may then anticipate that a defect may exist at, for example, 47 meters into the roll. The machine speed may be momentarily reduced to either prevent a web break or to allow the defective portion to more easily be removed, after which full speed may be resume. If a portion of the roll has inadequate basis weight, that portion may be automatically spliced out by the machine or with the assistance of human operators, following directions electronically conveyed in response to the process control system of the present invention.

[0138] In general, STORM or PIPE data for one component (e.g., a raw material or intermediate product), generated by any of the machines used in the production of that component, may be communicated to other machines that use the component in manufacturing. The data may be used to verify quality of the incoming components, to make adjustments to the component (e.g., removing portions with quality problems), or to make adjustments to the machines using the component. In the latter case, feed forward process control technology may be applied to adjust the machine in anticipation of changes in the component. Other suitable process control strategies may be used as well.

[0139] In one embodiment, the improved system may access multiple databases pertaining to the raw material using the “license plate” described more fully hereafter, in which a “license plate” bar code or other identifier on the material permits access to multiple databases of information pertaining to the materials. In other words, the license plate may be a pointer to multiple sources of data. The databases may have a common format for easy access to and display of information in a form usable by the manufacturer.

[0140] The problem that is anticipated need not be an absolutely verified problem, such as an observed defect, but may be one that is only probable or possible based on a detected event that is known to be associated with a quality problem (i.e., a deviation in the properties of the raw material). For example, based on past experience with manufacturing a roll good on a first machine, it may have been determined through data mining or other procedures that after the first machine goes down, the portion of a web being produced that was in contact with a heated section of the first machine when the first machine went down may have a 25% probability of being thermally damaged during the down time, resulting in an increased likelihood that the raw material may fail in a subsequent manufacturing process on a second machine. The process conditions for the subsequent manufacturing process employing the raw material may be temporarily adjusted near the time when the portion of the web in question is unwound and enters the second machine, in order to decrease the probability that a web break or other failure will occur. In this manner, the likelihood of waste or delay or a quality problem can be decreased in a manufacturing process by temporarily adjusting process conditions responsive to previously obtained manufacturing information about a raw material, wherein the manufacturing information is interpreted to indicate an increased probability of a quality problem or waste or delay if normal process conditions are maintained during manufacture of a product.

[0141] Temporary adjustments to process conditions that anticipate possible manufacturing problems due to deviations in the properties of a raw material may be done when a batch or unit of the raw material is “sequentially trackable,” meaning that data are available relating one or more identifiable portions of a raw material (each portion comprising substantially less than 100% of the raw material in this case, such as 10% or less, or 2% or less) to manufacturing or material property information about the one or more portions of the raw material. Sequentially trackable raw materials are typically produced in a known sequence and, in a subsequent manufacturing process, supplied in a known sequence that can be related to the sequence of manufacturing. For example, roll goods can be sequentially trackable. When roll goods are used as a raw material, they are generally used in the reverse sequence in which they were manufactured (the last portion made is the first portion used; though in some applications, the first portion made may be the first portion used). Webs in any form may be sequentially trackable, including cut stacks of web materials, festoons, and the like. Materials that are provided in a string or other fixed sequence may also be sequentially trackable (e.g., medications sealed in pouches along a continuous web of aluminum foil). Material in bales, or loose powders or liquids in tanks, or vats generally is not sequentially trackable because the material within the batch becomes mixed after production.

[0142] A batch of sequentially trackable raw material may be associated with a single identification code (e.g., a single barcode or single electronic product code from a smart tag for an entire roll of materials), but the identification code may provide access to sequential manufacturing information such as event data in a PIPE database and/or continuously monitored process data from any number of process sensors and other control devices, and the sequential information may then be associated with various portions of the raw material (e.g., identifying a basis weight deviation in a roll good at a specified location in the roll).

[0143] A bill of materials for the manufacturing of a product may specify or may be consulted to help specify what actions should be taken for an anticipated temporary deviation in the properties of a raw material. Some deviations of the properties may still permit production of the product within the targeted specifications for that product, while other deviations may require culling of the products affected by the deviation in the properties of the raw material, or may require rejecting the affected portion of the raw material so that it does not enter the machine.

[0144] The above-mentioned feed-forward system or system of machine-to-machine communication regarding raw materials and their use in products need not apply to components produced from a single manufacturer, but may also apply to any raw material used by a manufacturer, wherein data generated by vendors of the raw materials are obtained and stored for use in manufacturing systems according to the present invention.

[0145] Accessing the data may require a connection across a network involving vendor computers. Alternatively, the vendor may electronically supply data to a common manufacturer database that may later be accessed through the license plate system.

[0146] In another embodiment, probabilistic or “fuzzy” information may be used for improved feed-forward control. The probabilistic information may be obtained by correlations of past machine performance or product quality as a function of PIPE data or raw material data for a component used in a process. The correlations may indicate that the risk of a delay or quality problem may be greater unless machine conditions are modified, or may indicate mixed risks and opportunities that may be weighed for the greatest expected economic return. For example, in the production of a diaper, correlations of past quality results could indicate that a statistically significant increase in consumer complaints about adhesive failure occurred 35% of the time when, even though all product specifications were met, when a batch of hot melt adhesive having a molecular weight slightly below target was used, suggesting that an increased amount of adhesive may need to be used to secure a component, but at a higher cost. However, the PIPE data may also show that increasing the application level of adhesive historically results in a 10% increase in down time due to adhesive nozzle plugging, and may indicate that machine runnability improved on the average when the low molecular weight adhesive was used. These factors may be associated with their expected costs and optimized run conditions may be suggested or automatically implemented, optionally subject to human supervision. In general, the information used for feed-forward control need not be data directly describing quality problems or other waste and delay information pertaining to components of a process, but may be information inferred from past PIPE and other data, such as probabilistic predictions obtained by correlations or neural network mining of the data to suggest opportunities to be obtained (increased machine speed, for example) or possible problems to be avoided or probabilities of various costs and problems to be weighed in optimizing process conditions as the associated materials enter the system.

[0147] Data stored in process information databases such as PIPE enable the determination of non-obvious cause and effect relationships between manufacturing events. While many events are related in a trivial manner (e.g., a raw material splice will cause some product