Summary A variable-frequency drive (VFD) used to change the speed of beam-pumpmotors was tested on seven wells to control the rate of oil pumping. Increasedpower consumption was observed, but on certain wells, pumping. Increased powerconsumption was observed, but on certain wells, increased production rates canresult in overall benefits. Introduction The simplicity of the beam-pump unit, which is a maintenance and operatingadvantage, presents a problem when it becomes necessary to provide moreaccurate control of the pumping rate than previously possible with simple timeclocks for on/off operation. The previously possible with simple time clocksfor on/off operation. The increased use of fluid injection for EOR, whichcauses frequent changes in reservoir production rates, and reduced manpoweravailable in the field to make manual changes are making automatic control ofpumping rates highly desirable. A number of methods, such as time clocks and pumpoff controllers, are usedto match pump performance to well production rates. These devices detect whenthe pump is empty and shut down the pump for a fixed period of time. An MSDOS-based controller combined with a VFD for motor speed control was developedand field tested on several west Texas leases. The beam-pump wellhead controller varies the speed of the electric motordriving the pump by detection of decreasing fluid in the pump. Fig. 1 shows thegeneral installation. With this method of pump. Fig. 1 shows the generalinstallation. With this method of control, the electric motor is never turnedoff but varies in operating speed. Because the unit is not stopped every 15 to30 minutes, equipment maintenance is expected to be reduced. When the pump isoperated continuously with the fluid level at the bottom of the well, thereservoir can produce at a slightly higher rate, which is the major incentivefor use of this control system. Background Automatic beam-pump control can be separated into two major activities:detecting when the pump rate is too high or too low andcorrecting the rateto eliminate the problem. Most of today's control methods use the polished-rodload or the electric-motor load or speed for detection. These loads change whenthere is reduced liquid in the subsurface pump, indicating overpumping. Thepolished-rod load is usually determined directly with a load cell polished-rodload is usually determined directly with a load cell on the polished rod or bya strain gauge that measures the deflection of the walking beam. The electricpower can be measured directly, or motor speed can be used to indicate the loadon the motor. Most control systems in use today correct the pumping rate byshutting down the pumping unit for a period of time so that the average pumpingrate matches the reservoir production rate. Lea presented an excellent surveyof control methods. presented an excellent survey of control methods. For themaximum reservoir production rate, the ultimate objective of all pumpoffsensing methods is to determine the fluid level in the well and to keep it atthe perforation level. A number of different methods have been used to measurefluid level directly down the annulus, but none have been used widely forpumpoff control. A significant advantage of such a method over others is thatthe pump operates while filled with liquid; most other methods require pumpoperates while filled with liquid; most other methods require some gas in thepump for control purposes. A popular method of load sensing measures the polished-rod load at thesurface as an indicator of downhole pump filling. A load cell mounted betweenthe carrier bar and polished-rod clamp yields the most accurate measurement ofdownhole rod load. In most cases, use of the rod wave equation can givereasonably accurate pump loads from the surface data. Because installation andmaintenance of a load cell on the heavily loaded polished rod are difficult inthe field, some systems use a strain gauge mounted on the top of the walkingbeam and measure the deflection of the beam as the load varies. Although not absolutely necessary for field control, most of theabove-mentioned systems include a method of determining the position of thepolished rod during the stroke. Potentiometers have position of the polishedrod during the stroke. Potentiometers have been attached to the beam or to thecarrier bar to indicate stroke position. Other approximations to the rod motionhave also been position. Other approximations to the rod motion have also beenused. Most of the rod-load detection systems use the rod load at the beginning ofthe downstroke as an indicator of pumpoff occurrence in the well. Fig. 2illustrates a normal dynamometer load card showing polished-rod load on theupstroke and downstroke with the pump full of liquid. If the pump contains gas, the rods will not be supported by the liquid normally in the pump and will fallfaster and exert a greater downward force on the polished rod. When the pumpbecomes partially filled with gas, indicating that gas is entering the pump, the dynamometer card is changed most dramatically on the pump, the dynamometercard is changed most dramatically on the beginning of the downstroke (Fig. 3). This increase in load has been used by many pumpoff controllers to indicateboth pumpoff and the need to shut down the pump until more liquid is enteringit. It is possible to use either the load at a single point on the downstrokeor the average load, work, or power over some selected portion of thedownstroke. portion of the downstroke. Variable-Speed Pumping With the increasing reliability and lower cost of solid-state electronics, it is becoming feasible to use VFD's to change the speed of electric motorscontinuously and automatically. VFD's are used throughout the industry tocontrol water-injection pumps and electric submersible oil pumps. Although a VFD control system is expected to cost more than conventional pumpoff controlsystems, the benefits are anticipated to be numerous. Continuous pumping fromthe level of the perforations will increase production from the reservoir, asshown later. Elimination of the starting and stopping of the motor and pumpunit should reduce maintenance. Continuous flow in the wellbore may reducerequirements for paraffin removal and other downhole workover operations. Thesystem may also be used as a temporary portable control unit. The capabilityfor pumping at various fixed speeds may be valuable in well testing or, after aworkover or fracturing treatment, in pumping the returned fluid rapidly. The VFD system consists of a large rectifier that converts the alternatingcurrent at 60 Hz into direct current that is used to synthesize an alternatingcurrent of any frequency from nominally to 80 Hz. Skinner described a methodfor using a VFD to control beam-pumping speed with the electric powerconsumption of the motor as an indicator of a partially filled pump. Some pumpsuse a VFD to reduce upstroke speed and to partially filled pump. Some pumps usea VFD to reduce upstroke speed and to increase downstroke speed, thus reducingpeak loads and permitting higher beam-pumping speeds in heavy oil production. With the system described here, it is possible to operate continuously with thefluid level pumped down to the pump without turning off the motor. This ispumped down to the pump without turning off the motor. This is called"pumpdown," instead of pumpoff, control.

