Expert Decision System for Robot Selection

Introduction. Over the past two decades, an upward trend has been observed in the use of industrial robots because of quality, productivity, flexibility, health, and safety reasons. Robots can help manufacturers in virtually every industry to stay globally competitive. Robots can be programmed to keep a constant speed and a predetermined quality when performing a task repetitively. Robots can manage to work under conditions hazardous to human health, such as excessive heat or noise, heavy load, and toxic gases.

Therefore, manufacturers prefer to use robots in many industrial applications in which repetitive, difficult, or hazardous tasks need to be performed, such as spot welding, arc welding, machine loading, die casting, forging, plastic molding, spray painting, materials handling, assembly, and inspection. However, a wide selection of robot alternatives and the large number of performance attributes result in a major problem for potential robot users when deciding which robot to purchase.

In the absence of a robust decision aid, the robot selection decisions can be based on the recommendations of robot vendors, the recommendations of an expert hired for performing the evaluation task, or the user’s own experience. The recommendations of robot vendors may be biased because they have an inherent interest in selling their product. Basing robot selection decisions on expert advice may be highly costly because experts usually charge considerable amounts for their valuations. Relying on personal experience generally results in selecting a robot with which the user is most familiar, ignoring other crucial factors.

A robot that has the capability of affording heavy load at high speed, as well as good repeatability and accuracy, will contribute positively to the productivity and flexibility of the manufacturing process, which are of high importance when competitive market forces require the introduction of new products into the market. When product design changes need to be made repeatedly, owning a high performing robot will avoid replacement or modification. Many studies reported in the literature address the development of a robust decision tool that enables the potential robot user to select a high-performing robot.

Although it is usually assumed that the specified performance parameters are mutually independent, in general performance parameters provided by robot vendors are not achievable simultaneously. For instance, Offodile and Ugwu (1) reported that for a Unimation PUMA 560 robot manufacturer-specified repeatability deteriorated as the speed increased beyond 50% of the status speed and the weight increased beyond 0.91 kg. Furthermore, it is very difficult to determine the functional relationship between these parameters; thus, making this assumption introduces a risk of selecting a robot that might fail to provide the required performance.

In this article, integration of an expert system and a decision-support system is proposed to enhance the quality and efficiency of the robot selection process. Rather than seeking the advice of an expert or group of experts, the user may consult an expert system to determine the key performance attributes and a short list of acceptable robot alternatives. Expert systems are used to perform a wide variety of complicated tasks that can only be performed by highly trained human experts.

Although the problem domain for robot selection may be considered as narrow, which fits the expert system structure, it is also complex and requires a multicriteria decision making (MCDM) procedure. An expert system needs to access the database of a decision-making system to gather factual knowledge. Furthermore, the judgmental data obtained from experts can be incorporated into a decision-making system through an expert system.

A MCDM methodology is required in the expert decision system framework because the expert system provides a subset of robot alternatives based on the technical aspects, and an appropriate MCDM technique evaluates the short list of alternatives and determines the robot that best fits the user requirements.

The proposed decision-support system that employs quality function deployment and fuzzy linear regression integrates user demands with key specifications of robots. The developed decision-making approach has advantages compared with the techniques previously proposed for robot selection. Statistical methods, such as ordinary least squares, are based on determining a meaningful statistical relationship between performance parameters, which is difficult to achieve in practice and the problem aggravates for a small number of candidate robots.

Multiattribute decision-making techniques such as multiattribute utility theory (MAUT), analytic hierarchy process (AHP), technique for order preference by similarity to ideal solution (TOPSIS), assume that preferential independence of the performance parameters hold. However, that is a very critical assumption that usually fails to hold in real-world applications.

Although fuzzy multiattribute decisionmaking techniques enable qualitative attributes to be taken into account in an effective manner, they suffer from the same shortcoming as the other multiattribute decision-making techniques. Data envelopment analysis (DEA) does not require the preferential independence assumption of performance parameters. However, DEA assumes that every characteristic defined as output is related to every input. Profiling may be considered as an alternative to tackle that problem, but gathering the efficiency scores into a single efficiency score may also be problematic (2).