Examples include the interpretation images

Chapter eleven

Systems for interpretation and diagnosis

PROSPECTOR
This system interpreted geological data and made recommendations of suitable sites for mineral prospecting. The system made use of Bayesian updating as a means of handling uncertainty (see Chapter 3).

DENDRAL
This system interpreted mass-spectrometry data, and was notable for its use of a three-phase approach to the problem:

A large number of intelligent systems has been produced more recently, using many different techniques, to tackle a range of diagnosis and interpretation problems in science, technology, and engineering. Rule-based diagnostic systems have been applied to power plants [1], electronic circuits [2], furnaces [3], an oxygen generator for use on Mars [4], and batteries in the Hubble space telescope [5]. Bayesian updating and fuzzy logic have been used in nuclear power generation [6] and automobile assembly [7], respectively. Neural networks have been used for pump diagnosis [8], and a neural network–rule-based system hybrid has been applied to the diagnosis of electrical discharge machines [9]. One of the most important techniques for diagnosis is case-based reasoning (CBR), described in Chapter 6. CBR has been used to diagnose faults in electronic circuits [10], emergency battery backup systems [11], and software [12]. We will also see in this chapter the importance of models of physical systems such as power plants [13]. Applications of intelligent systems for the more general problem of interpretation include the interpretation of drawings [14], seismic data [15], optical spectra [16], and ultrasonic images [17]. The last is a hybrid system, described in a detailed case study in Section 11.5.

11.2 Deduction and abduction for diagnosis