The oil industry, spurred by the need for cost efficiency and the loss of experts in technical areas, has begun to apply Artificial Intelligence or Expert systems. This paper details the use of a pattern recognition scheme using Fourier Descriptors and a Neural Network model to develop an Artificial Intelli?ence system to analyze and classify beam pumping unit dynamometer diagrams. INTRODUCTION Background, Purpose and Method of StudyPattern recognition has, for nearly 30 years, had a high degree of academic and industrial research interest. This can be attested by observing the large volume of literature available on this subject. Though primarily utilized in communication and signal analysis it is also being applied in a number of other disciplines and industries. This technology is fast moving from research oriented technology to that of being applied throughout our society. The use of computer algorithms for pattern recognition is being used to provide faster and more economical analysis of procedures that are nornlally done by people. Typically human analyses are costly, time consuming, inconsistent, and error prone. The use of computer pattern recognition schemes can be used as stand-alone or in conjunction vvith other Artificial Intelligence (AI) systems. Though far from the ability of the human mind, these systems are accomplishing tasks that have historically needed the judgenlent and intuitive insight of the human brain. A few applications of pattern recognition schemes recently re-1 References and figures at end of paper 349 ported in the literature are : To Decipher medical problems in electrocardiograms [1,2], electroencephlogram, and blood pressure wave analysis [2]. Others mention classifying chromosomes, and interpretation of chest X-rays [3].To recognize parts and assure quality control in manufacturing and robotics [4] To classify objects in a scene [5] To classify archaeological artifacts [6] To recognize the written word In different languages [7,8,9,10,11] To map LANDSAT images [12]The discrimination of modulated signals [13] Brill [9] gives a good summarized review of the use and historical evolution of this subject to his respective time.

Piezoresistive soft composite materials exhibit a change in resistance when undergoing deformation. This combined with their optical, thermal, and mechanical properties makes these composites good candidates for force sensors. Tactile force sensors have long been studied for applications in healthcare, robot–human interactions, and displacement monitoring. The main goal in this work is to characterize a soft piezoresistive layer in both tension and compression to enable a model system for a piezoresistive tactile force sensor and a characterization platform. However, the mechanisms by which these composites exhibit piezoresistivity are complex and must be characterized before use not only in bulk but at the exact locations where contact is expected. In this paper, a cantilevered beam is proposed as a base-mounted force-sensing mechanism. This mechanism allows for characterization of the composites at multiple locations across the sample using a two-probe technique. Multiwalled carbon nanotubes (MWCNTs) are mixed by weight with a soft polyurethane in 15, 16, and 17 wt. % concentrations. Because the elastic modulus of the piezoresistive layer is not known, indentation tests using Hertz theory and numerical calculations are used to simulate the effective elastic modulus and average strain. These results are then compared with the experimental stress results. In general, these tests show a greater sensitivity in tension than in compression. However, the difference lessens as the concentration increases. A linear fit is applied to the ΔR/R versus strain graphs to calculate the gauge factors. Each sensor exhibits a positive and negative gauge factor over two different ranges. ΔR/R versus strain graphs for tension and compression show gauge factors between −19 and 24 with the range decreasing with increasing MWCNT percentage.

Flexible piezoresistive films, such as, carbon black/polydimethylsiloxane (C‐PDMS) composites, are often used as skin analogs and integrated into complex array sensors for tactile sensing. The uniformity of the sensor characteristics heavily depends on the homogeneity of the composite. Therefore, the ability to locally characterize a film that will be integrated into a complex force sensor could be critical. Here, a method to characterize the local sensitivity of flexible piezoresistive films is presented. Using a conductive sphere, which was chosen over a flat probe to eliminate misalignment issues, the surface of a thin film composite is indented to characterize the change in resistivity in terms of average strain. Experiments were performed with 15 and 18 wt% carbon black C‐PDMS films of varying thickness. The contact radius of the probe with the piezoresistive film was estimated using the Johnson‐Roberts‐Kendall contact theory. Theoretical contact area estimates were found to agree with contact radius measurements carried out using optically transparent PDMS films observed through an optical microscope. Results show that C‐PDMS with 15 wt% carbon black exhibit a higher rate if change of resistivity and gauge factor than films of same thickness with 18 wt% carbon black. On the other hand, thicker films exhibit higher gauge factors for the two tested carbon black contents. Tests carried out at multiple locations yielded consistent sensitivity values, making these types of composites suitable for array type force sensors.