Control of timber strength grading machines

Procedures for the control of timber strength grading machines have been established, including both machine control and output control. For the machine control so called control planks have been developed which enables a dynamic control of the whole grading system, i.e. the grading machine as well as infeed and outfeed equipment. Some generic control plank designs have been developed which may be applicable in the different machine types available on the market. Earlier studies have indicated that the currently used procedure for output control is too slow to show an out-of-control. Simulations have been carried out which show that the method is very sensitive to the actual strength distribution of the graded material. Using a log-normal distribution, which had the best goodness of fit to actual strength data, instead of a normal distribution, resulted in a much faster registration of a grading not fulfilling the requirements as well as a less frequent false alarms. INTRODUCTION Two different control systems are presently available for strength grading machines in Europe, output control and machine control. The output control system is based on a production control, i.e. samples are picked out from the production and tested with respect to the mechanical properties. This system is suitable for industries grading particular sizes, species and grades in repeated production runs of at least a duration of one working shift. The machine control system was developed in Europe in the late 1960's because of the large number of sizes, species and grades used. It was too expensive and labour-intensive to perform the proof-loading control several times per working shift. The machine control system relies on a more strictly control of the grading machine during the assessment, and on machine settings which are set to a certain value based on an extensive test procedure during the assessment. Studies have showed that the output control procedure may, at least under certain conditions, be too slow to register outof-control (Leicester and Breitinger 1986). Therefore it is necessary to further investigate, and develop, the output control process to ensure the same level of safety as in the machine control process. The machine control process relies heavily on the initial assessment of the machine. It is thereafter assumed that all machines of the same type behaves in a similar manner as the machines included in the assessment testing. There are today no common rules on how to reach this, leading to different practices in different countries. In a machine control system the control and calibration of the machine is essential. Due to the European harmonisation, and the development of new standards, a project was launched in 1998 in order to establish new routines for the output control system and to develop control procedures for the machine control system. The present standard on machine strength grading, EN 519:1995, is undergoing a revision in order to cover the new machine types as well as to ensure an approximately equal safety level for both machine and output control. MACHINE CONTROL CONTROL PLANKS In a machine controlled system all machines of the same type shall have a similar performance. This cannot be controlled only by a static calibration of the sensors in the machine. The whole installation must be controlled, and the control should be made when the machine is running. The work has therefore been focused on the development of designs of control 1 SP-Fire technology, Box 857, S-501 15 Borås, Sweden planks and their procedures of use. A control plank shall have the same response as normal timber when passed through the grading system, i.e. for a bending type machine the control plank should have the same dynamic behaviour and stiffness as timber. Since new machine types have reached the market, with new measuring techniques, it is not possible to have only one design of control plank. It is likely that each machine type will have a special control plank designed for this specific measurement technique or combination of techniques. Therefore it is difficult to standardise the design of the control plank. The control plank shall be used to check that the dynamic performance of the grading machine does not change over time. This means that the evidence of the machines performance is the repeatability of the response of the control plank. The control plank must therefore be constructed to have a known and repeatable indicating property/properties, i.e. the property/properties determined or measured by the machine used to establish the strength class of the timber. This may be obtained with many different plank designs, but it is not necessarily that a control plank designed for a X-ray machine works for a bending type machine. Several different materials and control plank designs have been studied on different machine types. The performance criteria used when testing the planks have been the repeatability of the measured properties. The basic criteria is that the indicating property shall be within ± 3 % of the stated value for a minimum of 500 passes through the machine, and that the effect of changes in environmental conditions on the indicating properties is less than 2 %. Of the materials and designs studied four different generic planks have showed a good performance. All control planks were designed to have an overall density and dimensions similar to that of wood. The measured property/properties of the control plank, i.e. the indicating property, shall be well defined. It would not be possible to use normal timber in for example a machine based on X-rays. In normal timber the density varies too much, both within the clear wood portion and between clear wood and knots, and thus this material would be difficult to be used for control of the machine. Instead a generic plank of a homogeneous material with inserts of metal discs would give a distinct reading of the changes in absorption of the radiation. For other machine types, such as a machine based on vibration, timber planks could be a possible alternative. Table 1 gives some generic plank designs and the machine types they could be used in. Figure 1 shows an example of the design of a control plank made of aluminium tubes. The only difference between the glass fibre tube plank and the aluminium plank is the material of the tubes. The plank called timber with coating is a normal plank of timber, or laminated veneer lumber (LVL), which is coated in order to give a protection against varying climate conditions. The foam plank consists of a core of an extruded foam with a density similar to that of wood. An outer layer of carbon or glass fibre reinforced epoxy give the plank its stiffness. Table 1. Machine types for which the generic plank designs can be used. Machine type Aluminium tubes Glass fibre epoxy tubes Timber with coating Extruded foam Bending type machines X X X Vibration wave propagation X X Optical inspection X X X Radiation emitting machines X X X Figure 1. Example of a generic control plank made of aluminium tubes. The use of the control planks should be simple. The control plank should be passed through the grading system at normal production speed and the indicating property shall be recorded in the same way as for normal timber. The interpretation of the results and the performance criteria may vary between different machine types. The recorded value of the indicating property shall be within certain limits established by the machine manufacturer and approved by an certification body. If the first run fails the plank shall be passed through the machine a second time and if also this run is a failure the machine must be recalibrated. OUTPUT CONTROL CUSUM CHART The currently used statistical process control chart, the cumulative sum control process (CUSUM), has shown to be too slow to indicate an out-of-control (Leicester and Breitinger 1986). A literature survey was carried out in order to find alternative methods, but the results from the study was that the CUSUM control procedure is the best suited method for output control of machine strength graded timber. The question was thus if the currently used method in some way could be modified in order to ensure a reliable output control with the same level of safety as in machine control. In output control a certain number of specimens of each grade and dimension is selected a certain number of times per working shift. Normally 5 pieces of timber are selected two times per shift for each grade and dimension. These specimens are proof loaded to a level equal to 0 96 , , ⋅ ⋅ k f h m k where kh is the size factor defined in EN 384:1995 and fm,k is the characteristic bending strength for 150 mm depth. The result from the proof loading is whether the specimen has a strength higher than the applied stress level or not, as well as a value on the modulus of elasticity. The CUSUM control chart operates with two separate charts, an attributes chart and a variables chart. In the attributes chart the number of pieces that fails when proof loaded is used, while in the variables chart the actual value of the modulus of elasticity is used. Simulations of the output control were carried out using both Monte Carlo and Markovian approaches. When comparing these two approaches the average run length to show out-of-control or false alarm was about the same for the two methods. The input data was based on material properties and their statistical variations from studies on Norway spruce (Picea abies) from the Nordic countries. Glass fibre and epoxy resin coating. Aluminium tubes Metal discs may be inserted into the tubes for radiation type machines. Applied artificial knots of varying contrast and size for optical systems, where applicable.